International Journal of Medical Engineering and Informatics (127 papers in press)

Regular Issues

• Open Access

  Heart disease detection using 1D transformer network: case of ECG signals and clinical data ( Free Full-text Access )

  by Amal Miloud Aouidate 
  Abstract: Early prediction of cardiovascular disease remains a critical public health challenge. This paper presents a 1D Transformer-based architecture for classifying patients as healthy or suffering from heart disease using ECG signals and clinical data. The model is evaluated on three benchmark databases: Cleveland Heart Disease (tabular data, 303 patients), PTB (ECG signals, 290 patients), and MIT-BIH (multi-class arrhythmia, 48 patients). Our approach achieves accuracies of 88.5% +- 1.2 (Cleveland), 94.2% +- 1.5 (PTB), and 89.2% +- 1.8 (MIT-BIH). The PTB dataset shows strong discriminative performance (AUC = 0.97), while Cleveland achieves AUC = 0.94. For MIT-BIH, class imbalance mitigation improves macro F1-score from 0.47 to 0.69. These results demonstrate the effectiveness of attention mechanisms for modelling biomedical time series, while highlighting the critical importance of proper validation protocols and imbalance mitigation for clinical applications.
  Keywords: early prediction; cardiovascular disease; 1D transformer model; ECG classification; heart disease detection; Heart Cleveland database; PTB database; Mitbih database; supervised training; attention mechanisms; biomedical time series.
  DOI: 10.1504/IJMEI.2026.10078275
   
  
• Tri-partition-based b-value optimisation for intravoxel incoherent motion magnetic resonance imaging of brain  
  by Jini Raju, C. Ushadevi Amma, Ansamma John, V. Jineesh 
  Abstract: Intravoxel incoherent motion (IVIM)-based magnetic resonance imaging (MRI) technique allows the simultaneous estimation of perfusion and diffusion without the use of contrast agents. As the number of b-values increases, the scan time also increases, which in turn causes patient discomfort. This necessitates the reduction in the number of b-values (b-value count) and the optimisation of absolute b-values that quantify both the perfusion and diffusion effects accurately. The two partition approach of biexponential model fails to consider the b-value regions where both the perfusion parameters have significance. The proposed work explores the possibility of finding minimal and optimal set of b-values using the images corresponding to a set of 21 b-values, using random sampling-based tri-partition method, by varying b-value counts from 10 to 4. Experimental results demonstrate that appropriate selection of b-values from the three partitions generate quality parametric maps.
  Keywords: intravoxel incoherent motion; IVIM; diffusion weighted imaging; b-value optimisation; random subsampling.
  DOI: 10.1504/IJMEI.2023.10058916
   
  
• A novel hybrid-based approach for detection of skin lesion using machine learning techniques  
  by Nikhil Singh, Sachin Kumar, Shriram K. Vasudevan 
  Abstract: As a result of medical sector treatment strategies, the incidence of skin cancer has increased globally over the past few decades. It is essential to develop automatic detection systems to aid doctors in the early diagnosis of skin cancer. The automated identification of skin lesions using dermoscopic pictures remains a difficult and complex endeavour. This proposal describes a unique method for detecting skin cancer using dermoscopy pictures. In order to enhance the performance, we combined the segmentation and classification techniques with the innovative hybrid methodology fuzzy brain storming optimisation (FBSO). In two steps, we performed lesion segmentation, noise reduction, and feature extraction before submitting our dataset to machine learning algorithms for optimisation and analysis. Many characteristics, including accuracy, sensitivity, specificity, precision, F1-score, and AUROC, have been used to validate the performance of the suggested model. The experiment demonstrates that random forest’s accuracy [91.0% for dataset 1 (ISIC) and 92.5% for dataset 2 (HAM10000)] and AUROC [96.1% for dataset 1 (ISIC) and 97% for dataset 2 (HAM10000)] are superior to those of traditional models.
  Keywords: machine learning; AI; healthcare; skin cancer; medicine; melanoma.
  DOI: 10.1504/IJMEI.2023.10060493
   
  
• Prediction of COVID-19 cases with epidemiological and time series models  
  by Aman Shakya, Anita Sharma, Sanjeeb Prasad Panday, Rom Kant Pandey 
  Abstract: This work analyses the official data of coronavirus and predicts the evolution of the epidemic in Nepal. The generalised SEIR model has been applied with hybrid of ETS-ARIMA time series model for the time series analysis and predictions of evolution of COVID-19 cases. The prediction has been made for 30 days using the past data of thirteen months. The prediction made by generalised SEIR model has been corrected using two time series models, ETS and ARIMA model. The predicted error by ARIMA model is added to the prediction made by generalised SEIR model. Use of generalised SEIR model along with ETS and ARIMA model improves the time series prediction of coronavirus spread in case of Nepal as compared to the generalised SEIR model. Also, the SEIR-ETS-ARIMA model reduces the estimation error as compared to SEIRD-ARIMA model. Improvement in all quality measures, MAE, MSE, RMSE and MAPE has been observed.
  Keywords: corona; SEIR; ETS; ARIMA; SEIRD-ARIMA.
  DOI: 10.1504/IJMEI.2023.10060974
   
  
• Deep transfer learning based two-stage multi-modal depressive tendency recognition using physiological and psychophysiological signals  
  by Gaurav Kumar Gupta, Dilip Kumar Sharma 
  Abstract: Depressive tendency recognition is challenging due to the limitations of labelled depressive knowledge and diverse variations in behaviour. This paper proposes a deep transfer learning-assisted two-stage multi-modal depressive tendency recognition system (DTL-TMD) by utilising physiological signals such as audio, transcript texts, and images and EEG as psychophysiological signals. The proposed system leverages the deep transfer learning (DTL) model to address and enhance the insufficient annotated physiological and Psychophysiological signals. Finally, it fuses the outcome of two stages with the assistance of a depression level-aware attention model in the adaptive neuro-fuzzy interface system (ANFIS) to detect depressed and non-depressed subjects effectively. Experimental results demonstrate superior performance and improve the depression recognition rate.
  Keywords: depression; electroencephalogram; EEG; convolutional neural network; CNN; adaptive neuro-fuzzy interface system; ANFIS; grow net; radial basis function neural network; RBFNN; deep Boltzmann machine; DBM; deep transfer learning; DTL; whale optimisation algorithm.
  DOI: 10.1504/IJMEI.2024.10061938
   
  
• Automated adaptive gamma improvement with weighting distribution in wireless capsule endoscopy  
  by Kanika Pasrija, Kavita Mittal 
  Abstract: This study focuses on enhancing the wireless capsule endoscopy (WCE) systems capabilities in disease localisation and bleeding detection within the gastrointestinal (GI) tract. By leveraging direction of arrival (DOA)-based localisation methods, the research proposes a novel approach involving the identification of regions of interest (ROIs) along the GI tract, streamlining feature extraction and bleeding image classification while reducing computational complexity. Additionally, a new bleeding detection method is introduced, utilising the Y.I/Q colour space as an efficient alternative to the conventional YIQ model. Experimental results demonstrate the effectiveness of the proposed computer-assisted diagnostic algorithm in accurately identifying gastritis and gastrointestinal bleeding, underscoring its potential to significantly enhance WCE technology for improved disease localisation and diagnostic precision. Our experiments yielded compelling results, with our computer-assisted diagnostic algorithm achieving an accuracy rate of over 90% in accurately identifying gastritis and gastrointestinal bleeding. This significant level of accuracy underscores the potential of our approach to greatly enhance the capabilities of wireless capsule endoscopy (WCE) technology, improving disease localisation and diagnostic precision within the gastrointestinal tract.
  Keywords: gastrointestinal; computer-assisted diagnostic-procedure; wireless capsule endoscope; luminance-in phase-quadrature; region of interest; deep neural network; disease localisation; bleeding detection; patient care; medical practices; early disease detection; telemedicine.
  DOI: 10.1504/IJMEI.2024.10062053
   
  
• Disease detection system based on haemogram report using random forest algorithm  
  by Deepali K. Gaikwad 
  Abstract: The healthcare systems require vast amount of medical data that is gathered post performing numerous medical tests. Medical data provides the essential information in order to diagnose diseases early or even before they occur. The classifiers use the characteristics of the complete blood count to forecast information about potential blood illnesses in early stages, which may improve the likelihood of a cure. Machine learning builds predictive models based on prior data. The illness prediction procedure may decrease as many as fatalities and improves the condition of living for those who tract with these diseases. In this study we applied random forest classifier on blood test report dataset to identify various blood diseases like anemia, leukemia, lymphoma, sickle-cell, etc.
  Keywords: machine learning; random forest; blood components; blood diseases.
  DOI: 10.1504/IJMEI.2024.10062222
   
  
• Analysis of type-2 diabetes datasets using sampling techniques  
  by Puneeth N. Thotad, Geeta R. Bharamagoudar, Basavaraj S. Anami 
  Abstract: Diabetes is one of the significant health disorders causing mortality in a developing country like India. Using technology to treat disease can lead to insights using data mining and machine learning techniques. Patient information plays a significant role in the diagnosis and prognosis of diabetes disease for decision-making. A comparative study is carried out using three imbalanced diabetes datasets. Synthetic minority oversampling technique is applied, and datasets are balanced. Balanced datasets gave higher accuracy, recall, precision, F1-score, and area under the curve compared to imbalanced datasets. Random forest has performed well on the balanced datasets. This work is appropriate for use in the medical field to manage diabetes efficiently.
  Keywords: data mining; machine learning; sampling techniques; random forest; decision tree; diabetic complications.
  DOI: 10.1504/IJMEI.2024.10062368
   
  
• Enhanced brain structure segmentation in schizophrenia MRI using bias correction and optimisation  
  by N. Swathi, J.M. Mathana, K. Sakthidasan 
  Abstract: Exploring the structural details of schizophrenia through neuroimaging, particularly MRI, has been essential in unravelling the disorder’s underlying pathophysiology. This research addresses the challenge of accurately segmenting MR images in schizophrenia studies, contending with issues like magnetic field inhomogeneity and noise. The central objective is to refine segmentation precision by effectively estimating bias fields. Employing multiplicative intrinsic component optimisation (MICO) and non-uniformity correction via brain SUITE for bias correction, and utilising firefly and partial swarm optimisation (PSO) algorithms for gray matter and ventricle segmentation, the study identifies the optimal combination through a comparative analysis. Evaluation metrics, including structural similarity measures (SSIM), feature similarity measures (FSIM) and Accuracy, highlight the superior performance of the MICO and PSO pairing in this proposed framework.
  Keywords: schizophrenia; SZ; image segmentation; bias correction; multiplicative intrinsic component optimisation; MICO; non-uniformity correction; brain suite; firefly; PSO.
  DOI: 10.1504/IJMEI.2024.10062369
   
  
• COPD: assessment of COPD prediction through machine learning techniques  
  by Mrinal Goswami, Arpita Nath Boruah 
  Abstract: Chronic obstructive pulmonary disease (COPD) is a progressive and debilitating respiratory condition characterised by persistent airflow limitation, typically associated with chronic bronchitis and emphysema. COPD represents a significant global health burden, affecting millions of individuals worldwide. In recent years, there has been growing interest in applying machine learning techniques to various aspects of COPD management, including diagnosis, treatment optimisation, etc. This work investigates the performance of different machine learning classifiers used in COPD prediction, especially in single and ensemble classification. A detailed performance comparison among all the classifiers is also done, considering accuracy, precision, recall, and F1 score.
  Keywords: chronic obstructive pulmonary disease; COPD; machine learning; ML; classification; ensemble learning; confusion matrix.
  DOI: 10.1504/IJMEI.2024.10064464
   
  
• Monitoring liver fibrosis in chronic liver disease patients using parametric electrical impedance tomography  
  by Shimon Hury, Shimon Abboud, Oranit Cohen Ezra, Maria Lichter, Davidov Yana, Ziv Ben Ari 
  Abstract: Liver fibrosis stage is the major factor that impacts liver morbidity and mortality. Previous studies suggest that liver dielectric properties, namely its bioimpedance, can be helpful in fibrosis stage classification. In this study, the feasibility of liver fibrosis classification using parametric electrical impedance tomography (pEIT) coupled with demographic & geometric clinical data is investigated. Data from chronic liver patients was collected and used to fit a machine learning model on the task of screening healthy subjects. An accuracy score of 85% was achieved. The exclusion of pEIT measurements, resulted in drop in accuracy to 72% (p-value < 0.005), sensitivity (85% vs. 70%), specificity (84% vs. 74%) and area under the receiver operating characteristic curve (AUC) (0.86 vs. 0.83).
  Keywords: electrical impedance tomography; liver fibrosis; machine learning; logistic regression.
  DOI: 10.1504/IJMEI.2024.10068351
   
  
• Clinical applications of magnetic resonance imaging and spectroscopy technology  
  by G.S. Uthayakumar, J. Jeneetha Jebanazer, A. Prabha, T.J. Nagalakshmi 
  Abstract: Magnetic resonance imaging has been refined as a non-invasive technique due to the superior contrasts it presents in soft tissues. Recent advances in instrumentation have allowed for measurements at ultra-high field strengths, leading to improved signal-to-noise ratios and increased resolution. This study discusses the magnet and gradient subsystems of MRI systems, as well as a number of complications that can occur from using a magnet. Furthermore, it illustrates the RF coils and transceiver’s finer characteristics and numerous constraints. It also showed the idea behind the data processing technology and the difficulties that come with it. In conclusion, the many artefacts inherent to MRI were elucidated. It also gives a quick rundown of the various problems that MRI have to deal with.
  Keywords: MRI image; spectroscopy; clinical application; medical image processing.
  DOI: 10.1504/IJMEI.2024.10064609
   
  
• Proposed novel approach for detecting Alzheimer’s disease in early stages  
  by Nadish Ayub, Syed Zubair Ahmad Shah, Rayees Ahmad Dar, Assif Assad, Abdullah Shah 
  Abstract: Alzheimers disease (AD), an incurable brain condition leading to memory loss, requires early detection for effective management. Utilising deep learning (DL) and computer vision (CV) in medical image analysis (MIA) shows promise, but acquiring annotated medical data is costly. Analysing 3D magnetic resonance images (MRIs) demands resource-intensive 3D convolutional neural networks (CNNs). This study introduces a 2D MRI slice-based transfer learning framework, employing a majority voting mechanism during testing. Experimentation reveals the optimal balance between data volume and accuracy is achieved with approximately 10 middle slices. The proposed approach attains a notable 93.91% accuracy, surpassing the state-of-the-art by 12.61% in distinguishing AD from cognitively normal (CN) cases.
  Keywords: deep learning; machine learning; Alzheimer’s disease; 3D MRI; transfer learning; 2D slices; convolutional neural networks; CNNs.
  DOI: 10.1504/IJMEI.2024.10064680
   
  
• Blood disease prediction system based on haemogram report using XGBoost algorithm  
  by Deepali K. Gaikwad, Ashok T. Gaikwad 
  Abstract: Healthcare systems gather extensive medical data from diverse tests spanning various fields. Unearthing concealed insights within this data remains critical for early disease detection or preventive measures. Classifiers utilise comprehensive blood count characteristics to forecast potential blood disorders in their initial phases, potentially amplifying the likelihood of effective treatment. Machine learning assumes a pivotal role in crafting predictive models based on historical data. This predictive procedure holds potential in curbing fatalities and enriching the lives of individuals managing these conditions. As an illustration, in this investigation, an XGBoost classifier was deployed on a blood test report dataset, successfully identifying a range of blood-related ailments like anemia, leukemia, lymphoma, sickle-cell anemia, and others.
  Keywords: machine learning algorithms; supervised learning; unsupervised learning; blood disease; XGBoost algorithm.
  DOI: 10.1504/IJMEI.2024.10064883
   
  
• Simulation of heart beat interval series - a neurophysiology based model in comparison with a nonlinear network oscillator model  
  by Sajitha Somasundaran Nair, Mini Maniyelil Govindankutty, Minimol Balakrishnan, Remya George 
  Abstract: Heart rate variability (HRV) analysis is a non-invasive method of autonomic nervous system (ANS) function assessment. In the present study a quantitative physiological model of the ANS is compared with a nonlinear network oscillator model in terms of power spectrum and poincare plots across three autonomic states. The physiological model more accurately reproduces HRV features compared to the nonlinear oscillator model. Rooted in actual physiology, the physiologic model allows extensive parameter tuning for various autonomic states, offering multiple modulable parameters. In contrast, the nonlinear oscillator model lacks physiological roots and permits only limited adjustments through coupling coefficients.
  Keywords: autonomic nervous system; ANS; heart rate variability; HRV; quantitative physiological model; neurotransmitter kinetics; nonlinear oscillator model; poincare plot; power spectrum.
  DOI: 10.1504/IJMEI.2024.10064942
   
  
• Automatic blood disease detection system based on haemogram report using decision tree algorithm  
  by Deepali K. Gaikwad, Ashok T. Gaikwad 
  Abstract: In healthcare systems, a vast quantity of medical data is collected from multiple medical tests conducted in various sectors. It remains required to extract hidden information from medical data to diagnose diseases at an early stage or even before they manifest. In order to predict information about possible blood disorders in their early stages and potentially increase the chance of a cure, the classifiers make use of the features of the full blood count. The development of predictive models based on historical data is the responsibility of machine learning. The process for predicting illness may reduce the number of deaths and enhance the lives of those who suffer from these conditions. Using a decision tree classifier applied to a dataset of blood test reports, this study identified several blood illnesses, including sickle cell disease, leukaemia, anaemia, and lymphoma.
  Keywords: blood disease; decision tree; machine learning; supervised; unsupervised.
  DOI: 10.1504/IJMEI.2024.10065459
   
  
• Cervical cancer detection and segmentation using ANFIS classifier and prediction by ensemble deep learning network  
  by D. Baskar, K. Manivanan 
  Abstract: In underdeveloped nations, the incidence frequencies of cervical cancer have been sharply rising while the health services for prevention, diagnosis, as well as therapy are still relatively few. Cancer screening procedures may lead to an early diagnosis, which increases the likelihood of successful treatment and, ultimately, the protection of cervical cancer. An approach for the identification, as well as fragmentation of cervical cancer depending on the adaptive neuro-fuzzy inference system (ANFIS), is provided here. A novel model has been built and given the name the colposcopy ensemble network (CYENET) to automatically diagnose cervical malignancies from colposcopy pictures. The cell classification feature is dependent on a compact vision geometry grouping (VGG) network termed compact VGG. The overall accuracy of the classification for VGG19 was 73.31%. The findings of the experiments indicate that the suggested CYENET displayed high levels of sensitivity (92.42%), specificity (96.22%), and kappa scores (88%).
  Keywords: ANFIS; machine learning; VGG19; cervical cancer.
  DOI: 10.1504/IJMEI.2024.10065831
   
  
• Detection of early biomarkers for Parkinson disease  
  by M.Z. Shaikh, R. Harini, Megala Jambulingam, M. Nalini 
  Abstract: The neuro-degenerative disorder known as Parkinsons disease (PD) is characterised by the gradual death of dopaminergic neurons. Individuals with Parkinsons disease often display a number of neurological as well as non-motor abnormalities. It is of the utmost importance to locate responsive as well as specific indications in order to meet the pressing requirement for improved earlier detection as well as subsequently disease-modifying therapy for Parkinsons disease (PD) individuals. There are four different types of markers that may be distinguished from one another. We extracted certain properties using a deeper artificial neuronal system, and then we categorised them as either belonging to Parkinsons disorder or not belonging to Parkinsons disorder. Utilising a method called ten-fold crossing validity, we were able to get extremely positive findings and the accuracy rate was ranging from 9293.5% for the input images used.
  Keywords: bio-marker; Parkinson’s disorder; deep learning; machine intelligence.
  DOI: 10.1504/IJMEI.2024.10066539
   
  
• Infrared hand joint dataset: a custom rheumatic hand thermal image dataset creation and its medically approved validation for extending research in diagnosing arthritis  
  by Rozina Naz, Gopal Sakarkar, Mohtashim Ahmad, Amina Vali, Noorul Amin 
  Abstract: Osteoarthritis (OA) and rheumatoid arthritis (RA) are distinct types of arthritis with different causes; OA results from joint wear and tear, while RA is an autoimmune disease. This research focuses on hand joints affected by arthritis, a leading cause of disability globally. Estimating the number of arthritis patients is challenging as many seek treatment only in severe stages. Conservative estimates suggest over 54 million adults and nearly 300,000 children have arthritis, with elderly individuals being more susceptible. However, 64% of affected adults are younger than 65. Arthritis can range from mild to severe, leading to chronic pain and disability. Severe arthritis can result in permanent joint damage and may affect other organs. The research aims to create a standardized thermal image dataset of hand joints, crucial for identifying inflamed areas. This dataset is well labelled and validated, addressing issues in previous datasets lacking proper annotation and validation.
  Keywords: osteoarthritis; infrared camera; thermal imaging; rheumatoid arthritis; thermal dataset.
  DOI: 10.1504/IJMEI.2024.10066772
   
  
• Prediction of cardiac disease based on whale optimisation algorithm and residual U-Net  
  by M. Balamurugan, P. Blessed Prince, M. Umapathy, S.P. Jeno Lovesum 
  Abstract: Due to the prevalence of the disease and the high death rate it is connected with, heart disease has become a significant public health problem in recent years. Therefore, predicting heart disease using certain simple physical indications that may be gained through having frequent physical examinations at an early stage has emerged as an important topic of study. UWOA is a deep neural network (U-Net) and whale optimisation (WOA) optimised technique created and applied to pick the ideal features that may boost the accuracy of heart disease prediction. The method that has been suggested makes use of a hybrid algorithm that will be broken up into two stages. The algorithm above can extract features, and the values that it generates show a high accuracy range of 97.34%, with improved accuracy by an average of 3.77 percentage points for heart attack prediction.
  Keywords: cardiovascular disease; whale optimisation algorithm; residual U-Net architecture.
  DOI: 10.1504/IJMEI.2024.10066862
   
  
• Image segmentation through advanced distance regularised level set evolution in chest X-ray analysis  
  by Ruchika Arora, Indu Saini, Neetu Sood 
  Abstract: This paper introduces a pioneering solution for the problem of weak edges and intensity inhomogeneities in medical images. Improved distance regularised level set evolution (I-DRLSE) model harnesses the power of local edge gradient features, local fitting variances, and meticulously weighted regional intensity information to orchestrate a seamless evolution of image contours. The variational level set method optimises diffusion rate through a refined quad-well potential function with range-based adaptability. It is validated on publicly available Japanese society of radiological technology (JSRT) dataset by achieving remarkable average dice similarity coefficient (DSC) and intersection over union (IoU) scores of 0.996 and 0.985, respectively.
  Keywords: image segmentation; synthetic images; CXR images; level set method; LSM; DRLSE; CV; adaptive bilateral filter; ABF.
  DOI: 10.1504/IJMEI.2024.10066938
   
  
• The role of thyroid medicine in relieve discomfort and health sustainability a study of patient health behaviour  
  by Yan Xu 
  Abstract: The objective of research is to determine the role of thyroid medicine in relieving discomfort and health sustainability from the perspective of patient health behaviour. This research collected 594 responses from thyroid patients as the sample size, and data analysis for this research is taken by Smart PLS 4. The study has confirmed that the impact of thyroid medicine is significant on relieve discomfort and health sustainability. This research also highlighted the importance of patients’ health behaviour as a moderating variable between thyroid medicine, relieving discomfort, and health sustainability. The research asserts that patients to get thyroid medicine for advancing their health sustainability.
  Keywords: patient health behaviour; thyroid medicine; health sustainability; relieve discomfort; health improvement.
  DOI: 10.1504/IJMEI.2024.10067004
   
  
• BiLSTM-net: a simple and efficient model for foetal compromise detection during intrapartum and antepartum phase  
  by Vidya Sujit Kurtadikar, Himangi Milind Pande 
  Abstract: Cardiotocography (CTG) is a well-known device used to assess foetus health by monitoring foetal heart rate (FHR) and uterine contraction (UC) simultaneously during the antepartum and intrapartum phases. The visual interpretation of CTG signals has associated high false positive and false negative rates due to its complex structure. Automatic detection of foetal state can be done using machine learning methods, thereby reducing undesirable adverse perinatal outcomes and unnecessary caesarean section. We propose a deep learning model suitable for time series FHR signals based on a bidirectional long short-term memory (BiLSTM) network. We used two open datasets and one private dataset, which are diverse enough to test the robustness of the deep learning model. The experiments on all three datasets achieved excellent and consistent results for all parameters. The results are encouraging enough to demonstrate the proposed models good predictive capacity.
  Keywords: cardiotocography; CTG; foetal monitoring; foetal compromise; bidirectional long short-term memory; BiLSTM; deep learning.
  DOI: 10.1504/IJMEI.2024.10067030
   
  
• SMART2C - a healthcare digital platform based on process mining for managing COVID-19  
  by Nesrine Missaoui, Asma Mejri, Wissem Hachfi, Amel Letaief, Asma Daassa, Emna Ammar, Hajer Chackroun, Haifa Boudriga, Marwa Seghaier, Ricardo Martinho, Sahar Toumia, Yemna Sayeb, Henda Hajjami Ben Ghezala, Sonia Ayachi Ghannouchi 
  Abstract: This paper addresses the difficulties and challenges that affected the Tunisian healthcare system during the spread of the COVID-19 virus. It proposes a digital platform to improve healthcare COVID-19 related processes, by applying process management methodologies and integrating process mining and decision mining techniques to analyse process execution traces. This integration assists stakeholders in making informed decisions and optimising resource utilisation, offering recommendations for healthcare providers to enhance their decision making. Additionally, it highlights emerging healthcare solutions with potential to mitigate the impact of pandemic-related events and safeguard against pandemics and other natural disasters.
  Keywords: COVID-19; digital platform; process mining; security; process implementation.
  DOI: 10.1504/IJMEI.2024.10067213
   
  
• Comprehensive assessment of enhanced transfer learning-based MRI classification for accurate brain tumour detection and grading  
  by S. Amsavalli, K. Saminathan, M.Chithra Devi 
  Abstract: Brain tumour identification and classification improve early detection and therapy planning. Brain cancer diagnosis with MRI scans is essential for fast and effective treatment. This study improves transfer learning-based MRI categorisation to improve brain tumour diagnosis and grading. Our method uses a varied collection of normal brain MRIs and glioma tumour pictures from different reliable databases. The preprocessing pipeline captures unique image features using Z-score normalisation and SURF model feature extraction. The work uses pre-trained weights and the VGG19 transfer learning model, fine-tuned on the MRI dataset, to classify brain tumours. Attention mechanisms in the VGG19 architecture prioritise informative regions in input images to improve discrimination. The Python technique uses TensorFlow and Keras deep learning libraries. Our brain tumour categorisation method achieved 99.2% accuracy in experiments. Our method surpasses Neural Networks, ANN, CNN, and CRNN by 2.1% in brain tumour detection and grading from MRI scans. According to extensive investigation and evaluation, our framework’s robustness and reproducibility could change neuroimaging diagnosis and improve clinical outcomes.
  Keywords: brain tumour detection; MRI images; transfer learning; speeded-up robust features; SURF; VGG19; Z-score normalisation; Keras deep learning libraries.
  DOI: 10.1504/IJMEI.2024.10067243
   
  
• Eye ball sensor controlled wheel chairs for disabled patients  
  by J. Gowrishankar, S.B. Mohan, Sreeja Vijay, P. Sivakumar 
  Abstract: This article proposes a system for eye ball motion controlled wheelchair, a device designed to help the elderly and those with mobility issues get around. An eye frame equipped with three proximity infrared (IR) sensor modules monitors the iris. If an infrared camera detects something white, it will generate a new string of binary digits. Arduino boards will these signals to operate the wheelchairs motors. Individuals experiencing difficulties with their motor nerves are the focus of this research. Even though they are disabled, this method will allow them to steer an electric wheelchair using their eyes alone. The user can steer and accelerate the wheelchair by looking left or right, and can also initiate and halt movement with their gaze. The proposed method runs well with an average reaction time of 3.2 seconds for eye control.
  Keywords: wheelchair; IR sensor; ultrasonic sensor; accelerometer.
  DOI: 10.1504/IJMEI.2024.10067333
   
  
• Age-based comparison of resting-state brain activity in autism spectrum disorder: an fMRI study  
  by B. Divya, A. Kavitha, R. Sowmya, B.S. Sneha, T.V. Santhoshiya 
  Abstract: Atypical alterations in brain development is a defining feature of autism spectrum disorder (ASD), but age-related changes in regional brain function and spontaneous activity may be more significant. ASD is linked to irregularities in neurological connections. Current rs-fMRI diagnosis lacks age differentiation, causing inconsistent findings. This study compares functional connectivity in adolescent and adult ASD patients. An age-based comparison shows connectivity declines with age in most brain regions but increases in the insula. Ring connectivity representation reveals better connections in adolescents, particularly in language regions. Connectivity matrices results are validated using fractional amplitude of low-frequency fluctuations (fALFF), which reduces physiological noise and increases accuracy in detecting brain activity. fALFF may serve as an early ASD detection tool or predict sibling risk, potentially through machine learning models. This research broadens the understanding of ASD brain connectivity development across age groups and opens ways for more focused, individualised interventions.
  Keywords: autism spectrum disorder; ASD; connectivity matrix; ring connectivity; amplitude of low-frequency fluctuations; ALFF; fALFF.
  DOI: 10.1504/IJMEI.2024.10067379
   
  
• A research based study of quantitative analysis and segmentation of thermal images for assessing rheumatoid arthritis in the hand joints  
  by Rozina Naz, Gopal Sakarkar, Mohtashim Ahmad, Noorul Amin, Amina Vali 
  Abstract: To investigate rheumatoid arthritis and observing bodys heat flow in the body and monitoring the changes in temperature of the skin. We will compare the difference in skin temperature between people with rheumatoid arthritis and people without it, in individuals with joint pain, the algorithms called the fuzzy c-means algorithm and expectation maximisation (EM) algorithm is used to pinpoint unusual areas on the hand (Backhaus et al., 1999; Tipa and Baltag, 2006; Kuruganti and Qi, 2002; Selvarasu et al., 2009; Snekhalatha et al., 2012). This study reveals how to examine thermogram measurement with heat distribution index (HDI) and estimate skin temperature. The temperature examination demonstrated a temperature increment of 0.970 C within the hand region of those patients with RA as compared to the hands of regular individual. Analysis of the patients data showed that there was a strong connection between HDI and skin temperature.
  Keywords: rheumatoid arthritis; thermograms; thermal segmentation; HDI; infrared images; hand joints.
  DOI: 10.1504/IJMEI.2024.10067688
   
  
• Osteoporosis diagnosis by femur bone modelling using COMSOL multi-physics  
  by Ammu Anna Mathew, Aastha Kapoor, Vivekanandan Sundaram 
  Abstract: Analysing bone behaviour requires a mathematical understanding of the functional history of bone and modelling since bone tissue can change shape to fit its function, much like a structural material. This study aims to model a femur bone using three-dimensional scanning by utilising the physical characteristics of a healthy and diseased one to detect osteoporosis as a pre-clinical evaluation tool. The Von Mises stress and Tresca stress values, which are the resulting point-load responses, were then recorded as part of the expanded inquiry, and a comparative analysis was conducted to get a thorough conclusion with the simulated model by calculating the bone reaction to mechanical changes.
  Keywords: COMSOL multi-physics; femur bone; stress distribution; osteoporosis; SolidWorks; Wolff’s law of remodelling.
  DOI: 10.1504/IJMEI.2024.10067727
   
  
• Genetic algorithm with ensemble feature ranking for feature selection in dengue prediction using convolutional neural networks  
  by K. Saraswathi, K. Rohini 
  Abstract: Early intervention and successful treatment depend on accurate disease diagnosis in medical diagnostics. With its high frequency and severe complications, dengue fever, a mosquito-borne viral infection in tropical and subtropical climates, is a major public health problem. Enhancing predictive models for dengue detection requires effective feature selection, especially when dealing with high-dimensional datasets. This study introduces a novel approach for feature selection in dengue prediction by integrating CNNs with GA-eFR. The GA-eFR algorithm leverages the feature ranking capabilities and the search efficiency of genetic algorithms to identify the most relevant features in the dataset. The methodology involves pre-processing the dataset, designing the CNN architecture, and incorporating GA-eFR into the feature selection process. Experimental results demonstrate that the proposed method significantly enhances the performance of CNN classifiers in detecting dengue, achieving an F-score of 0.92, precision of 0.90, recall of 0.88, and 92.85% accuracy on the OpenDengue dataset. The effectiveness of the proposed feature selection method is further validated by comparing it with other techniques, such as particle swarm optimisation (metaheuristic/swarm-based), recursive feature elimination (wrapper-based), and correlation-based feature selection (filter-based).
  Keywords: genetic algorithm; GA; feature ranking; feature selection; convolutional neural networks; CNN; dengue detection; medical diagnostics; optimisation algorithms; GA combined with ensemble feature ranking; GA-eFR.
  DOI: 10.1504/IJMEI.2024.10067742
   
  
• AI-based radiodiagnosis: a comprehensive review  
  by Jay Sawant, Niti Chikhale, Ayushi Tanna, Gajanan Nagare 
  Abstract: AI plays a critical role in the interpretation and diagnosis of medical conditions based on radiographic images. This review explores the symbiotic relationship between artificial intelligence (AI) and medical imaging, focusing on conventional and emerging modalities. Beginning with an overview of imaging modalities and AI fundamentals, a dive into diverse AI applications in the medical field has been made, emphasising advancements in image recognition and segmentation. Dedicated sections review the role of AI integrated with conventional as well as emerging radiology techniques, showcasing its adaptive potential for rapid diagnostics. Concurrently, the challenges in implementing AI in medical settings, including ethical considerations and validation frameworks have been addressed. By consolidating insights into the transformative impact of AI on diagnostic accuracy and efficiency, this review underscores its pivotal role in shaping the future of medical imaging.
  Keywords: artificial intelligence; healthcare.
  DOI: 10.1504/IJMEI.2024.10067812
   
  
• Cranial abnormalities: a deep learning-based approach for classification and measurement, enhanced treatment sensitivity through 3D printing  
  by Osman Nuri Uçan, Incilay Yıldız, Mehmet Ateş, Fulya Oduncu, Emir Kaan Cengiz 
  Abstract: This article emphasises that technological advances have a significant impact on the treatment process and are essential for the detection and correction of cranial abnormalities. Using the MUG500+ dataset, a deep learning-based classification was used. The aim of the research is to increase measurement precision and accuracy by simplifying the processes involved in helmet therapy measurements. The artificial intelligence solution focusing on deep learning makes a significant contribution to the identification of skull anomalies. The research provides a mathematical model for scaling the abnormal segment using the ResNet-18 algorithm for classification, which further improves the processes involved in 3D printing of the abnormal segment. Examining the results, it appears that the accuracy measure performed successfully in the testing phase (95% success rate). The accuracy and reliability of the proposed model is demonstrated by the sensitivity measure reaching 100%.
  Keywords: artificial intelligence; deep learning; cranial abnormalities; classification; 3D printing.
  DOI: 10.1504/IJMEI.2025.10070941
   
  
• Detection of sleeping disorders in Parkinson's affected patients using multi-head attention-based BiLSTM model  
  by Sk.Wasim Akram, A.P. Siva Kumar 
  Abstract: The proposed study aims to introduce a novel technique for identifying sleeping difficulties in PD patients by overcoming the existing limitations. Initially, the input samples are collected from the available dataset, and to remove unwanted noises in the input signals, pre-processing is carried out by using improved finite impulse response filter (FIRF), Bessel filter (BF), and continuous wavelet transform (CWT). After pre-processing, the complexity problem is avoided by extracting the most significant features by introducing an adaptive residual dense network (AResDenseNET) model. Based on the extracted features, a novel multi-head attention-based BiLSTM (MultiHeadAtt _BiLSTM) is proposed for detecting sleeping disorders. Finally, to improve the efficiency of the proposed model, its parameters are fine-tuned by an adaptive aquila optimiser (AAO) approach. As compared with existing approaches, the proposed model achieves a better accuracy result of 98.7%, and hence, it proves the efficacy of the proposed model.
  Keywords: time domain; linear shift-invariant system; scale factor; skip connections; backward information; candidate solution; forward information.
  DOI: 10.1504/IJMEI.2024.10068369
   
  
• Framework for sentimental analysis in healthcare data using deep Biaffine attention enabled CNN classifier  
  by Rushali Patil, Arunkumar Molapalayam Sekar 
  Abstract: Sentiment analysis associated with the COVID-19 pandemic tweets are paramount for informed decision-making and effective public health interventions. Hence, this research presents an effective deep Biaffine attention enabled CNN classifier model to handle the nuances and subjectivity inherent in COVID-19 tweets. This innovative model combines CNN with deep Biaffine attention mechanisms to comprehensively capture intricate word relationships within the COVID-19 tweets. Specifically, the attention mechanism enables the model to identify subtle emotional cues often ignored by traditional approaches, thereby enhancing the precision of sentiment analysis. Our framework demonstrated remarkable performance with an accuracy of 96.16%.
  Keywords: sentiment analysis; Biaffine attention enabled CNN; Aurelia swarm search optimisation; healthcare data; COVID-19 tweets.
  DOI: 10.1504/IJMEI.2024.10068395
   
  
• Challenges and issues in detection of seizure in early stages using artificial intelligence approach  
  by Pratima Ranjit Patil, Deepa Sachin Deshpande 
  Abstract: This study finds the challenges and issues in the early-stage detection of seizures using the artificial intelligence (AI) approach. This study examines the advances made in detection, calculation, and tracking of epileptic seizures utilising EEG technology. So, primary contribution of this study is to analyse numerous EEG-based seizure detection studies and how they address some of the challenges in seizure detection. The research investigated a number of methodologies, such as the internet of things framework, deep learning, and machine learning. This review focuses on: 1) detailed survey on elliptic seizure detection systems employing AI techniques; 2) survey conducted on AI approaches; 3) survey on the basis of EEG signal-based seizure detection in children; 4) survey on importance of thermal images in seizure detection; 5) dataset utilisation in seizure detection; 6) challenges and advantages in seizure detection; 7) the overall conclusion of this review is described in detail.
  Keywords: seizure detection; artificial intelligence; electroencephalography; epileptic seizures; early stages; thermal images; issues in seizure detection.
  DOI: 10.1504/IJMEI.2024.10068483
   
  
• Optimising heart disease prediction model parameters using pelican optimisation algorithm and Talos hyperparameter tuning  
  by R. Vinitha, Viji Vinod 
  Abstract: Heart disease (HD) is a major global health issue requiring reliable and efficient prediction models. Recent advances in machine learning and optimisation methods promise early identification and diagnosis. This systematic study uses data preparation, several optimisation algorithms, and Talos hyperparameter adjustment to enhance heart disease prediction. The goal Is to create a reliable prediction model with high sensitivity and specificity. Starting with heart disease data collection and preparation, the research is systematic. The processed data is then optimised using three algorithms: POA, CO, and Incomprehensible but Intelligent-in-time (ILAIBI) logic. After optimisation, statistical analysis evaluates sensitivity and specificity. Talos hyperparameter tuning optimises kernel parameters to refine model parameters. The POA method with optimised parameters improves prediction accuracy to 99% in the final prediction model. With such great accuracy, the model may be useful in early heart disease identification. Sensitivity and specificity analyses verify model reliability and robustness. The suggested cardiac disease prediction model uses advanced optimisation algorithms and hyperparameter tuning to be very accurate. The POA with Talos works well for clinical practice, the findings suggest. Future research should investigate model validation and optimisation strategies for different patient populations.
  Keywords: heart disease; HD; Talos; optimisation; pelican optimisation algorithm; POA; cheetah optimiser algorithm; CO; hyperparameter tuning; pre-processing and enhancement.
  DOI: 10.1504/IJMEI.2024.10068484
   
  
• Polynomial regressive projection-based pruning extreme learning classifier prediction of autism spectrum disorder by age group  
  by S. Sreevidya, Lipsa Nayak 
  Abstract: Autism spectrum illness (ASD) is a neurodevelopmental disorder that affects social interaction, communication, and behaviour. Early ASD discovery allows for prompt therapy. Multiple machine learning methods have been developed for autism identification; however, reliable prediction is difficult and time-consuming. PRCDP-PELC is a new method that improves ASD prediction accuracy and reduces time. Initially, data on toddlers, children, adolescents, and adults are collected from four databases. In the preprocessing phase, missing values are handled with polynomial regression interpolation, Tukeys range statistical test identifies noisy data, and generalised Blomqvists correlative Sammon projection reduces dimensionality. Finally, Tversky indexive pruning extreme learning classifier is used to classify data samples as ASD or not. In this way, the ASD prediction accuracy is improved, and prediction time is minimised. The results of the PRCDP-PELC technique provide better performance with higher accuracy, precision, and sensitivity and lesser prediction time when compared to existing methods.
  Keywords: autism spectrum disorder prediction; different age groups; data preprocessing; polynomial regression interpolation; generalised Blomqvists correlative Sammon projection; Tversky indexing pruning extreme learning classifier.
  DOI: 10.1504/IJMEI.2025.10068627
   
  
• Brain-computer interface and eye tracking for smart homes  
  by Prabhu, Nithya Jenev, Kamalakannan Ramachandran, Ashok Vajravelu 
  Abstract: The term brain-computer interface (BCI) refers to a sophisticated and diverse active research topic that is based on neurology, signal processing, biomedical sensors, hardware, and other related fields of study. Individuals who are paralysed may regain some of their freedom with the use of brain-computer interfaces (BCIs), which use brain impulses to operate prostheses or activate functional electrical stimulation. We examine the possibility of controlling a smart home environment using an augmented reality system that incorporates eyetracking as well as a brain-computer interface. We show controls that are depending on the context to the user through a head-mounted display, and the user chooses which controls to utilise by focusing their attention on those controls. In an average of 86% of the instances, the desired meaning was accurately inferred using a combination of EEG and eye tracking data.
  Keywords: electroencephalography; brain-computer interface; BCI applications; smart home; BCI challenges.
  DOI: 10.1504/IJMEI.2025.10068831
   
  
• Efficient clinical decision-making in nephrology: predictive analytics for chronic kidney disease management  
  by S. Vairachilai, Shreyash Jaiswal, Thota Chandan, S. Periyanayagi, S.P. Raja 
  Abstract: Chronic kidney disease (CKD) significantly impacts global health, making early detection and management crucial. This study explores the use of machine learning (ML), ensemble learning (EL), and deep learning (DL) models to enhance clinical decision-making in nephrology. We processed a comprehensive dataset, identifying feature importance using SHAP and splitting it into training and testing sets. Among the models, support vector machine (SVM) achieved 97.5% accuracy, while XGboost and Adaboost scored 98.7%. Deep learning models, especially the multilayer perceptron, also performed well. Our findings underscore predictive analytics potential to personalise CKD management and improve clinical outcomes in nephrology.
  Keywords: chronic kidney disease; CKD; nephrology; clinical decision-making: predictive analytics; machine learning; ML; ensemble learning; EL; deep learning; DL; SHapley additive explanation; SHAP; feature selection; evaluation metrics.
  DOI: 10.1504/IJMEI.2025.10068832
   
  
• Real-time visual quality enhancement and low-complexity image reconstruction via GPU-based deep learning  
  by Yogesh Mahadev Kamble, Raj B. Kulkarni 
  Abstract: This article presents a novel approach for reconstructing human figures from skeletal images captured at two different angles, utilising NVIDIAs CUDA framework for parallel image processing. The study enhances image quality and processing speed, particularly for optical-fibre endoscopic optical coherence tomography (OCT) systems, where high image quality is crucial. The research introduces an improved van Herk/Gill-Werman morphology algorithm for image component extraction and optimises operations with NVMe SSDs for better storage efficiency. Validation results demonstrate the methods superior performance, with a high dice coefficient (0.964), accuracy (0.993), correlation (0.97), and Structural Similarity Index (0.985), outperforming existing techniques.
  Keywords: image reconstruction; visual quality; NVIDIA GPU; parallel image processing; temporal making mechanism; generalisation.
  DOI: 10.1504/IJMEI.2025.10068833
   
  
• Time-resolved high-resolution imaging with enhanced contrast and SNR via multithreaded GPU-based autoregressive interpolation  
  by Yogesh Mahadev Kamble, Raj B. Kulkarni 
  Abstract: This study introduces a novel image reconstruction approach using a piece-wise autoregressive (PAR) model for predicting unknown pixels in high-resolution images. The method employs a block-based prediction strategy to enhance reconstruction accuracy. To optimise performance, a parallel processing strategy leveraging graphics processing units (GPUs) is implemented. CUDA threads are used to launch each particle in the optimisation framework, while CUDA streams divide the swarm into smaller groups for efficient task offloading. The particle swarm optimisation (PSO) technique integrates a particle coding strategy and a gradient-based local search to improve accuracy and speed up convergence. The approach is evaluated using the DIV2K and APTOS datasets, demonstrating its ability to reconstruct high-resolution images with minimal errors. Experimental results show that the CUDA-accelerated PSO method significantly enhances reconstruction performance, highlighting its effectiveness in handling noisy or incomplete data, making it a promising solution for high-quality image reconstruction tasks.
  Keywords: image reconstruction; time-resolved high resolution; contrast-enhanced imaging; autoregressive image interpolation; residual filtering; and conjugate gradient method; CG.
  DOI: 10.1504/IJMEI.2025.10068834
   
  
• High-resolution magnetic resonance image reconstruction model for early epilepsy detection using hybrid optimisation approach  
  by D. Chandraprakash, Kapil Gupta, Sagarkumar S. Badhiye, Senthil Kumar Chandrasekaran, Pavan Kumar Varma Kothapalli 
  Abstract: Epilepsy, a complex neurological disorder, requires early detection and intervention to reduce its impact. In MRI-based diagnostic imaging, reconstructing images from sparse data is essential for accurate diagnosis. This study introduces a novel approach for high-resolution MRI reconstruction aimed at early epilepsy detection. Initially, a multi-scale adaptive mean-Wiener filtering technique is used to reduce noise in collected MRI data. The processed images are then reconstructed using a generative adversarial network (GAN) to enhance high-level feature extraction. Subsequently, a hybrid model integrating a deep convolutional spiking neural network with the artificial gorilla troops optimiser algorithm is proposed, combining traditional iterative methods with machine learning for optimised reconstruction. The model, implemented using MATLAB, demonstrates superior performance with a remarkable accuracy of 99.57%, surpassing existing methods such as MCC-TL, HANQL, and MBINet-SONN. This advancement promises enhanced diagnostic accuracy and improved early intervention for epilepsy and other neurological disorders.
  Keywords: magnetic resonance image; reconstruction; epilepsy seizure detection; generative adversarial network; GAN; deep convolutional spiking neural network; DCSNN.
  DOI: 10.1504/IJMEI.2025.10068873
   
  
• Efficient skin disease detection using optimised AlexNet50 and feature correlation subspace vector modulation  
  by D. Ushanandini, T. Kamala Kannan 
  Abstract: In identifying skin cancer, image processing algorithms have increasingly relied on object recognition in isolated regions. However, most current approaches ignore crucial parameters like feature dimension and disease-specific impact, resulting in large classification mistakes, especially during early identification. These errors lower precision, recall, and F1 scores. This research proposes a feature-constrained support vector machine with optimised ALEX-CNN to overcome these restrictions and identify skin cancer. Image pre-processing and normalisation are done in the algorithm. We identify sick regions using the former and Canny edge detection. To segment the impacted region, we use limited feature subset selection. The optimised ALEX-CNN evaluates the strongest disease-relevant feature margins ranked by FCSVM. The experimental results demonstrate that our suggested method outperforms all other comparable studies in skin cancer diagnosis accuracy (97.2%), precision (96.4%), recall (97.1%), and false detection rate. This shows that the strategy may improve categorisation and early cancer diagnosis.
  Keywords: skin disease detection; feature selection and classification; AlexNet50; cross-fold slice segmentation; CFSS; AlexNet50 convolutional neural network; ALEX-CNN; Canny edge morphing; CEM; feature correlation subspace vector modulation; FCSVM.
  DOI: 10.1504/IJMEI.2025.10070978
   
  
• Deep learning techniques-based prediction of chronic insomnia  
  by Rajalakshmi Ramanathan, K. Nithiya, S. Gayathri, C. Sivakumaran 
  Abstract: The potential of technologies has resulted in significant changes to the globe during the course of that steadily for the past decades. Insomnia is one of the many health problems that are on the rise as a result of persons increased participation in digital activities, decreased engagement in physically active pursuits, and increased reliance on electronic gadgets that emit radio waves. This may happen on its own or as a consequence of another issue. Both scenarios are possible. Insomnia that persists over a long period of time, often known as severe insomnia, might cause significant and irreparable harm to a persons brain. Sleeplessness is a condition that may be identified by a variety of clinical testing in accordance with the many internal elements that influence sleep. This method, on the other hand, is not only costly but seems to be equally time-consuming as well as labour-intensive.
  Keywords: deep learning; artificial intelligence; machine learning; insomnia.
  DOI: 10.1504/IJMEI.2025.10071308
   
  
• Optimised palm-print recognition system using deep learning models with image processing approaches  
  by J. Sheela Mercy, S.Silvia Priscila 
  Abstract: Palm print recognition enables biometric authentication in forensics, access management, and security. Traditional palm print identification systems sometimes have segmentation errors, noise interference, and limited feature extraction. New deep learning (DL) and image processing approaches can improve the accuracy and reliability of palm print recognition systems. This paper presents a revolutionary palm print identification method using cutting-edge DL and image processing approaches. After removing noise with an AWF, marker-controlled watershed (MCW) segmentation is used. ResNet, GoogLeNet, and AlexNet are DCNN frameworks for feature extraction and categorisation. The squirrel search algorithm optimises results. ResNet with adaptive Weiner filtering, marker-controlled watershed segmentation and SSA perform well in accuracy (95.10%), sensitivity (0.93), specificity (0.91), and F1-score (0.95). This integrated system improved palm print recognition technology. This study uses DL and image processing to present a palm print identification approach with high performance and reliability.
  Keywords: palm-print recognition; deep learning; ResNet; image processing; segmentation; marker-controlled watershed; MCW; classification; convolutional neural network; CNN; squirrel search algorithm; SSA.
  DOI: 10.1504/IJMEI.2025.10069107
   
  
• An intellectual framework for brain tumour detection and survival prediction using deep learning methods  
  by K. Shanmuga Priya, G. Rosline Nesa Kumari 
  Abstract: This paper aims to design a deep learning-based brain tumour detection and survival prediction method. Initially, the process starts with standard database and then the images are processed through a segmentation phase to segment the abnormality by multiscale MobileUNet++ (MMUnet++). Next, the feature extraction process is performed by using segmented images. Finally, the resultant features are given to vision transformer-based residual bidirectional long short-term memory (ViT-ResBiLSTM) for tumour detection. The diverse metrics are considered for the validating model efficiency. This integration facilitates the detection process, allowing for the identification of brain tumours with better accuracy and provides performance.
  Keywords: tumour segmentation; brain tumour detection; survival prediction; multiscale MobileUNet++; MMUnet++; vision transformer-based residual bidirectional LSTM; ViT-ResBiLSTM.
  DOI: 10.1504/IJMEI.2025.10069116
   
  
• An Intelligent model of brain tumour detection and survival prediction using adaptive MobileNet with multihead cross attention module  
  by K. Shanmuga Priya, G. Rosline Nesa Kumari 
  Abstract: A new deep learning method is developed for identifying brain tumours and forecasting patient survival outcomes is developed. The requisite images are garnered from standard sources. Thereafter, the aggregated images are forwarded to the segmentation phase, which is carried out by the multi-dilated MobileUnet3+ (MDMUnet3+). The obtained segmented image is fed into the adaptive MobileNet with multihead cross attention module (AM-MCAM) for getting the classified outcome; here the parameters are optimised by employing enhanced garter snake optimisation (EGSO). Finally, the experimental analysis is carried out on the designed brain tumour identification and survival-prediction approach to confirm its effectiveness.
  Keywords: brain tumour detection; survival prediction; segmentation; multi-dilated MobileUnet3+; MDMUnet3+; adaptive mobilenet with multihead cross attention module; enhanced garter snake optimisation.
  DOI: 10.1504/IJMEI.2025.10069117
   
  
• Enhancing maize disease detection and classification using optimised hyper-tuned EfficientNet architecture  
  by Bhawna Kumawat, Reeba Korah, Ashok Kandipati 
  Abstract: Climate variability exacerbates diseases that pose serious production difficulties for maize, a staple crop for global food security. Manual inspection is the foundation of traditional disease detection techniques, which are beset by scaling issues, errors and inefficiencies. This study presents a state-of-the-art convolutional neural network (CNN) tailored for the detection of maize disease: the hyper-tuned EfficientNet (HTENet) architecture. The unique spatial-channel attention method of the proposed model reduces the influence of complicated image backgrounds and improves disease-specific feature extraction by separating important visual characteristics. The study emphasises the significance of hyperparameter optimisation, which includes adaptive pooling techniques, learning rate adjustments, and data augmentation. The study uses criteria like precision, recall, F1-score and mean average precision (mAP) to show how much better HTENet is at detecting and categorising maize leaf diseases. With the help of sophisticated transfer learning and hyperparameter tuning, HTENet outperforms benchmark models like YOLOv5, ResNet50 and VGG16 in terms of efficiency and precision.
  Keywords: maize disease detection; deep learning; hyper-tuned EfficientNet; HTENet; convolutional neural networks; CNN; transfer learning; agricultural technology.
  DOI: 10.1504/IJMEI.2025.10069118
   
  
• A comprehensive approach to integrating multimodal data in healthcare AI  
  by Zunzarrao Vilasrao Thorat, Amina Kotwal, Mahesh M. Bulhe 
  Abstract: Integrating multimodal data in healthcare AI such as clinical notes, lab results, imaging, genetic profiles, and wearable data presents challenges like data silos, interoperability, privacy concerns, algorithmic bias, and interpretability. To optimise patient care, standardised frameworks for data integration are needed, along with enhanced interoperability, improved AI accuracy, and stronger privacy measures. Techniques like multi-criteria collaborative filtering (MCCF) enhance personalised treatment recommendations, while the synthetic minority over-sampling technique (SMOTE) addresses class imbalances and improves predictive accuracy. Network access control (NAC) ensures security. Future advancements should focus on real-time integration, AI-driven analytics, patient engagement, and ethical standards.
  Keywords: healthcare AI; multimodal data; multi-criteria collaborative filtering; MCCF; synthetic minority over-sampling technique; SMOTE; network access control; NAC; clinical decision making.
  
  
• Projection of chronic disease using machine algorithms  
  by P. Nithya, C.D. Nandakumar, S. Srinivasan 
  Abstract: The study aims to project the future prevalence of chronic diseases in India using data from the World Health Organisations 2024 report, focusing on nine key factors related to non-communicable diseases (NCDs). The study analysed data from 2002 to 2023, using Holt-Winters exponential smoothing for missing values and Gaussian hidden Markov models to predict NCD occurrences from 2024 to 2030. Key areas for intervention identified include obesity, physical inactivity, and elevated fasting glucose. Four machine learning models were evaluated for forecasting NCD mortality, with linear regression yielding the best accuracy (R2 of 0.989), outperforming Random Forest and Gradient Boosting models.
  Keywords: chronic diseases; Hidden Markov models; machine learning; mortality; forecasting.
  DOI: 10.1504/IJMEI.2025.10069647
   
  
• Breast cancer forecast using enhanced support vector machine model  
  by M.M. Agarwal 
  Abstract: Breast cancer (BC) significantly affects womens health worldwide. Timely identification and precise risk evaluation are essential for successful treatment and improved patient outcomes. Among women globally, breast cancer is the second deadliest cancer, with only lung cancer causing more fatalities. Contemporary advancements have allowed specialists to utilise various innovative approaches alongside conventional methods for breast cancer diagnosis in women. Machine learning techniques have been successfully applied in the realm of computational cancer treatment. Studies have demonstrated that diagnostic precision can be enhanced using machine-learning techniques. The breast cancer risk predictor was developed through a dedicated study using the Breast Cancer Wisconsin (diagnostic) dataset and a refined support vector machine (SVM) learning algorithm. SVMs provide a novel method for transforming nonlinear data, enabling the use of a linear algorithm to fit a linear model to data. The resulting model exhibited an accuracy of approximately 95% when applied to the training data.
  Keywords: breast cancer; supervised learning; support vector machine; SVM; cancer risk predictor; breast cancer forecast.
  
  
• Impact of demographic, temporal, and socio-economic factors on Lyme disease incidence in the Northeastern USA  
  by Ali Ben Khalil, Dinesh P. Mital, Shankar Srinivasan 
  Abstract: Lyme disease ranks as the leading vector-transmitted illness across the USA, with its incidence demonstrating significant variations across different demographics. This paper examines the impact of demographic (age, gender, race/ethnicity), temporal (seasonality), and socio-economic factors on the incidence of Lyme disease in the Northeastern USA. Data from the Healthcare Cost and Utilization Project (HCUP) from 2017 to 2020 were analysed to identify disparities in disease incidence and aid in the development of targeted public health interventions. The results indicated peak Lyme disease admissions during July (18.7%) and the least in January (4.4%). Mortality during hospitalisation was low (1.2%), with the disease predominantly affecting males (56.1%) and individuals identified as White (88.8%). 2 A.B. Khalil et al. Socio-economic analysis revealed a higher incidence in the highest income quartile (42.9%) and the lowest (8.8%). Age-wise distribution showed the highest incidence in older adults 65 years and above (42.0%).
  Keywords: Lyme disease; Lyme borreliosis; demographic factors; temporal factors; socio-economic status; epidemiology; geographic variability; disease surveillance; the northern USA.
  
  
• On the reduction of pulsations of piezoelectric micropump using a microneedle valve for drug delivery applications  
  by Sachin R. Gavali, Prashant M. Pawar 
  Abstract: Piezoelectric (PZT) diaphragm-type mechanical micropumps produce pulsing fluid flow. However, continuous fluid flow devoid of any flow pulsation is necessary for applications such as drug delivery systems. To reduce the pulsations, the micropump is paired with microneedle valves for medication applications. This study incorporated a microneedle valve at the delivery side to investigate the flow pulsations produced by the piezoelectric micro pump. The study reveals that by incorporating a microneedle valve into the micropump pulsations can be significantly reduced. Inculcating a microneedle valve to the piezoelectric pump utilised in the study reduced the overall pulsation by 93.67%.
  Keywords: pulsations in flow; degree of pulsation; piezoelectric micropump; microneedle valve; drug delivery.
  DOI: 10.1504/IJMEI.2025.10069816
   
  
• Multimodal authentication for healthcare data protection using gated recurrent networks  
  by Maddila Suresh Kumar, Satyanarayana Botsa 
  Abstract: Protecting biometric data in healthcare is critical. This study introduces a multimodal authentication model utilising gated recurrent networks (GRNs) to enhance security. This study presents a multimodal authentication model utilising gated recurrent networks (GRNs) to enhance the security of biometric data in healthcare. The examination presents the usage of a squeeze-and-excitation network (SE-Net) consideration instrument, combined with an example taken on the wavelet method, for productive element extraction. The gated recurrent unit (GRU) addresses long-range dependencies in the verification process. Tested in MATLAB, the model achieved a classification accuracy of 99.12%. Future work aims to improve real-time authentication efficiency, incorporate additional biometric modalities, and develop adaptive security measures against evolving cyber threats in healthcare.
  Keywords: healthcare; biometric data; multimodal authentication; gated recurrent network; iCanClean algorithm; squeeze-and-excitation network; SE-Net.
  DOI: 10.1504/IJMEI.2025.10069817
   
  
• Retinal structure segmentation for glaucoma classification using fundus and OCT images  
  by Esra'a Mahmoud Jamil Al Sariera, M.C. Padma, Thamer Mitib Ahmad Al Sariera 
  Abstract: Glaucoma is a frequent chronic condition that causes irreversible vision loss. We propose a new technique to build an automated feature-learning system for detecting glaucoma in colour retinal fundus images and optical coherence tomography (OCT) images. Three phases are described in this paper. First, fundus and OCT image enhancement are performed. Secondly, the ROI is segmented and glaucoma features are extracted. Finally, a fuzzy logic classifier is used to classify the features as either glaucoma or normal. We were able to reach 96% accuracy, which was significantly less expensive to process than the state-of-the-art. The proposed system is evaluated using the data on OCT, fundus images and DRISHTI-GS for glaucoma detection.
  Keywords: glaucoma; optical coherence tomography images; OCT; fuzzy logic classifier; DRISHTI-GS dataset.
  DOI: 10.1504/IJMEI.2025.10069956
   
  
• Smart IoT-based system for efficient neonatal respiration monitoring and control  
  by Sukhwinder Sharma, L. Maria Michael Visuwasam, R. Deeptha, T. Shynu, S. Suman Rajest 
  Abstract: This work presents a system that enables the monitoring of infants in incubators, collecting relevant data for diagnosis and research at neonatal intensive care units (NICU). It comprises the incubator with breath, heartbeat, heat detectors, and a MEMS detector that tracks an infants heart monitor. The receiver part sends this information (respiration rate, pulse, and MEMS sensor levels) to a mobile device with an Arduino microcontroller. When there is an abnormal heart rate, a DC motor triggers to provide oxygen by giving commands through the microcontroller. Doctors could even use a local area network to check the babys status before it goes home for remote telehealth monitoring. The data comes to a mobile device via IoT, and an app manages therapy progress. We have developed this framework to illustrate a passive sensing application designed to enhance future remote health monitoring solutions that can benefit care and outcomes for infants in the NICU. This early persuasive evidence also suggests that the system integration of real-time transduction and automated responses represents a credible path to facilitate more efficient neonatal care with higher performance objectives.
  Keywords: automated respiration control; IoT-based sensors; neonatal health tracking; oxygen supply regulation; real-time data analysis; remote health monitoring; smart infant care system.
  DOI: 10.1504/IJMEI.2025.10070120
   
  
• A review: liposomal formulations in the management of idiopathic pulmonary fibrosis  
  by Nehaba Natwarsinh Solanki, Vijaykumar K. Parmar 
  Abstract: Idiopathic pulmonary fibrosis (IPF) is progressive lung illness that affects lung function and is characterised by fibrosis. New strategies are needed since traditional treatments frequently have severe side effects and little effectiveness. Drug distribution, bioavailability, and toxicity reduction in the treatment of IPF may all be enhanced by liposomal formulations, which are made up of lipid bilayers encasing medicinal substances. Liposomal drug research, including anti-fibrotic medicines, their physical characteristics (size, charge), and encapsulation efficiency are all included in this review. In addition, issues like stability and repeatability are discussed, emphasising how liposomal formulations may improve the outcomes for IPF patients.
  Keywords: idiopathic pulmonary fibrosis; IPF; liposomal formulation; aerosol characterisation; anti-fibrotic agents; pulmonary drug delivery technologies.
  DOI: 10.1504/IJMEI.2025.10070227
   
  
• Healthy eating followed by circadian rhythm help to reverse diabetes  
  by Madhavi Vijaykumar Chekatla, Amiya Bhaumik, Mohammad Gousuddin, Vijaykumar Chekatla 
  Abstract: The transformative potential of coordinating lifestyle treatments with circadian rhythms for the management and possible reversal of type 2 diabetes mellitus (T2DM) is investigated in this study. 36 people, ages 30 to 60, participated in the 180-day experiment, which focused on individualised dietary modifications, exercise, intermittent fasting, and meal scheduling based on circadian rhythms. There were notable glycemic improvements: postprandial glucose levels plummeted by 60%, fasting glucose levels dropped by 52.5%, and HbA1c levels dropped by 26%. As a result, diabetic drugs were completely stopped. Among the supplementary advantages were a 10% decrease in body mass index, increased vigour, and glowing skin; participants reported feeling 1015 years younger. The results highlight the importance of circadian rhythms in promoting metabolic health, with time-restricted meals improving insulin sensitivity and regulating blood pressure. By combining exercise and detoxification techniques, the studys integrative approach emphasises the comprehensive advantages of managing diabetes that go beyond glucose control to include general health and well-being. The findings support a paradigm change in diabetes care by highlighting individualised, circadian-aligned therapies.
  Keywords: circadian rhythms; type 2 diabetes mellitus; T2DM; glycemic control; lifestyle interventions; intermittent fasting; circadian-aligned therapies.
  DOI: 10.1504/IJMEI.2025.10070298
   
  
• Depression prediction based on socio-economic factors: a comparative study  
  by C. Gunavathi, Aadya Ranjan 
  Abstract: This article addresses the impact of the COVID-19 pandemic on mental health and evaluates various feature selection methods with supervised classification algorithms to predict mental health conditions using socio-economic data. Results show information gain and Gaussian Naive Bayes as the most effective combination with 84.75% accuracy. The study emphasises the importance of feature selection for improved predictive accuracy and sheds light on algorithm performances when paired with different feature selection methods. Notably, the decision tree algorithm with fisher score yields the lowest accuracy at 71.98%. Insights gained from this comprehensive comparison include identifying common attributes significantly influencing model predictions.
  Keywords: classification methods; supervised learning; machine learning; mental health; feature selection methods; classification models; Chi square test; information gain; Gaussian Naïve Bayes; random forest.
  DOI: 10.1504/IJMEI.2025.10070323
   
  
• Hybrid classifier of adaptive potential in biotechnical systems of rehabilitation type  
  by Sergey Filist, Riad Taha Al-Kasasbeh, Alexey Kiselev, Olga Vladimirovna Shatalova, Nikolay Korenevskiy, Osama M. Al-Habahbeh, Ashraf Shaqadan, Maxim Ilyash 
  Abstract: For the rehabilitation of persons with disabilities, a method for classifying adaptation potential is proposed. It allows you to monitor and manage the functional state of patients during therapy or a rehabilitation session. A hybrid classifier of adaptive potential has been developed, based on the aggregation of heterogeneous data characterising the functioning of various organs and subsystems of the body. It is used to predict the risk of recurrent myocardial infarction. The hybrid classifier was tested on an experimental group of post-infarction patients. The quality indicators of the medical risk classification of the synthesised hybrid classifier allow us to recommend it for biotechnical rehabilitation systems that carry out therapeutic and restorative procedures in post-infarction patients. Diagnostic efficiency by 11%.
  Keywords: adaptive potential; hybrid classifier; HC; virtual model; algorithm; recurrent myocardial infarction; cumulative survival.
  DOI: 10.1504/IJMEI.2025.10070409
   
  
• EEG-driven mental state detection using residual networks  
  by Soumen Ghosh, Tanmay Sinha Roy, Prem Prakash, Raja Gupta 
  Abstract: The study explores the use of EEG, a method for measuring electrical brain activity, in recognising emotions. EEGs role in emotion recognition is crucial for enhancing human-computer interaction and brain-computer interfaces (BCI). The dataset includes EEG data from two participants (one male and one female), observed for three minutes per emotional state (positive, neutral, negative). The research employs a residual neural network (ResNet) architecture, achieving a 95.12% accuracy in classifying emotions. ResNet models, typically used in image recognition, prove effective in this context as well. The study concludes that this deep learning approach outperforms traditional machine learning models in emotion classification.
  Keywords: residual neural network; human emotions; emotional classification; deep learning; ResNet architecture.
  DOI: 10.1504/IJMEI.2025.10070763
   
  
• The internet of things-based platform for real-time classifications with autism spectrum disorder detection using modified SVM algorithm  
  by Subair Ali Liayakath Ali Khan, K. Saravanan 
  Abstract: There are many challenges for the children who are diagnosed with autism spectrum disorder (ASD) in every aspect of their lives, including social engagement, interpersonal communication, and so on. The symptoms of ASD differ from one individual to another and can range from mild to severe. Autistic children use fewer communication signals than typically developing children. Parents may be able to identify and recognise the activities of their kids and comprehend their ideas with practice. Others, on the other hand, have difficulties in interpreting their gestures. An internet of things (IoT) based platform is proposed to identify the posture and signals of ASD children using multiple classification approaches. The data set is created by performance of ten children execute every movement roughly ten times to build a dataset. The primary objective of this research is to develop an IoT-based system for recognising gestures of kids with ASD.
  Keywords: autism spectrum disorder; ASD; machine learning; ML; random forest; SVM; XGB; internet of things; IoT.
  DOI: 10.1504/IJMEI.2025.10070768
   
  
• Endoscopic ultrasonography in differential diagnostics of benign and malignant pathology of the common bile ducts utilising fuzzy mathematical technologies  
  by Etab Taha Al-Kasasbeh, Nikolay A. Korenevskiy, Vladimir Anatolievich Belozerov, Riad Taha Al-Kasasbeh, Sofia Nikolaevna Rodionova, Sergey Filist, Ksenia V. Razumova, Emad Saleh Al-Tarawneh, Osama M. Al-Habahbeh, Moath Musa Al Smady, Mahdi Salman Alshamasin, Elena Shalimova, Vladislav Krutskikh, Maxim Ilyash 
  Abstract: The study aims to enhance the outcomes of differential diagnosis of obstructive pathology of the common bile duct by leveraging the results of endoscopic ultrasonography (EUS) using hybrid fuzzy decision-making technologies. The main diagnostic methods are endosonography and the hybrid fuzzy decision rule synthesis methodology, developed at Southwest State University in Kursk, Russia. Based on expert assessments of endosonograms depicting common bile duct pathology, and the hybrid fuzzy decision rule synthesis methodology, a hybrid fuzzy model for differential diagnosis of the studied pathology was developed. Mathematical modelling and statistical calculations demonstrated the effectiveness of this differential diagnostic model, providing a confidence level of 0.93 in the desired diagnosis which is an acceptable result when working with poorly formalised initial data.
  Keywords: fuzzy mathematical models; common bile duct obstruction; reference images.
  DOI: 10.1504/IJMEI.2025.10070769
   
  
• ARAN-ASPP: efficient pulmonary emphysema disease classification framework using adaptive residual attention network with atrous spatial pyramid pooling network  
  by Indira Linginani, Akkalakshmi Muddana 
  Abstract: An effective deep network-based pulmonary emphysema disease classification framework is proposed. Initially, the images are collected from standard dataset. Next, the images are forwarded to the segmentation phase which is performed by developed SegUNet++ with novel loss function (SegUNet++-NLF). The loss function is adopted to enhance generalisation capacity of the SegUNet++. The segmented images are passed to the adaptive residual attention network with atrous spatial pyramid pooling (ARAN-ASPP) to get efficient outcome. The parameters in ARAN-ASPP are tuned by diversified fitness-based skill optimisation algorithm (DFSOA). Finally, the validation process is carried out to confirm the efficacy of the developed model.
  Keywords: adaptive residual attention network; atrous spatial pyramid pooling; diversified fitness-based skill optimisation algorithm; DFSOA; pulmonary emphysema disease classification; SegUNet++ with novel loss function; SegUNet++-NLF.
  DOI: 10.1504/IJMEI.2025.10071131
   
  
• Acoustic measurements for diagnosis and evaluation of prosodic disorders in a patient with spastic dysarthria case study  
  by Ali Kaddour, Kamel Ferrat 
  Abstract: This case-control study evaluates acoustic markers for diagnosing prosodic disorders in spastic dysarthria through comparative analysis of a cerebral palsy patient (18 years) and a neurotypical control (19 years). Using Praat ® software, we quantified six parameters during monologue, text reading, and syllable repetition: fundamental frequency variability (F0SD), intensity modulation (Int SD), pause frequency (NoP), disfluent word percentage (PDW), speech rhythm, and rhythm variation (COV). Results revealed significant intergroup differences: the patient exhibited markedly higher F0SD (31.08 Hz vs. 2.6 Hz), Int SD (14.4 dB vs. 3 dB), NoP (42 vs. 7 pauses), PDW (6.59% vs. 1.51%), and COV (5.17% vs. 1.5%), alongside slower rhythm (3.2 vs. 7.25 syllables/sec). Perceptual assessment using the clinical assessment battery of dysarthria (BECD) confirmed severe prosodic abnormalities in the patient with spastic dysarthria. Our findings validate a multimodal assessment framework combining perceptual evaluation with computerised acoustic analysis, offering speech and language pathologists (SLP) quantifiable tools for targeted intervention in spastic dysarthria.
  Keywords: dysarthria; spastic dysarthria; acoustic measurements; diagnostic; Praat® software; speech-language pathologist; SLP.
  DOI: 10.1504/IJMEI.2025.10071257
   
  
• Efficient pancreatic cancer detection framework using CGU-net segmentation and HResST network  
  by Garigipati Rama Krishna, K. B. V. Brahma Rao 
  Abstract: The DL-based approaches are used in the proposed methodology for detecting pancreatic cancer. Computed tomography (CT) scan images are used to detect positive cases. Initially, the intensity guided Fourier transform (IGFT) augmentation model is employed to achieve a data augmentation technique that tackles overfitting problems. After data augmentation, pancreatic tumour regions are segmented using cascaded gated U-Net (CGU-Net), which removes unnecessary noise information from the CT image and extracts the region of interest (ROI). Finally, the hybrid residual Swin transformer network (HResSTNet) is utilised for pancreatic cancer detection. In order to extract the significant features, the Swin transformer multi-scale residual-dense network blocks (MsR-DB) are utilised. The proposed model is trained and evaluated using the Medical-Segmentation Decathlon (MSD) Challenge dataset. The HResSTNet models performance is assessed through a variety of performance metrics. This model achieves 99.25% accuracy and 99.23% precision. Thus, the proposed model enhances the pancreatic cancer detection performance.
  Keywords: pancreatic cancer detection; deep learning; CT scan images; cascaded gated U-net; Swin transformation; image segmentation; data augmentation.
  DOI: 10.1504/IJMEI.2025.10071565
   
  
• A comprehensive review of computational intelligence approaches in type 2 diabetes prediction and classification  
  by Hitesh B. Patel, Keyur Brahmbhatt 
  Abstract: High glucose levels in diabetic individuals cause damage to the eyes, kidneys, and nerves, which is called microvascular problems. This paper provides a comprehensive review of deep learning and machine learning techniques and identifies their Key features and advantages for type 2 diabetes (T2D) prediction. The deep learning techniques, including neural networks, convolutional neural networks (CNN), U-Nets, and DNNs, along with machine learning models including support vector machine (SVM), decision tree (DT), and logistic regression, are mainly used for type 2 diabetes prediction. Most prediction models utilise ANOVA tests, chi-squared tests, and recursive feature elimination procedures for enhanced model accuracy. The deep learning approaches have the potential to identify early disease conditions; hence, preventing long-term issues associated with T2D is possible. Despite variances in both analytic approaches, incorporating them into medical health practices will assist patients in adopting a healthy lifestyle while lowering significant healthcare costs for an individuals family and society.
  Keywords: type 2 diabetes; T2D; convolutional neural networks; CNN; support vector machine; SVM; decision tree; DT; U-Nets; ANOVA.
  DOI: 10.1504/IJMEI.2025.10071889
   
  
• An efficient liver tumour detection using elephant swarm optimised pyramid deep convolutional neural network  
  by Baswanthrao Patil, Sachinkumar Veerashetty 
  Abstract: The research aims to enhance the detection of liver tumours through the utilisation of an elephant swarm optimised pyramid deep convolutional neural network (DCNN) classifier. The pyramid deep CNN classifier uses these features as input after integrating a pyramid net to progressively increase the feature map dimension and boost its efficacy. A novel elephant swarm optimisation method is presented that combines particle swarm optimisation with elephant herding optimisation to improve the classifier tuning. This approach shows reduced prediction errors and enhanced performance in the identification of liver tumours. Performance metrics of the proposed method are evaluated by the training point and K-fold with the accuracy, specificity, and sensitivity.
  Keywords: deep learning; pyramid deep CNN classifier; liver tumour detection; elephant swarm optimisation; ESO; particle swarm optimisation; PSO.
  DOI: 10.1504/IJMEI.2025.10072172
   
  
• Analyses of Lyme disease discharge data in the Northeastern United States using machine learning techniques  
  by Ali Ben Khalil, Dinesh P. Mital, Shankar Srinivasan 
  Abstract: Lyme disease, the most prevalent vector-borne disease in the USA, demonstrates distinct demographic and temporal patterns, especially within the Northeastern region. This research utilises an extensive dataset sourced from the healthcare cost and utilisation project (HCUP), covering the years 2017 to 2020, to employ machine learning in forecasting Lyme disease outbreaks. Utilising the dataset containing 4,867,103 patient records for Lyme disease prevalence, this investigation employs logistic regression within a MATLAB and SAS framework to anticipate Lyme disease outbreaks in the Northeastern US the models validation was conducted using MATLAB through an ROC analysis, comparing various algorithms, and despite random forest exhibiting a higher AUC, logistic regression was selected for its clinical interpretability. In SAS, logistic regressions robustness was verified, with notable reductions in AIC, SC, and -2 Log L, indicating a robust model fit. The general methodology of the study combines predictive accuracy with usability and the importance of interpretable models within healthcare contexts.
  Keywords: Lyme disease; machine learning; random forest; neural networks; predictive modelling; healthcare data; epidemiology; HCUP; Northeastern United States; public health.
  DOI: 10.1504/IJMEI.2025.10072951
   
  
• A systematic review on non-invasive and non-voluntary bio-signals intended for cost-effective depression detection  
  by Rupali Pawar, Prachi Mukherji, Leena Phadke, Sharmishtha Deshpande 
  Abstract: Depression is the common denominator of mental affliction and can result in a large medical and socio-economic burden. The extensive investigation of bio-signals to find biomarkers to aid the accurate detection and diagnosis of depression is ongoing to reduce this load. The non-invasive modalities offer an easy and affordable way to detect depression. Unfortunately, a comprehensive study showing the applicability of non-invasive modalities for depression detection is not available. This article investigates non-invasive modalities to detect depression from multiple perspectives. The study reviewed 76 articles from IEEE Xplore, PubMed, and Web of Science by adopting the PRISMA protocol and Boolean search operators covering ten years from 2013-2023. The review found that EEG was used in 78% of studies, with unimodal studies at 93% and multimodal studies at only 7%. The review suggests creating a multimodal database and probing significant biomarkers from non-invasive modalities to enhance depression detection accuracy.
  Keywords: depression detection; major depressive disorder; MDD; non-invasive; non-voluntary; bio-signals; biomarkers; multimodal; machine learning; deep learning.
  DOI: 10.1504/IJMEI.2025.10073119
   
  
• Proactive ransomware attack detection and enhanced cybersecurity in telehealth services using hybrid model  
  by Maddila Suresh Kumar, Satyanarayana Botsa 
  Abstract: The rise of telehealth services has revolutionised healthcare accessibility but introduced new cybersecurity challenges, particularly ransomware attacks. This study investigates a hybrid model combining machine learning techniques and static code analysis to proactively detect these threats. Data was cleaned and pre-processed using a median filter and transformed through extract-load-transform (ELT) pipelines. The proposed model integrates a blockchain framework with a cascaded convolutional neural network (C-CNN), implemented in Python for real-world applicability. Results showed that with a learning rate of 0.01, the model achieved 90% accuracy, outperforming the 81% accuracy of the 0.1 learning rate, indicating strong potential for proactive ransomware defence in telehealth.
  Keywords: ransomware attack; cybersecurity; telehealth; blockchain model; extract-load-transform; cascaded convolutional neural network; C-CNN.
  DOI: 10.1504/IJMEI.2025.10073171
   
  
• A new enhanced deep learning approach for Parkinsons disease diagnosis in MRI brain images  
  by B. Dhanalaxmi, Madadi Vijaya Kamal, Narendhar Mulugu, G. N. Beena Bethel, V. Srilakshmi 
  Abstract: Parkinsons disease (PD) is currently one of the most common neurological conditions. In this research, we have proposed a new, enhanced deep-learning approach for PD classification and segmentation. This research proposes an enhanced InceptionV3 (EINceptionV3) model for classification, the interested region of PD is segmented using the improved VGG-19 (IVGG-19) model for segmentation. Then, the image features are extracted using the ResNet-34 model to improve the classification performance. The feature selection process is performed using the Chaotic sand cat optimisation algorithm (CSCOA). The publicly available MRI image-based PD database is used to analyse the experiment.
  Keywords: Parkinson’s disease; PD; deep learning; classification; segmentation; enhanced INceptionv3; improved VGG-19.
  DOI: 10.1504/IJMEI.2025.10073204
   
  
• Multimodal biomedical imaging for patchy area segmentation in complex brain disease diagnosis using machine learning approach  
  by Anil Zumberlal Chhangani, Aejazul Hasan Khan, Abhishek A. Vichare 
  Abstract: This study combines MRI and PET for accurate brain disease segmentation, aiming to enhance diagnosis and treatment planning. The system utilises advanced pre-processing techniques like adaptive optimum weighted mean filter (AOWMF) and stacked auto encoder algorithms to improve imaging quality and reduce noise. A deep convolutional spiking capsule neural network classifier is introduced to categorise segmented brain areas into disease categories. The method achieves 99.9% accuracy in segmenting patchy areas, demonstrating superior performance in both speed and precision, significantly improving diagnostic accuracy and providing valuable insights for personalised therapeutic interventions in neurology and psychiatry.
  Keywords: multimodal biomedical image; brain disease diagnosis; machine learning; patchy area segmentation; magnetic resonance imaging; MRI; positron emission tomography; PET.
  DOI: 10.1504/IJMEI.2025.10073341
   
  
• Detection and prediction of diabetes mellitus through predictive analysis using ensemble methods in machine learning and deep learning  
  by Prema Subhash Kadam, Sachin P. Godse, Parikshit N. Mahalle 
  Abstract: This review explores advancements in diabetes mellitus (DM) detection and prediction using machine learning (ML) and deep learning (DL) techniques. Ensemble methods like bagging, boosting, and stacking enhance predictive accuracy, while DL models such as CNNs and RNNs effectively capture complex patterns. The role of publicly available datasets and cross-validation in ensuring model reliability is emphasised. Challenges like data quality, model interpretability, and computational demands are discussed, with potential solutions. Future research focuses on hybrid models combining ML and DL strengths and personalised approaches for more effective diabetes management.
  Keywords: diabetes mellitus detection; diabetes prediction; machine learning; ML; deep learning; DL; ensemble-based prediction.
  DOI: 10.1504/IJMEI.2025.10073342
   
  
• An innovative squeeze-and-excitation CNN for prostate cancer histopathology Gleason grading  
  by Maulika Patel, Parag Sanghani, Niraj Shah 
  Abstract: Accurate Gleason grading is critical for diagnosing prostate cancer and guiding treatment decisions. Traditional manual evaluation of histopathological images is often subjective and inconsistent, highlighting the need for reliable automated solutions. This study introduces a deep learning model based on a squeeze-and-excitation convolutional neural network (SE-CNN) to perform automated Gleason grading. By integrating SE blocks into standard CNN architectures, the model effectively emphasises key image features, improving its ability to distinguish subtle differences between Gleason patterns. Trained on a custom dataset of prostate cancer images, the SE-CNN achieved a 92% classification accuracy, outperforming baseline CNN and ResNet models. The performance demonstrates the models potential to deliver consistent and interpretable results, reducing diagnostic variability and enhancing clinical decision-making. The proposed SE-CNN framework offers a robust and efficient approach to prostate cancer grading, supporting pathologists with a scalable tool for improved accuracy and workflow efficiency in pathology practices.
  Keywords: prostate cancer; Gleason grading; histopathology; squeeze-and-excitation CNN; SE-CNN; deep learning.
  DOI: 10.1504/IJMEI.2025.10073436
   
  
• XAI-endowed predictive phenotyping data in healthcare for unveiling hidden patterns  
  by Arvind Uttiramerur 
  Abstract: This study explores the integration of explainable artificial intelligence (XAI) in predictive phenotyping for healthcare, aiming to uncover hidden patterns in medical datasets. It uses dimensionality reduction via principal component analysis (PCA) and feature extraction through LIME and SHAP to enhance model interpretability. XAI models, including XNNs, random forests, gradient boosting machines (GBM), and long short-term memory networks (LSTM), are developed to ensure high predictive accuracy while maintaining transparency. The highest accuracy achieved is 95.445%. This approach promises to revolutionise disease diagnosis, prognosis, and treatment by providing interpretable insights from complex healthcare data.
  Keywords: explainable artificial intelligence; XAI; phenotyping data; healthcare; unveiling hidden patterns; principal component analysis; PCA; local interpretable model-agnostic explanations; LIME.
  DOI: 10.1504/IJMEI.2025.10073509
   
  
• Anomaly detection network in internet of medical things using upgraded inception gated recurrent unit with confidence scoring  
  by Anilkumar Vishwanath Brahmane, Bhagyashree Ashok Tingare, Archana Ajit Chaugule, Pratik Sudhakar Chopade, Vijay Arun Kotkar, Rashana Avinash Golande 
  Abstract: Anomaly detection in medical imaging is critical for early disease diagnosis and clinical decision-making. However, existing deep learning models often struggle with subtle abnormalities, high false-positive rates, and limited interpretability, particularly in complex and high-dimensional medical datasets. This study proposes a novel deep learning framework that integrates a deep convolutional bidirectional modified variational autoencoder (DCBiMVAE) for robust feature extraction and an inception gated recurrent unitmulti-layer perceptron (IGRU-MLP) for accurate anomaly detection. The framework is optimised using an improved bald eagle search optimisation (IBESO) algorithm, and a confidence prediction network (CPN) is employed to assign confidence scores to the predictions, reducing uncertainty and false alarms. Evaluated on X-ray and MRI datasets, the proposed model achieves 96.12% AUC, 95.33% sensitivity, and 98.01% specificity, outperforming state-of-the-art methods. This work demonstrates the potential of a hybrid, confidence-aware approach for reliable and automated anomaly detection in medical imaging.
  Keywords: convolution kernel weight; anomaly identification; univariate Gaussian distribution; confidence interval parameter; threshold function; Brownian motion.
  DOI: 10.1504/IJMEI.2025.10073557
   
  
• Adaptive brain tumour segmentation with dilated residual DenseNet-attention mechanism-aided multi-classification model on analysis of 3D MRI brain images  
  by D. Ramya, C. Lakshmi 
  Abstract: This work presents an advanced deep-learning mechanism for brain tumour segmentation and classification models to provide highly accurate outcomes. The essential 3D MRI images are given into the developed adaptive Trans-UNet3+ (ATUNet3+) for segmenting the brain tumour regions. From ATUNet3+, the parameter tuning is executed with the support of the enhanced election-based optimisation algorithm (EEBOA). The segmented outcomes are forwarded to the dilated residual DenseNet with attention mechanism (DRDenseNet-AM) for classification. The experimental analysis is performed to analyse the efficacy of the suggested scheme by contrasting it with the traditional approaches. The classification accuracy of the implemented DRDenseNet-AM-based brain tumour classification model is 91.35%, which is greater than the previously developed convolutional neural network (CNN) models.
  Keywords: brain tumour segmentation and multi-classification; MRI brain images; dilated residual DenseNet with attention mechanism; enhanced election-based optimisation algorithm; EEBOA.
  DOI: 10.1504/IJMEI.2025.10073741
   
  
• An automated parameter tuning twin support vector machine model for diabetic retinopathy risk prediction in patients with type 2 diabetes mellitus  
  by Mayuri Diwakar Kulkarni, Shailesh S. Deore 
  Abstract: This study presents machine learning-based prediction models for faster and more accurate diabetic retinopathy risk prediction in patients with type 2 diabetes. To enhance model performance, the input data are first enhanced using the synthetic minority oversampling technique (SMOTE) and the generative adversarial network (GAN) model. Next, the input data are pre-processed using min-max normalisation and discretisation to remove unnecessary data and improve learning efficiency. Then, the features are selected utilising the adaptive red fox optimisation (ARFO) algorithm with improved accuracy. Finally, the risk of diabetic retinopathy in type 2 diabetes mellitus patients is predicted using a new automated parameter tuning-driven twin support vector machine (APT-TSVM) model. Additionally, the prediction parameters are optimally fine-tuned using an enhanced spotted hyena optimisation (ESHO) algorithm. The obtained simulation results show that the proposed model outperforms other existing approaches with accuracy (98%) for the nomogram dataset and (97.08%) for dynamic nomogram datasets.
  Keywords: synthetic minority oversampling technique; SMOTE; machine learning; generative adversarial network; GAN; adaptive red fox optimisation algorithm; ARFO; twin support vector regression; spotted hyena optimisation algorithm; ESHO.
  DOI: 10.1504/IJMEI.2025.10073742
   
  
• Automated brain tumour segmentation and classification using twin attention-based dropout network  
  by G.B. Umesha, T. Manjula, K.S. Babu 
  Abstract: Timely and accurate diagnosis of brain tumours is crucial for improving treatment outcomes and increasing patient survival rates. However, manual diagnosis is time-consuming and often prone to subjectivity. To address these challenges, this study introduces an advanced computer-aided diagnosis system that leverages deep learning for brain tumour segmentation and classification. MRI scans from the publicly available BRATS dataset are first pre-processed to remove noise and normalise image intensity. Tumour segmentation is then performed using a position attention-based modified convolutional U-Net (PCU-net), which enhances the models focus on spatially relevant regions and improves segmentation accuracy. For classification, a novel twin attention-based dense bidirectional gated dropout network (TA-DeBiDr) is proposed to capture both contextual and sequential patterns while preventing overfitting. The proposed model achieves a classification accuracy of 99.25%, precision of 98.41%, recall of 100%, and an F1-score of 98.51%. These results demonstrate the effectiveness of the proposed approach and its potential to support radiologists in real-time clinical decision making.
  Keywords: filtering; segmentation; twin attention; channel attention; overfitting; dropout; fine-tuning.
  DOI: 10.1504/IJMEI.2025.10073743
   
  
• Optimised deep intelligent system for arrhythmia prediction from electrocardiogram data  
  by Radhika Nampalli, Sujatha Dandu 
  Abstract: A novel red panda radial basis prediction system (RPRBPS) is developed for detecting abnormal heart rhythms from ECG signals. This system uses ECG data from Kaggle and undergoes a filtering phase to remove artefacts and noise. Feature selection process to identify most important characters using red pandas best solution. The RPRBPS model detects heart rate abnormalities and classifies specific conditions, assisting medical professionals in early diagnosis and intervention. The metrics like precision, recall, accuracy, F-score, error rate, and accuracy was measured. The proposed model attained 0.001% prediction error rate, which demonstrates the proposed RPRBPS model remarkable disease detection performance.
  Keywords: disease classification; filtering ECG signal; red panda optimisation; arrhythmia prediction.
  DOI: 10.1504/IJMEI.2025.10073822
   
  
• Secure and efficient kidney image classification using AI and deep learning techniques  
  by Anil Zumberlal Chhangani, Aejazul Hasan Khan, Harish Keshav Barapatre 
  Abstract: The rapid advancement of artificial intelligence (AI) and deep learning techniques has transformed various fields, including healthcare and medical imaging. Utilising public datasets like the kidney dataset in open access and NIH image collections, automates image pre-processing to enhance quality. A deep learning model is optimised using Bayesian algorithms to improve performance without manual tuning, while neural architecture search (NAS) identifies the best neural network design. Federated Learning allows for distributed training across medical institutions without sharing sensitive data. The implementation in Python achieves a 96% accuracy rate in kidney image classification. This research promises significant advancements in secure, efficient, and scalable diagnostic tools for healthcare, with broad future potential.
  Keywords: kidney image classification; deep learning optimisation; artificial intelligence; neural architecture search; NAS; federated learning; Bayesian optimisation algorithms.
  DOI: 10.1504/IJMEI.2025.10073845
   
  
• Preoperative predictors of severe postoperative pain in orthopaedic surgeries - a hospital based observational prospective study  
  by Sameer Haveri, Varisth Vardhan, Farhana Tahseen Taj 
  Abstract: This prospective observational study aimed to identify preoperative predictors of severe postoperative pain in patients undergoing major orthopaedic surgeries. Conducted in a tertiary care hospital in India, the study included 197 patients. A structured 20-point preoperative questionnaire was used, and pain was assessed using the visual analogue score (VAS) scale on postoperative days 0 to 2. Severe postoperative pain (VAS 75) was recorded in 107 patients. Statistically significant predictors included severe preoperative pain, emergency surgery, younger age, female sex, higher body mass index, diabetes, smoking, anxiety, depression, general anaesthesia, longer surgical duration, and larger incision size. The findings highlight the multifactorial nature of postoperative pain and underline the need for early identification of high-risk patients. Incorporating these predictors into clinical protocols may support the development of personalised pain management strategies, ultimately improving patient recovery and outcomes following orthopaedic surgery.
  Keywords: risk factors; postoperative pain; preoperative predictors; orthopaedic surgeries; pain management; visual analogue score scale; VAS scale.
  DOI: 10.1504/IJMEI.2025.10073984
   
  
• Optimising the identification of Parkinsons disease diagnosis with machine learning in healthcare AI  
  by Girish Jaysing Navale, Archana Tukaram Bhise, Kishor Sadashiv Wagh 
  Abstract: Parkinsons disease (PD) is a neurodegenerative disorder causing motor and cognitive impairments. Early diagnosis is critical but often delayed due to subjective observations and overlapping symptoms with other conditions. This study develops machine learning (ML) algorithms for early PD detection, improving diagnostic accuracy and sensitivity. Techniques like DWSAF reduce noise for precise classification, while CPSGA optimises feature selection for a more efficient and personalised diagnosis. The recurrent neural network-support vector machine (RNN-SVM) model captures temporal patterns for early detection. These advancements ensure scalable, adaptable ML models for enhanced diagnosis and integration into clinical decision support systems (CDSS). The results show that PD patients have a higher motor score (6.5) compared to controls, with advanced stages showing an even higher score (16.5). Accuracy and processing time remain consistent across groups. Future ML advancements will improve early detection, accuracy, and personalised treatment.
  Keywords: Parkinson’s disease; diffusion wiener spline adaptive filtering; DWSAF; support vector machine; SVM; chaotic particle swarm genetic algorithm; CPSGA; recurrent neural network; RNN; early diagnosis.
  DOI: 10.1504/IJMEI.2025.10073985
   
  
• A framework for classifying thyroid cancer using APPO_HyRZNet from ultrasound images  
  by T. Merlin Jaba, Lipsa Nayak 
  Abstract: A novel deep learning method for thyroid cancer classification is proposed in this work using ultrasound images. Initially, the image pre-processing is done using an adaptive bilateral filter. After that, the cancer region is detected using a pyramidal attention based network (PAY-Net) with artificial protozoa pelican optimisation (APPO). Here, the APPO is developed by the combination of artificial protozoa optimisation (APO) and pelican optimisation algorithm (POA). Further, features, like gray level co-occurrence matrix (GLCM), complete local binary pattern histogram (CLBP), and local vector pattern (LVP) are extracted. Then, thyroid cancer classification is done by hybrid deep learning hybrid residual zeiler and fergus network (HyRZNet). Moreover, the hyperparameter of the HyRZNet will be tuned based on the proposed APPO. Moreover, the proposed APPO_HyRZNet attained better true positive rate (TPR), accuracy, and true negative rate (TNR) of 95.886%, 92.986%, and 93.998%.
  Keywords: thyroid cancer; thyroid cancer detection; artificial protozoa pelican optimisation; APPO; deep learning; ultrasound images.
  DOI: 10.1504/IJMEI.2025.10074000
   
  
• A review on deep learning models for speech emotion recognition: challenges, databases, features and performance analysis  
  by Srinivas Pichuka Veera Venkata Satya, Asritha Aravapalli, Dushyanth Chalasani, Oswell Sianyengele, Vignesh Suprith Busetty, Sampath Patchigolla 
  Abstract: Emotions play a crucial role in human relationships and have a big impact on consumer satisfaction and opinion. The development of applications involving human-computer interaction (HCI) also heavily relies on speech emotion recognition (SER) modules. Over the past few decades, a huge amount of SER systems have been developed. Its been demonstrated that various deep learning algorithms are effective instruments for extracting information from audio-visual content that is unevenly time-distributed. This articles effectiveness is increased by providing a review of current enhancements of deep learning models in SER, details about databases, the difficulties with SER, its uses, and suggestions for the future.
  Keywords: deep neural networks; DNNs; recurrent neural networks; gated recurrent unit; GRU; capsule network; long short-term memory; LSTM human-computer interaction; HCI; speech emotion recognition; SER.
  DOI: 10.1504/IJMEI.2025.10074137
   
  
• Deep learning and transfer learning techniques for glaucoma detection via analysing green channels on SMDG-19 fundus imagery  
  by A. Anushya, Nawal A. Alonezi, Meiad Jrad Aljrad, Kholood Mubark, Afrah Naif Eid Alsagri Alsagri, Awatef Alreshidi 
  Abstract: The study uses transfer learning on the SMDG-19 dataset to explore glaucoma detection in fundus images. It examines the performance of vision transformer, EfficientNet, and DenseNet201, across training, validation, and testing phases enhanced by contrast limited adaptive histogram equalisation (CLAHE) image pre-processing. Although vision transformer flashed superior training accuracy, it floundered on validation and test sets. EfficientNet delivered stable results with an insignificant dip in test accuracy. DenseNet201 achieved the highest accuracy, sensitivity, and specificity, however, it needed 40 epochs for fine-tuning. An ensemble of all models was used to balance execution and shrink computational load.
  Keywords: contrast limited adaptive histogram equalisation; CLAHE; DenseNet201; EfficientNet; glaucoma detection; vision transformer; ViT.
  DOI: 10.1504/IJMEI.2025.10074292
   
  
• A multimodal image feature-fusion using novel vision transformer model for Alzheimers disease classification  
  by P. V. V.S. Srinivas, Ziya Afreen, Vadlamudi Saranya, Chandu Uma Mahesh, Pasupuleti Lokesh Babu 
  Abstract: One of the most common neurodegenerative disorders affecting elderly people is Alzheimers disease (AD). Since there is no known cure, it is crucial to diagnose the disease as early as possible. This study introduces a novel multimodal image feature fusion method to fuse and learn feature representation for classifying AD. Initially, the images are collected and pre-processing is done to remove the noise by exponentially modified Gaussian filter (EMGF). Next, feature extraction is carried out by employing a fusion network attention integrated shallow latent Gaussian variational autoencoder (FA-SLGV). After that, both the MRI and PET image features are fused and given as input to the classifier, which classifies the AD by the optimised coordinate attention-based modified vision transformer model (OCA-MVT). Finally, hyperparameters of the model are fine-tuned using dung beetle optimisation (DBO). The experimental results show that the proposed approach attains an accuracy of 98.1% demonstrates promising performance in classification.
  Keywords: multimodality brain images; Alzheimer’s disease; improved vision transformer; Gaussian variational autoencoder.
  DOI: 10.1504/IJMEI.2025.10074338
   
  
• Advanced feature selection in cardiovascular disease using improved relief extension feature selection and classification techniques with gradient boosting classifier  
  by G. Ezhilvani, G. Thailambal 
  Abstract: Advanced MRIs and CTC scans can treat cardiovascular heart disease. Electronic equipment involved in medical issues generates a significant amount of data. Healthcare has a more complex disease landscape and hidden information, making decisions to end diseases and disorders difficult. IoT sensors detect this serious heart problem and aid in its diagnosis and used existing methods to improve AI-based approaches like machine learning. The aggregated properties of multiple forests were classified using the novel extra trees classifier (NETC) to generate minimal judgements. Kaggles cardiovascular dataset was pre-processed to train unique classes and handle many decisions to help choose the appropriate settings. A strategy is proposed to address the research gap in feature prediction by selecting precise features using fine-tuned hyperparameters. We use NETC hyperparameters to fine-tune relevant feature selection using improved relief extension feature selection (IREFS). Relief a cardiovascular dataset selection operator uses a gradient boosting classifier to gather features during training. Implemented measures include accuracy over other methodologies, performance analysis score, confusion matrix, threshold range, ROC curve, precision prediction ratio, and instance-based recall identification. Performance analysis using the CVD dataset data showed CVD disease prediction validation accuracy of 0.9583.
  Keywords: artificial intelligence; gradient boosting classifier; GBC; heart disease; improved relief extension feature selection; IREFS; machine learning; novel extra trees classifier; NETC; prediction.
  DOI: 10.1504/IJMEI.2025.10074377
   
  
• State of the art survey on brain tumour classification and detection using MRI images in QFL  
  by P. Rajyalakshmi, Namineni Gireesh 
  Abstract: This study focuses on highlighting the strengths and limitations of previous classification methods. Initially, this survey outlines the general process of brain tumour (BT) classification using MRI images. Subsequently, the classification techniques are divided into five main categories: machine learning (ML)-based, deep learning (DL)-based, optimisation-based, hybrid methods, and others. This study is based on an analysis of 50 articles that focus on various methodologies employed for BT classification. This research identifies that DL modules are frequently used techniques in numerous studies. Finally, the evaluation is categorised by publication year, research technique, performance measures, and the success of methodologies.
  Keywords: federated learning; FL; quantum federated learning; QFL; deep learning; DL; transfer learning; TL; machine learning; ML.
  DOI: 10.1504/IJMEI.2025.10074559
   
  
• Classification of COVID-19 using multi-scale fuzzy entropy model: application in medical imaging  
  by S. Salini 
  Abstract: In order to measure and describe lung tissues that have been affected by dangerous diseases like coronavirus disease 2019 (COVID-19), radiologists and doctors in general need the right information. Texture-based analysis has been used in the setting of different lung diseases. Making sure that COVID-19 patients are properly separated from healthy subjects and patients with other breathing diseases is very important. Our goal with COVID-19 and idiopathic pulmonary fibrosis (IPF) is to measure the changes that happen in the lungs that come with these two lung diseases. The three-dimensional multiscale fuzzy entropy (MFE3D) method is what we recommend as a way to reach this goal. The three groups that were looked at were COVID-19 patients, IPF patients, and healthy people. We can tell the difference between healthy subjects and COVID-19 patients by using a complexity index (CI) that is found by adding up the entropy numbers.
  Keywords: classification of COVID-19; data pre-processing; feature extraction.
  DOI: 10.1504/IJMEI.2025.10074573
   
  
• Novel framework for timely and precise diagnosis of retinopathy of prematurity using deep learning models  
  by G. Hubert, S.Silvia Priscila 
  Abstract: Retinopathy of prematurity (ROP) is a critical eye disease influencing early birth children, requiring primary and exact diagnosis to avoid loss of vision. Initial-stage diagnosis is crucial for preventing sight issues, and automated ROP image separation is a significant step in developing a consistent analytical device. This research recommends a healthy ROP diagnosis structure using innovative image pre-processing, separation, feature retrieval, and grouping approaches. Primarily, ROP images are subjected to bilateral filtering to reduce noise while preserving edge values, followed by adaptive histogram equalisation (AHE) to enhance contrast. Segmentation is achieved using a U-Net approach to segregate pertinent retinal features. Features are then mined using the AlexNet deep learning (DL) framework. To optimise the classification task, apply the Emperor Penguin optimiser (EPO), which boosts AlexNets performance by fine-tuning its attributes. The combination of AlexNet with EPO significantly outperforms conventional optimisation models, such as the genetic algorithm (GA) and particle swarm optimisation (PSO), in terms of accuracy, sensitivity, and specificity values. This optimised design provides a more effective tool for the early recognition of ROP, potentially yielding more accurate results for affected children.
  Keywords: retinopathy of prematurity; ROP; adaptive histogram equalisation; AHE; AlexNet and U-Net approach; particle swarm optimisation; PSO; Emperor Penguin optimiser; EPO; genetic algorithm; GA.
  DOI: 10.1504/IJMEI.2025.10074617
   
  
• Enhanced deep learning-based method for early-stage Parkinsons disease diagnosis  
  by Kumbham Bhargavi, Soanpet Sree Lakshmi, G.Uma Devi 
  Abstract: The neurological condition known as Parkinsons disease (PD) affects the nerve cells. To develop an accurate and reliable early-stage PD detection model, we present an improved deep learning-based approach in this research. The speech signal features are correctly predicted using the improved ResNet-34 (IResNet-34) model to determine whether PD is present. The most pertinent features needed for PD detection are chosen using the enhanced remora optimisation algorithm (EROA). To deal with unbalanced data, the synthetic minority oversampling technique (SMOTE) is applied. The results showed that the proposed IResNet-34-based technique shows higher performance than existing models, with an accuracy of 98.72%.
  Keywords: Parkinson’s disease; PD; speech signal attributes; correlation matrix; visualisation; IResNet-34.
  DOI: 10.1504/IJMEI.2025.10074669
   
  
• Privacy-enhanced federated learning for multilingual speaker identification in IoT driven disaster management system  
  by Nijara Kalita, Aniruddha Deka 
  Abstract: This research proposes a federated learning-assisted hybrid DL model for emergency speaker recognition that addresses these difficulties. This article has three sections: pre-processing, feature extraction, and identification. Adaptive discrete wavelet transform pre-processes the input signal. The complicated pre-processing method adaptively decomposes the signal to increase compression, noise reduction, and feature quality. Distilled long short-term assisted audio Bert (DisLSBERT) extracts the feature. This feature extraction model created a rich feature vector comprising audio signal temporal and contextual dynamics. The signal is supplied into federated training to generate a local model and update the global model weight after feature extraction. The global aggregated the weight and passed it to the local model to identify the speaker in an emergency. The Hy-RegVIT hybrid region distillation-based residual transformer model does this. This federated learning method protects user data. The speaker identification experiment uses South Korean and LibriSpeech datasets attains accuracy, precision 98.79%, and 97.12%.
  Keywords: speaker identification; federated learning; wavelet transform; audio Bert; region distillation; vision transformer.
  DOI: 10.1504/IJMEI.2025.10074688
   
  
• AB-CLOA: advanced Bi-LSTM with chaotic lyrebird optimisation for arrhythmia disease segmentation and classification  
  by Chanchal G. Agrawal, Nilesh J. Uke 
  Abstract: This study presents an advanced framework for ECG and PCG classification using the PhysioNet/CinC and EPHNOGRAM datasets. The upgraded fractional Daubechies wavelet (UpFDW) model enhances signal quality, while Gramian angular field (GAF) transformation aids segmentation via extended U-Net (EU-Net). A bi-directional LSTM optimised by the chaotic lyrebird optimisation algorithm (CLOA) refines hidden neuron selection, improving classification accuracy. The model achieves 99.1% accuracy on PhysioNet/CinC and 98.8% on EPHNOGRAM, demonstrating high reliability. This approach enhances cardiac abnormality detection, providing a promising solution for automated arrhythmia diagnosis and potential real-world clinical applications, ensuring improved patient monitoring and care.
  Keywords: arrhythmia disease; electrocardiograms; ECG; PCG; wavelet transform; Gramian angular field; GAF; U-Net; bi-directional long short term memory; lyrebird optimisation.
  DOI: 10.1504/IJMEI.2025.10074910
   
  
• Enhancing Alzheimer's disease diagnosis through deep learning  
  by Sonali Deshpande, Nilima Kulkarni 
  Abstract: Alzheimers disease remains a significant healthcare issue that demands proactive risk reduction, early intervention, and timely, accurate diagnosis. Literature highlights the need for early and accurate detection of Alzheimers disease. This study modified a deep learning method to diagnose Alzheimers disease. Using the ADNI and Kaggle datasets, this model is trained on a large dataset from trusted sources and pre-processed for disease identification. The accuracy of 98.13% and 98.42% is obtained on both datasets respectively. Standard evaluation measures showed that the model outperformed conventional techniques. This deep learning approach has practical applications. Future studies should use real-time clinical data to evaluate these in varied datasets.
  Keywords: deep learning; Alzheimer’s disease detection; convolutional neural network; CNN; neurological diseases; medical imaging; ADNI dataset; Kaggle dataset.
  DOI: 10.1504/IJMEI.2025.10075133
   
  
• Improving early diagnosis of skin cancer: a deep learning approach with VGG16  
  by Yassmina Saadna, Saliha Mezzoudj 
  Abstract: Skin cancer poses a significant health challenge due to its increasing prevalence and the difficulty in distinguishing melanoma from nevus. This study investigates deep learning techniques for skin cancer classification, utilising the VGG16 architecture in two roles: as a classifier and as a feature extractor. The HAM10000 dataset is employed for training and testing, with fine-tuning, re-sampling, and decision tree methods enhancing performance. Results show that fine-tuning VGG16 achieves 88.6% accuracy, highlighting the potential of deep learning to improve early detection and diagnosis, thereby supporting dermatologists in combating this critical disease.
  Keywords: skin cancer; VGG16; HAM1000; transfer learning.
  DOI: 10.1504/IJMEI.2025.10075134
   
  
• Automated schizophrenia detection using deep lightweight features extracted from time series EEG image  
  by Hardik Thakkar, Bikesh Kumar Singh, Saurabh Gupta, Sai Krishna Tikka, Lokesh Kumar Singh 
  Abstract: Schizophrenia (SCZ) has become a serious mental condition which is affecting the daily life of the patients. The researchers and psychiatrist are supporting and discovering new methods using electroencephalogram (EEG) signals for early diagnose of SCZ symptoms. In this study, an automated SCZ detection method is proposed based upon the time series EEG image (TSEI) using the deep lightweight features. The resting state 19-channel EEG signal of 24 healthy control (HC) and 24 SCZ patients is acquired and pre-processed. Further, each acquired EEG signal is segmented and converted into an RGB image, i.e., TSEI. A deep convolution neural network, i.e., ShuffleNet-based feature extraction technique is proposed. Significant features are selected and applied to support vector machine model with different kernel combination for detection of SCZ. The highest classification accuracy achieved with proposed methodology is 99.31%. The proposed methodology outperforms exiting techniques for time series EEG based SCZ detection.
  Keywords: schizophrenia; SCZ; EEG signal; machine learning; ML; transfer learning.
  DOI: 10.1504/IJMEI.2025.10075339
   
  
• Optimising orthopaedic diseases prediction with hybrid machine learning model  
  by S. Vairachilai, Sambhav Jain, Telugunti Akhil, Xiomara Patricia Blanco Valencia, S.P. Raja 
  Abstract: This study aims to use six features to develop a machine-learning model to predict orthopaedic disease. The vertebral column dataset related to orthopaedic conditions was divided into training and testing subsets. The model uses these features to predict the disease status of new patients. Our findings indicated that the logistic regression algorithm achieved the highest performance, with an accuracy of 83.10%, precision of 72.7%, recall of 88.9%, an F1-score of 80.00%, and an AUC-ROC of 0.94. In ensemble learning algorithms, AdaBoost and ensemble stacking had the best performance, with an accuracy of 87.10%, a precision of 72.7%, a recall of 88.9%, an AUC-ROC of 0.94, and an AUC-ROC of 0.94. In deep learning algorithms, multilayer perceptron had the best performance, with an accuracy of 79.0%, a precision of 63.2%, a recall of 66.7%, a 64.9%, and an AUC-ROC of 0.88.
  Keywords: orthopaedic disease; machine learning algorithm; ensemble algorithm; deep learning algorithm.
  DOI: 10.1504/IJMEI.2025.10075374
   
  
• Design and analysis of 3D U-Net-based automated chest CT image segmentation and classification  
  by S. Salini 
  Abstract: The 2019 coronavirus disease, also known as COVID-19, has quickly spread around the globe. In order to determine how the disease will progress, it is very essential to do a rapid and accurate automated segmentation of COVID-19 infected areas on chest computed tomography (CT) scans. We present a convolutional neural network that we have given the name 3D CU-Net and which is capable of automatically detecting COVID-19 contaminated areas from 3D chest CT scans. The architecture of 3D U-Net provides the basis for the development of 3D CU-Net. We build a pyramid fusion module with larger convolutions at the very end of the encoding process in order to merge information about high-level qualities. The results of the tests indicate that 3D CU-Net is capable of producing satisfactory results in terms of segmentation. In areas infected with COVID-19 and the lung, the Dice similarity coefficients were, respectively, 96.3% and 77.8%.
  Keywords: coronavirus disease 2019; COVID-19; 3D chest CT images; 3D CU-Net; automatic segmentation.
  DOI: 10.1504/IJMEI.2026.10075484
   
  
• Muscular energy and fatigue: possible metrics to evaluate focal forearm dystonia  
  by Prathisha Mani, Kirthi Mariappan, Akilan Arulselvam, Kalpana Ramakrishnan 
  Abstract: Focal forearm dystonia is a movement disorder involving involuntary muscle contractions. We analysed EMG signals from healthy and dystonic subjects at rest, during voluntary movement, and under increasing load. Muscular energy was computed using the Morlet wavelet transform, while fatigue was estimated using Katz fractal dimension. Dystonic muscles showed higher wavelet-based energy and lower fractal dimension under load, suggesting increased fatigue and reduced signal complexity. These results indicate an inverse relationship between energy and complexity in dystonia. Our findings highlight the potential of combining wavelet energy and fractal metrics as objective EMG biomarkers to assess fatigue and guide rehabilitation strategies.
  Keywords: fractal dimension; wavelet transform; electromyography; EMG; physical activity.
  DOI: 10.1504/IJMEI.2026.10075485
   
  
• Hybrid fuzzy logic modelling for early diagnosis of cognitive function disorders for operators attention in human-machine systems  
  by Nikolay A. Korenevskiy, Riad Taha Al-Kasasbeh, Sofia Nikolaevna Rodionova, Ksenia Razumova, Sergey Filist, Osama M. Al-Habahbeh, Ashraf Adel Shaqadan, Safaa Saber Salah Alghaswyneh, Mohammad Al-Jundi, Etab Tah Al-Kasasbeh, Mahdi Salman Alshamasin, Altyn Amanzholovna Aikeyeva 
  Abstract: This study aims to improve the early diagnosis of attention-related cognitive disorders using a hybrid fuzzy logic approach integrated with a bio-signal monitoring device. It addresses the limitations of conventional cognitive assessments, which are often inadequate for early prediction in human-machine system operators. The proposed hybrid fuzzy logic model evaluates cognitive functions such as attention and memory using bio-indicators reflecting psycho-emotional stress, fatigue, energy imbalance at biologically active points, and functional reserves. A mathematical framework was developed to estimate the probability of early-stage disorders, focusing on attention metrics including concentration, selectivity, distribution, and stability. Statistical analysis of verification samples from computer operators showed a diagnostic confidence level above 0.91, particularly for attention concentration. This multidisciplinary work combining psychophysiology, neurology, engineering psychology, and psychiatry enhances forecasting, early detection, and classification of cognitive states, supporting timely prevention and correction of emerging cognitive dysfunctions.
  Keywords: human-machine systems; biological active points; cognitive function; classification; disorders; psycho-emotional stress; forecasting.
  DOI: 10.1504/IJMEI.2026.10075602
   
  
• Brain tumour classification: a transfer learning approach with pooling layer optimisation  
  by Soumyarashmi Panigrahi, Dibya Ranjan Das Adhikary, Binod Kumar Pattanayak 
  Abstract: Brain tumours are a deadly illness with high mortality rates worldwide. Automated detection methods are crucial for early diagnosis. CNNs are effective, but require substantial computational resources. To address this, we proposed a modified EfficientNetB7 architecture with added layers, including various pooling layers. We analysed the effectiveness of MaxPooling, AveragePooling, GlobalMaxPooling, and GlobalAveragePooling layers on pre-trained CNN models like VGG19 and ResNet50V2 using the Br35H: Brain Tumor Detection 2020 dataset. Our proposed model with MaxPooling achieved the highest accuracy of 99.66%. These results significantly improve diagnostic accuracy and computational effectiveness, supporting decision-making processes.
  Keywords: brain tumour; convolutional neural networks; CNNs; MaxPooling; AveragePooling; GlobalMaxPooling; GlobalAveragePooling.
  DOI: 10.1504/IJMEI.2026.10075789
   
  
• An optimised ensemble machine learning models for breast cancer detection in mammography images  
  by Sayali Yadavrao Bamble, Nilesh Janardan Uke 
  Abstract: This paper proposes a progressive optimal ensemble machine learning (OEML) model for the early detection of breast cancer. Initially, the mammography images are pre-processed to remove noise, and the region of interest for cancer cell detection is identified by using the adaptive histogram thresholding and contour clustering algorithm. The higher-level deep features are then retrieved using the convolutional VGG-16 model to improve the accuracy of detection. Finally, the classification is performed by ensembling machine learning algorithms such as support vector machine, XGBoost, and random forest. As a result, the proposed OEML model outperforms the existing method with an accuracy of 98.37%.
  Keywords: breast cancer; Gaussian bilateral filter; GBF; adaptive histogram thresholding; AHT; convolutional VGG-16; ConV-16; contour clustering ensemble machine learning; pelican optimisation; and classification.
  DOI: 10.1504/IJMEI.2026.10075850
   
  
• Analyses of trends and behavioural risk factors in the health status of African immigrants in the USA  
  by Nana K. Yeboah, Shankar Srinivasan, Dinesh P. Mital, Riddhi Vyas 
  Abstract: The population of African immigrants (AI) in the US has grown significantly, yet their health status remains underexplored. This study analyses NHIS data (2007-2017) using multivariate logistic regression to examine health trends by length of stay (>10 years: 52.9%; <10 years: 47.1%). African immigrants with longer residence were older, more educated, had higher income and health insurance, but exhibited elevated rates of chronic conditions such as diabetes, obesity, and hypertension along with reduced physical activity. Age and socioeconomic factors significantly influence health outcomes. Findings highlight the need for tailored early interventions to tackle AI-specific health risks and lifestyle changes.
  Keywords: immigrants; African; immigrant health; communicable diseases; non-communicable diseases; behavioural risk; odds ratio; USA.
  DOI: 10.1504/IJMEI.2026.10075974
   
  
• Accurate glioma classification using gamma fusion and Gabor filters for histological images  
  by Linda Ait Mohammed, Fatiha Alim-Ferhat, Farid Talbi 
  Abstract: Digital pathology has revolutionised the storage and sharing of biological data, enabling pathologists to diagnose conditions based on extensive quantitative measurements alongside visual observations. Cancer diagnosis typically involves biopsy and histological analysis, necessitating reliable tools to aid decision-making and data interpretation. In this study, we leveraged machine learning and a novel feature extraction technique based on gamma fusion of images and Gabor filter features. The developed algorithm accurately classified gliomas into oligodendroglioma, astrocytoma, and glioblastoma subtypes, achieving a 99% accuracy on the CPM-RadPath 2020 dataset. This approach can serve as a valuable second reader and assistive tool, alleviating the workload on pathologists while enhancing diagnostic accuracy and efficiency.
  Keywords: gliomas; digital pathology; machine learning; Gabor filter; MLP; fusion.
  DOI: 10.1504/IJMEI.2026.10075975
   
  
• Diabetic retinopathy classification through advanced deep learning with GANs  
  by Indu Saini 
  Abstract: Diabetic retinopathy is a significant cause of vision loss and blindness in people with diabetes. Early detection and accurate diagnosis are essential for effective treatment and preventing complications. This study presents a new AI-based method using generative adversarial networks (GANs) to detect diabetic retinopathy from retinal images automatically. GANs generate realistic synthetic photos, increasing the dataset size and improving the models ability to work with new data. The proposed GAN-CNN model combines GANs with convolutional neural networks (CNNs). The CNN is fine-tuned using an expanded dataset containing original and generated images. Fine-tuning involves adjusting model parameters using optimisation techniques like gradient descent and backpropagation. The model is trained on the IDRiD dataset, which includes 516 retinal images from 54 Indian diabetic patients. It achieves 96.5% accuracy on the test dataset, demonstrating strong potential for real-world clinical use in detecting diabetic retinopathy.
  Keywords: deep learning; machine learning; generative adversarial networks; GAN; retinal images; image classification; disease detection; convolutional neural networks; CNNs.
  DOI: 10.1504/IJMEI.2026.10076489
   
  
• Autonomic and behavioural consequences of heat-induced stress  
  by Anjali Kumari, Rakesh Kumar Sinha 
  Abstract: To investigate the changes in behaviour and heart rate variability (HRV) correlation and potential HRV biomarkers under heat stress. 80 young male wistar rats were exposed to control (24 C), 38 C, 40 C or 42 C (n = 20 each). ECG-derived HRV (Kubios) and behaviour (open field, elevated plus maze, Y maze) analysed. Exposure to high temperature increased heart rate, LF, LF/HF and decreased RMSSD, HF along the side of impaired cognition and increased anxiety. Specific HRV indices (LF, HF, LF/HF, RMSSD) reliably reflect neurocognitive decline under heat stress, supporting their use as biomarkers for early detection of heat-induced cognitive impairment.
  Keywords: heat stress; heart rate variability; rats; cognitive behaviour; BOD incubator; electrocardiogram.
  DOI: 10.1504/IJMEI.2026.10076490
   
  
• Deep learning and attention mechanisms for speech emotion recognition: a review  
  by Irfan Abdullatif Chaugule, Satish R. Sankaye 
  Abstract: Speech emotion recognition (SER) is an essential component of modern speech processing and human-computer interaction in recent times. It seeks to identify emotions embedded in speech by analysing acoustic and spectral features. Despite rapid advances, SER faces challenges related to noise, variability, and feature selection. This paper reviews recent developments in SER, emphasising deep learning models, attention mechanisms, and higher-order spectral analysis. The survey organises studies by pre-processing, feature extraction, and classification techniques, and evaluates commonly used performance metrics. Finally, it highlights research gaps and outlines future directions for building robust, multi-modal, and generalisable SER systems suitable for real-world applications.
  Keywords: speech emotion recognition; SER; k-nearest neighbourhood; Gaussian mixture models; GMMs; automatic speaker recognition; convolutional neural network.
  DOI: 10.1504/IJMEI.2026.10076491
   
  
• Circulatory tumour cells and circulatory tumour DNA as a marker for breast cancer finding  
  by Ramya Ravindran, Ramadas Nayak, Kavitha Kanjirakkad Parameswaran, Vipin Viswanath 
  Abstract: One of the primary causes of cancer-related fatalities globally is breast cancer, and enhancing patient survival requires early identification. Conventional diagnostic techniques frequently do not allow for real-time monitoring or early disease detection. New developments in liquid biopsy technologies, including those involving circulating tumour cells (CTCs) and circulating tumour DNA (ctDNA), present encouraging, minimally invasive methods for early diagnosis. In order to evaluate the diagnostic utility of both biomarkers, this study included 60 breast cancer patients and 60 healthy controls. Both techniques successfully separated breast cancer-positive cases from healthy controls in a direct comparison in a clinical context, with CTCs demonstrating somewhat higher sensitivity and accuracy. RT-PCR was used to identify HER2-positive ctDNA, and multiparameter flow cytometry was used to characterise CTCs. These results demonstrate how CTC and ctDNA detection may be combined to improve early breast cancer identification, which would ultimately improve clinical outcomes and individualised treatment.
  Keywords: breast cancer; tumour; circulating tumour cell; CTC; circulating tumour DNA; ctDNA; reverse transcription-polymerase chain reaction; RT-PCR.
  DOI: 10.1504/IJMEI.2026.10077112
   
  
• Evaluating the impact of GDPR on electronic health records management in European hospitals  
  by D.Ramya Dorai, A.Bathsheba Parimala, B. Judy Flavia, I.M.Christina Febiula, K.K. Sreedeve, Lachimipriya Kathirvel 
  Abstract: data as electronic records are increasingly used in healthcare. A representative sample of 75 urbans, suburban, and rural hospitals from 12 EU nations was studied. The regional snapshot sample illustrates the difficulties of failing to control for hospital size and resource variability. The study quantifies compliance practice, IT departmental issues, and organisational accommodation using qualitative and quantitative data. Important findings include increased investment in data security infrastructure, reduced administrative costs, and greater patient confidence and openness. Methodological rigour was achieved through the use of R for statistical modelling, Tableau for visualisation, and NVivo for qualitative thematic coding. The study uses architectural and functional diagrams to show how EHR systems add consent management, access controls and encryption. Comparative and historical trend charts show the impact of GDPR on breach reduction, compliance costs and user behaviour. The study helps policymakers and institutions balance patient-centred care with data privacy as European healthcare rapidly digitises.
  Keywords: General Data Protection Regulation; GDPR; electronic health records; EHRs; data privacy; hospital IT; European healthcare.
  DOI: 10.1504/IJMEI.2026.10077137
   
  
• Integrating cutting-edge algorithms with wearable sensor data for predict personalised health status monitoring and intervention  
  by Asha Pandit Ghodake, Hanamant Bhagwan Sale, Rahul Raghvendra Joshi 
  Abstract: Integration of advanced algorithms with wearable sensor data to enable personalised health monitoring and intervention. Key challenges addressed include data privacy, real-time data processing, and algorithm validation. The objective is to develop accurate health status predictions using the Wang-Mendel (WM) algorithm and sequence-to-sequence (Seq2Seq) model. Wearable sensor data is processed using MATLAB tools to generate customised health alerts. The research highlights the potential for improved, tailored health monitoring systems, with results showing 95% oxygen saturation accuracy. This approach fosters proactive healthcare management, enhancing individual well-being and promoting preventive care for better healthcare outcomes.
  Keywords: Wang-Mendel algorithm; sequence-to-sequence model; Seq2Seq; cutting edge algorithms; wearable sensor data; health status monitoring; intervention.
  DOI: 10.1504/IJMEI.2026.10077247
   
  
• An efficient skin cancer detection and classification performance using adaptive residual DenseNet with attention mechanism  
  by S. Jenita Christy, G. Rosline Nesa Kumari 
  Abstract: Skin cancer is a disease that occurs when skin cells grow abnormally. The research implements a novel technique to detect and classify skin cancer. Initially, the input images are gathered from standard sources and fed into the segmentation phase, where the transformer-based residual dense UNet++ (TRDUNet++) method generates segmented images. Then, the detection process is done by adaptive residual DenseNet with attention mechanism (AResDAM), and its parameters are tuned by enhanced random variable-based lyrebird optimisation algorithm (ERV-LOA). Then, the overall performance is validated by utilising several performance measures and it attains 94.3% accuracy, 90% sensitivity, and 96% specificity values.
  Keywords: skin cancer detection; transformer-based residual dense UNet++; adaptive residual DenseNet with attention mechanism; enhanced random variable-based lyrebird optimisation algorithm.
  DOI: 10.1504/IJMEI.2026.10077487
   
  
• A systematic exploration of kidney tumor restoration and detection frameworks using CT imaging  
  by Prudhvi Raj Budumuru, Sri Rama Krishna Kalva, P. Murugapandiyan 
  Abstract: This study provides a thorough examination of the approaches and applications of artificial intelligence (AI) currently utilised in kidney tumour diagnosis, image restoration, and detection. This review article provides a basic overview of important developments in AI, Deep learning (DL), and machine learning (ML), and focuses on the most current developments and applications of AI in kidney cancer diagnosis. The purpose of this study is to provide a brief analysis of difficulties and potential applications for kidney disorders. This review primarily highlights the importance of conducting additional research and improving AI to transform the detection and treatment of kidney cancer.
  Keywords: artificial intelligence; AI; deep learning; DL; segmentation approaches; computed tomography images; machine learning; ML; kidney tumor image restoration and detection model.
  DOI: 10.1504/IJMEI.2026.10077488
   
  
• SRideNN-FHNet: support RideNN forward harmonic network for mental health detection  
  by Amit Dnyaneshwar Narote, Uttam Dnyanu Kolekar, Manoj Muralidhrrao Deshpande, Vijaykumar Nageshrao Pawar 
  Abstract: In this research, a new deep learning method for mental illness named support RideNN forward harmonic network (SRideNN-FHNet) is proposed. Initially, the acquisition of input data from the specified dataset is done and is sent to the attribute extraction phase. Thereafter, the recursive feature elimination (RFE) is used for selecting the necessary features, and the Synthetic Minority Oversampling Technique (SMOTE) is then applied to augment the data. Finally, the SRideNN-FHNet approach is used for detecting mental health issues. Here, the SRideNN-FHNet is obtained by integrating support vector machine (SVM) and RideNN, with harmonic analysis. Furthermore, the SRideNN-FHNet computed a maximum true negative rate (TNR) of 91.205%, true positive rate (TPR) of 93.153%, and accuracy of 92.051%.
  Keywords: mental health; support vector machine; SVM; mental health detection; RideNN,; machine learning; recursive feature elimination; RFE.
  DOI: 10.1504/IJMEI.2026.10077769
   
  
• The properties and clinical trial results of titanium and titanium alloys as biomedical materials on orthopaedic implants: a review  
  by Diah Ayu Fitriani, Damisih Damisih, Putri Sayyida Ashfiya, Ika Maria Ulfah, Siti Amalina Azahra, Prabowo Puranto, Aghni Ulma Saudi, Razie Hanafi, I. Nyoman Jujur, Dwi Gustiono, Bambang Triwibowo, Agus Nugroho, Kusuma Putri Suwondo, Muhammad Kozin, Muhammad Prisla Kamil 
  Abstract: Biomedical materials have recently attracted much attention due to the increasing life expectancy and aging population issues. Titanium and its alloys have garnered significant attention in biomedical applications, particularly orthopaedic implants, due to their excellent biocompatibility, mechanical strength, and corrosion resistance. The present review comprehensively covers the requirement of orthopaedic implant material and examines the properties of titanium and its alloys, including mechanical properties, and biocompatibility. Several types of titanium are also discussed in detail, such as commercial pure titanium (Cp-Ti) and alloys such as Ti-6Al-4V, Ti-6Al-4V ELI, and Ti-6Al-7Nb. Additionally, this review evaluates numerous clinical trials to assess its efficacy and safety-based implants, focusing on the success rate, implant failure, and potential side effects.
  Keywords: biomedical material; titanium; titanium alloy; orthopaedic implant; clinical trial.
  DOI: 10.1504/IJMEI.2026.10077803
   
  
• Assessing the role of access control mechanisms in protecting patient data  
  by Sonal Yuvraj Borase, R. Sujeetha, S. Poornima, I.M.Christina Febiula, A. Lillyroslin, R.L. Shyja 
  Abstract: In the digital age, safeguarding patient data is paramount. Access control mechanisms are critical tools for ensuring the confidentiality, integrity, and availability of electronic health records (EHRs). This research paper examines the effectiveness of three access control models - role-based access control (RBAC), attribute-based access control (ABAC), and risk-adaptive access control (RAdAC) - in protecting patient information in healthcare systems. Through a comprehensive literature review, we examine the strengths and limitations of each model and highlight their applicability across different healthcare scenarios. The study further examines the integration of advanced technologies, such as blockchain and smart contracts, in enhancing access control frameworks. By analysing real-world case studies and existing literature, we identify best practices and potential pitfalls in implementing these mechanisms. The findings underscore the necessity for a hybrid approach that combines the rigidity of RBAC, the flexibility of ABAC, and the adaptability of RAdAC to address the dynamic nature of healthcare environments. Ultimately, this paper aims to provide healthcare organisations with insights into optimising access control strategies to enhance patient data security.
  Keywords: access control mechanisms; electronic health records; EHRs; role-based access control; RBAC; attribute-based access control; ABAC; patient data security.
  DOI: 10.1504/IJMEI.2026.10077804
   
  
• Predictive modelling of infectious disease spread using machine learning algorithms  
  by R. Ramyadevi, T. Karpagam, M.S. Bennet Praba, Ajay Kumar, Koppuravuari Gurnadha Gupta, V. Maria Christy 
  Abstract: The biggest public health issue worldwide is the spread. Predictive modelling helps identify illness patterns and enables early intervention. This article models infectious disease spread using machine learning (ML) techniques from epidemiological, mobility, and environmental data. Supervised learning approaches, such as random forests, support vector machines and gradient boosting, are combined with time-series models, such as LSTM and Prophet, to predict infection rates, geolocate outbreaks, and identify influential factors. WHO and ECDC real-time data were processed in Python using scikit-learn, TensorFlow, and pandas. Data ingestion, preprocessing, feature selection, and model deployment are integrated into a scalable pipeline. Experimental results show that ensemble-based models outperform baseline models for both short- and long-term predictions, achieving significant increases in accuracy. Visualisation tools: waterfall and mesh plots showed varying effects and spatial distributions. The findings show that ML can improve pandemic control epidemiology by improving prediction, response time, and policy. The research offers a replicable, modular approach that can be applied to other diseases and regions. The study proposes a framework for integrating ML into public health surveillance systems to reduce the impact of outbreaks.
  Keywords: infectious diseases; predictive modelling; machine learning; disease forecasting; public health analytics.
  DOI: 10.1504/IJMEI.2026.10077805
   
  
• An efficient approach for myeloma cancer detection using adaptive efficient Unet++ with spatial attention module  
  by M.M. Shina, D. Pamela, G. Glan Devadhas 
  Abstract: The identification of Myeloma cancer that occurs in bone marrow can be difficult to analyse in early stage. The diverse visual pattern and the irregular shapes tend to struggle in the earlier deep learning approaches and provide less accuracy and blurred images in results. An advanced deep learning-based detection approach is introduced. It provides the reliable outcome that can be achieved through the integration of adaptive efficient Unet++ with spatial attention module (AEUnet++-SAM) for identifying the myeloma cancer by attaining fine details in images. This process is enhanced by parameter optimisation with the aid of the fitness-based muscle selected botox optimisation (FMSBO) algorithm. Finally, the outstanding performance of the proposed technology is determined.
  Keywords: myeloma cancer detection; microscopic image segmentation; fitness-based optimisation; adaptive efficient Unet++ with spatial attention module; AEUnet++-SAM.
  DOI: 10.1504/IJMEI.2026.10078236
   
  
• Early detection of Alzheimer’s disease using enhanced CNN models and image balancing techniques: a different perspective  
  by Archana Thakran, Yogesh Kumar Gupta 
  Abstract: Early detection of Alzheimers disease can be achieved by analysing MRI scans with computer-aided diagnostic (CAD) systems. Existing CAD systems lack robustness and generalisation capabilities due to suboptimal model architectures and imbalanced datasets. To address these issues, we propose three key contributions: 1) developing enhanced convolutional neural network (CNN) models based on pretrained VGG16 and InceptionV3; 2) optimising performance of enhanced CNN by hyperparameter tuning; 3) a novel balanced adaptive synthetic ensemble (BASE) approach to generate synthetic images and balance datasets. In this paper, we utilised two publicly available datasets, OASIS and KAGGLE and achieved accuracies of 93% and 91% using enhanced CNN and BASE.
  Keywords: Alzheimer’s disease; MRI; computer-aided diagnostic systems; CAD; enhanced CNN; hyperparameter tuning; imbalanced datasets; balanced adaptive synthetic ensemble.
  DOI: 10.1504/IJMEI.2026.10078411
   
  
• Human emotion classification enabled by EEG signal analysis and machine learning  
  by Dattaprasad A. Torse, Mahadev M. Bagade 
  Abstract: In recent years, automated human emotion recognition system poses numerous challenges with large and complex data and its vast computation. In this work, we consider a three-dimensional continuous valence-arousal-dominance framework for data representation in space using the DEAP dataset of 32 participants. An experiment was carried out to test practicality of the proposed system using the EMOTIV Insight headset electroencephalogram (EEG) data from four channels. We utilise a tunable-Q wavelet transform (TQWT) algorithm to extract frequency domain characteristics of the signals and classify using time windows of two seconds. The extracted power of the signals was identified as features from different frequency bands and the K-nearest neighbour (KNN) and random forest (RF) classifiers resulted in an accuracy of 97.8%.
  Keywords: emotion recognition; tunable-Q wavelet transform; TQWT; K-nearest neighbour; KNN; random forest; RF.
  DOI: 10.1504/IJMEI.2023.10058261
   
  
• Revolutionising organ donation and transplantation for a better future: a blockchain based approach  
  by Rachana Y. Patil, Yogesh H. Patil 
  Abstract: Prolonged monitoring of life-sustaining organs and their supply chain is challenging. Patients have to wait for a longer time due to numerous lacunas in the organ donation and transplant system. The objectives of this article are manifold, first to address the major challenges in successful organ donation and transplantation systems worldwide. Second, to focus opportunities in workforce expansion, developing modern infrastructure and creating social awareness for organ donation. Third is to regulate the legal, ethical and administrative management among government authorities and medico-legal firms. To overcome these irregularities, we have proposed a secure, distributed and immutable blockchain - based approach to improve and streamline organ donation and transplantation procedures. This ensures the availability of organs for the needy patents. This blockchain-specific approach helps to prevent and monitor organ trafficking by auditable medical transactions. This opens up blockchain-based smart health services to defend doctors and patients' rights.
  Keywords: organ donation; transplantation; blockchain; challenges; opportunities.
  DOI: 10.1504/IJMEI.2023.10058768
   
  
• Respiratory waveform variability during oral feeding in preterm infants  
  by Caryn Bradley, Joanne Ly, Anton Palma, William C. Tang 
  Abstract: Preterm infants admitted to a neonatal intensive care unit (NICU) require successful oral feeding for discharge, but often struggle with bottle feeding due to neurodevelopmental immaturities. We have identified an objective criterion, respiratory waveform variability (RWV), to quantify the adaptability of breathing during bottle feeding. RWV is calculated using sample entropy (SE), a statistical measure of the change of complexity of physiologic processes. Analysis of RWV may be an early predictor of a neuromotor system capable of meeting the demands of feeding and subsequent discharge from the hospital. Results from this exploratory pilot study (n = 10) revealed that higher RWV was associated with significantly fewer days to discharge [β = -1.8 days per 0.1 unit change in SE (95% CI: -3.4, -0.2)]. These findings suggest that RWV is a potentially informative biomarker of physiology and may provide an objective metric to measure infant oral feeding skills in the NICU.
  Keywords: preterm infant; oral feeding; breathing patterns; respiratory waveform variability; dynamic systems theory; oral feeding adaptability; feeding outcomes.
  DOI: 10.1504/IJMEI.2026.10076049
   
  
• A novel deep learning approach for b-value optimisation in intravoxel incoherent motion magnetic resonance imaging on simulated data  
  by Abin Shoby, Jerome Francis, Jini Raju, C. Ushadevi Amma, Ansamma John 
  Abstract: Intravoxel incoherent motion magnetic resonance imaging (IVIM MRI) is a non-invasive technique which measures the perfusion and diffusion effects present in a tissue. One of the major challenges in IVIM imaging is the prolonged scan time since multiple b-value images are required for estimating IVIM parameters. The proposed work introduces a novel approach for reducing the number of b-values required for the generation of IVIM signals from unknown b-values using long short-term memory (LSTM) network. Experimental results show that LSTM network has the capability of accurately estimating IVIM parameters even with 4 b-values, with the estimated values are in agreement with the literature. If an IVIM machine takes 37.5 seconds for the acquisition of a b-value signal, then it will take only 2.5 minutes for acquiring IVIM signals for 4 b-values. This will reduce the patient discomfort and increases the clinical acceptance of IVIM imaging.
  Keywords: intravoxel incoherent motion imaging; long short-term memory; LSTM; area error; optimal b-values.
  DOI: 10.1504/IJMEI.2023.10058833
   
  
• Solution for I-RFID-based smart infrastructure health monitoring's security and privacy  
  by A. Rajasekar, K. Vidya, G. Adharsh, C. Sivakumaran 
  Abstract: The idea of a 'smart city' and 'smart infrastructure' is a relatively recent one, and it refers to the management and control of the various infrastructures of a city via the integration of internet and cellular networks. The addition of internet of things functionality to cellular networks results in an increase in both the scalability and dependability of the whole system. The most current innovation in integrated radio frequency identification (I-RFID) sensor technology allows for the data collected by the sensor to be sent over greater distances in a more secure manner using access point and base station. We constructed a network of intelligent nodes, each of which will consist of a radio-frequency identification tag, a reduced function RFID reader and sensors. The proposed method exhibits minimal data delivery losses and a considerable reduction in the amount of time spent on transmission delays.
  Keywords: radio frequency identification; RFID; healthcare monitor; security; smart infrastructure.
  DOI: 10.1504/IJMEI.2023.10060421
   
  
• Early-stage leukaemia detection using sophisticated machine learning algorithms  
  by Pawan Whig, Anant Aggarwal, Dhaya Sindhu Battina, Srinivas Venkata, Shama Kouser, Ashima Bhatnagar Bhatia 
  Abstract: This paper investigates the efficacy of convolutional neural networks (CNNs), a deep learning technique, in early-stage leukaemia detection - a crucial task for improving outcomes. Comparing support vector machines, random forests, artificial neural networks, and CNNs, we assess performance on a dataset of blood samples from leukaemia patients and healthy subjects. Results reveal high accuracy across models, with CNN outperforming other methods in both accuracy and efficiency. CNN's capacity to learn complex patterns from raw data, such as blood samples, sets it apart from traditional algorithms. This study underscores CNN's potential to revolutionise early-stage leukaemia detection, demonstrating its significance in advancing cancer diagnosis.
  Keywords: early-stage leukaemia; machine learning; deep learning; convolutional neural network; CNN; classification; feature extraction; diagnosis.
  DOI: 10.1504/IJMEI.2023.10060130
   
  
• Breast cancer diagnosis based on thermography images using deep learning and forest optimisation algorithm  
  by M. Sri Geetha, A. Grace Selvarani, R. Vinodhini, M. Murugan 
  Abstract: Breast cancer is one of the most lethal types of cancer affecting women. The method of obtaining mammograms is a painful and unpleasant procedure for women since it requires compression of the breasts. This article would analyse thermal breast images for symptoms of the disease using image-processing techniques and algorithms. With this procedure, breast cancer could be discovered at an earlier stage of development. We describe a novel approach for extracting breast differentiating features utilising bio-data, image analysis, and image statistics. These features were obtained from thermal images acquired by a camera, and they will be classified by convolutional neural networks (CNNs) as normal or suspicious. The proposed method gives an accuracy rate of 98.95% for the thermal image dataset.
  Keywords: breast cancer; breast thermal image; convolutional neural network; CNN; image analysis; thermography.
  DOI: 10.1504/IJMEI.2024.10063877