Forthcoming and Online First Articles

International Journal of Medical Engineering and Informatics

International Journal of Medical Engineering and Informatics (IJMEI)

Forthcoming articles have been peer-reviewed and accepted for publication but are pending final changes, are not yet published and may not appear here in their final order of publication until they are assigned to issues. Therefore, the content conforms to our standards but the presentation (e.g. typesetting and proof-reading) is not necessarily up to the Inderscience standard. Additionally, titles, authors, abstracts and keywords may change before publication. Articles will not be published until the final proofs are validated by their authors.

Forthcoming articles must be purchased for the purposes of research, teaching and private study only. These articles can be cited using the expression "in press". For example: Smith, J. (in press). Article Title. Journal Title.

Articles marked with this shopping trolley icon are available for purchase - click on the icon to send an email request to purchase.

Online First articles are published online here, before they appear in a journal issue. Online First articles are fully citeable, complete with a DOI. They can be cited, read, and downloaded. Online First articles are published as Open Access (OA) articles to make the latest research available as early as possible.

Open AccessArticles marked with this Open Access icon are Online First articles. They are freely available and openly accessible to all without any restriction except the ones stated in their respective CC licenses.

Register for our alerting service, which notifies you by email when new issues are published online.

We also offer which provide timely updates of tables of contents, newly published articles and calls for papers.

International Journal of Medical Engineering and Informatics (124 papers in press)

Regular Issues

  • Computerised detection of autism spectrum disorder using EEG signals   Order a copy of this article
    by Aarti Sharma 
    Abstract: Autism spectrum disorder (ASD) is one of the most common neurological disorders. Detection of ASD is based on behavioural analysis made by clinician by conducting interviews with the parents of the child. This paper presents computer aided diagnostic tool to detect autism disorder. This paper presents the detection of ASD based on biological markers. An early diagnosis is essential to confirm that the child have ASD. In this paper, power spectral density and phase locked values have been extracted from gamma band of EEG signals for autistic and normal subjects. Gamma band from EEG signals is extracted using band pass filter. Significant decrement in these features is observed for the autistic subjects in comparison to normal subjects. Above findings are statistically validated though ANOVA. Anomalies in EEG signals can be used as potential biomarker for detection of ASD.
    Keywords: autism; electroencephalogram; EEG; gamma wave; computer aided diagnosis.
    DOI: 10.1504/IJMEI.2021.10043643
     
  • Biothermodynamics may be a highly useful tool to help medical practitioners to detect and cure morbidities   Order a copy of this article
    by Mustafa Özilgen 
    Abstract: Biothermodynamics may help clinicians to diagnose and cure many medical cases, slow down ageing, correct malnutrition and repair the distorted tissues. Thermodynamically detected decrease in cardiac or renal filtration efficiencies indicate a malfunctioning system. Thermodynamic assessment of the circulatory system after amputation and restoring the circulatory pressure drop may prevent deaths. Research on thermodynamics of fertilisation may help to medical practitioners to fight against infertility. Athletes of some sports have substantially shorter lifespan than the other people. Biothermodynamics may offers some help to those athletes via prescribing appropriate energy intake. Distortion of the energy management in the body is observed in various health problems, including cancer. The need for multidisciplinary research to achieve additional understanding of these phenomena has been suggested by numerous researchers. Biothermodynamics may offer the best tool to achieve this goal.
    Keywords: biothermodynamics; internal work performance; external work performance; energy and exergy efficiency; helping medical practitioners; diagnosing and curing morbidities.
    DOI: 10.1504/IJMEI.2021.10043802
     
  • Risk stratification of cardiovascular disease in type 2 diabetes using LDA and CNN for clinical decision management - a multi-centre study in eastern India   Order a copy of this article
    by Suparna Dutta, Saswati Mukherjee, Medha Nag, Sujoy Majumdar, Ghanshyam Goyal 
    Abstract: Approximately 72.9 million patients of type 2 diabetes mellitus (T2DM) in India are at a potential risk of cardiovascular diseases (CVDs), strokes and peripheral gangrene. CVD is a major cause of disability and death is one of the major areas of risk severity stratification study. Unlike well-known prediction score models of CVD herein, a unique assessment deep learning model is proposed to stratify the cardiovascular events in different risk grades in T2DM individuals This risk assessment tool can aid clinicians in decision management of CVD risk severity It is a retrospective cross-sectional observational study that stratifies risks using linear discriminant analysis (LDA) and convolution neural network (CNN). Class separability feature of LDA helps to achieve optimal performance. The model is externally validated in a cohort of 4,719 individuals with T2DM to assess performance heterogeneity across different settings.
    Keywords: convolution neural network; CNN; linear discriminant analysis; LDA; support vector machine; SVM; risk stratification; cardiovascular disease; CVD; type 2 diabetes mellitus; T2DM; India.
    DOI: 10.1504/IJMEI.2021.10043803
     
  • Cloud and online system of 3D printing for serving multi-client of hospitals and medical colleges in different locations   Order a copy of this article
    by Ignatius Luddy Indra Purnama, Alva Edy Tontowi, Herianto 
    Abstract: This manuscript presents a 3D printing method, especially human bone anatomy, in a cloud and online system. The human bone anatomy, specific skull and foot bone, of the patient’s digital imaging and communications in medicine (DICOM) file, is uploaded to the so-called 3DPNet-DICOM cloud. After printing, the 3DPNet-DICOM cloud sends a notification and the 3D printed bone model to the hospital by courier. The 3D images constructed from the DICOM file and the 3D printed model was identifiable with no significant dimensional errors. The result is to eliminate the human operators’ activity, and queuing at the order process.
    Keywords: 3D printing; bone model; cloud application; online application; DICOM; network access.
    DOI: 10.1504/IJMEI.2021.10043869
     
  • A conceptual model to improve the patient flow during COVID-19   Order a copy of this article
    by Abdesselam Bougdira, Asmae Mazti, Hayat Sedrati, Fayssal Jhilal, Saaid Amzazi, Chakib Nejjari, Hassan Ghazal 
    Abstract: In the COVID-19 context, hospitals across the globe endeavour to manage the unprecedented flow of patients. Challenges are related to the unexpected increase in patients, extra waiting time for cleaning and decontamination of inpatient beds, congested queueing in treatment areas, and waiting time in crowding rooms. This research introduces a new conceptual model to bring a broader, patient-centred, and traceability-based view of patient flow in a clinical setting in the context of COVID-19. The model design opts for four interrelated modes procedural, technical, operational, and contextual. These modes lay the basis of a patient flow solution before starting its development. The feasibility of the proposed model is shown through a usage scenario and developed user interfaces. Results have shown that a knowledge representation of the patient’s real-time information can enable better patient flow monitoring. In addition, it would support intelligent healthcare systems that reuse and share pieces of declarative knowledge. Such functionality would enhance the management of patient flow and improve dealing with the patient flow challenges imposed by the COVID-19 pandemic. Moreover, the presented proof of concept laid the ground for future applicability in various clinical settings facing similar infectious disease crises.
    Keywords: patient flow; clinical setting; COVID-19; traceability; ontology; artificial intelligence; conceptual modelling; hospital systems.
    DOI: 10.1504/IJMEI.2022.10044040
     
  • Spotting congenital heart diseases using palm print based on faster R-CNN and spatial method   Order a copy of this article
    by Y. Mahesha, C. Nagaraju 
    Abstract: This paper proposes a machine learning method to detect congenital heart diseases (CHDs) using a palm pattern known as axial triradius. This article spreads light on three things. First, Faster R-CNN Inception v2 has been used to identify triradii on the palm image. Secondly, a novel spatial method has been proposed to select leftmost, rightmost and axial triradii. Finally, the angle at axial triradius has been calculated on the palm images of healthy people and of patients suffering from tetralogy of Fallot (TOF), atrial septal defect (ASD), ventricular septal defect (VSD) and coarctation of aorta (CoA). The result shows that the proposed method can be used as a screening method to predict CHDs.
    Keywords: axial triradius; CHDs; faster R-CNN; inception v2; spatial method.
    DOI: 10.1504/IJMEI.2022.10044267
     
  • Real-time healthcare monitoring system through haversine distance calculation-based global positioning system   Order a copy of this article
    by Leo John Baptist, Linesh Raja, Suresh Shanmugasundaram 
    Abstract: In the present world, new technologies are highly-developed which drastically changes the conventional tasks of medical and the healthcare systems. Evolving mobile health (m-health) systems belong to these technologies with advanced data communication, artificial intelligence, deep learning, big data, cloud computing, and other machine learning approaches. Data collected using sensors are sent to the local databases via cellular networks which are the stored in the cloud. Data residing in the cloud or medical centres are used for analysis. Machine learning approaches are utilised for predicting the disease appropriately and for classification purpose. This research paper details the m-health systems, their framework with GPS assistance using machine learning. Moreover, this model can save the human life for some time as the ambulance location is tracked using GPS. Heart beat rate and temperature of patient are sensed and the information is sent to the hospital or to the mobile of the doctor.
    Keywords: global positioning system; GPS; healthcare; longitude; latitude; m-health.
    DOI: 10.1504/IJMEI.2022.10044463
     
  • Empirical assessment of COVID-19 infections and information diffusion: a data science approach   Order a copy of this article
    by Isaac Kofi Nti, Adebayo Felix Adekoya, Owusu Nyarko-Boateng, Ponnadurai Ramasami 
    Abstract: The spread of the novel coronavirus disease, SARS-CoV-19 (COVID-19), has affected human activities everywhere, resulting in fear and panic among all age groups. Hence, this study implements a novel data science process to empirically model the daily reported cases and Google search queries in 14 countries. We observed a strong positive association (0.79-0.96) among reported cases of COVID-19 in the 14 countries. Furthermore, there is an inverse correlation of -0.18 to -0.62 between information diffusion on the virus and reported cases (new cases and deaths). Our outcome shows that contagious diseases are highly predictable using historical records from other countries and information spread on the disease.
    Keywords: data analytics; machine learning; data science; coronavirus; SARS-CoV-2; COVID-19; Goggle search engine; infections.
    DOI: 10.1504/IJMEI.2022.10044968
     
  • How health knowledge that can influence patient outcome   Order a copy of this article
    by Leandro Pereira, Carlos H. Jeronimo, André Salgado, Álvaro Dias, Renato Lopes Da Costa, Rui Gonçalves 
    Abstract: Since COVID-19 was declared a global pandemic, it has killed more than two million people worldwide. Some directly due to complications and symptoms of the virus others due to lack of resources to take care of all the patients affected by it. A proper triage can go a long way to an efficient resource allocation, which makes it extremely relevant to understand which factors can affect COVID-19 complications or mortality risk. With the analysis of survey data collect form front-line Portuguese doctors, it was possible to identify the main comorbidities and health issues that can influence patient outcome.
    Keywords: COVID-19; mortality risk; comorbidities; severity of symptoms; knowledge management.
    DOI: 10.1504/IJMEI.2022.10045013
     
  • Colorectal cancer risk factor assessment in Palestine using machine learning models   Order a copy of this article
    by Mohammad A. Z. Abu Zuhri, Mohammed Awad, Shahnaz Najjar, Nuha El Sharif, Issa Ghrouz 
    Abstract: Colorectal cancer (CRC) has a prevalence of 15% among men and 14.6% among women of all cancers. This research was carried out to assess behavioural risk factors that affected Palestinian reported CRC cases, and to make use of machine learning (ML) tools that might be used in CRC prediction, where the use of a public CRC classification and prediction tool based on accurate ML tools might help individuals in addressing their behavioural CRC risk factors and enhancing their engagement with their health. In this research, Palestinian dataset that consists of 57 predictors was collected, the dataset consists of 216 instances. Statistical models were used to determine the important features. The study found that the most important risk factors to consider are age, past medical history, diet behaviours, physical activity, and obesity. Consequently, ML models were applied to classify and predict CRC risk factors. Results showed that ANNs outperformed all models.
    Keywords: colorectal cancer; CRC; data mining; risk factors; machine learning; classification; Palestine.
    DOI: 10.1504/IJMEI.2022.10045260
     
  • Computational study of the progression of Alzheimer's disease and changes in hippocampal theta rhythm activities due to beta-amyloid altered calcium dependent ionic channels   Order a copy of this article
    by Akanksha Kaushik, Jyotsna Singh, Shilpa Mahajan 
    Abstract: Although, Amyloid beta (B-amyloid) and neurofibrillary tangles are the assay mark of Alzheimer’s disease (AD), cognitive decline is best concerned with synaptic loss, rather than tangles or plaques. The pyramidal neurons in Hippocampus are highly affected by AD. Therefore, pyramidal neurons are prime focus in our study. Pyramidal neurons have extensively developed calcium signalling, a phenomenon that controls the neuronal rhythms desirable for memory processing and cognition, for regulating wide range of functions like controlling rhythmic activities, information processing and memory binding. The focus of our work is to inspect the impact of B-amyloid on calcium signaling and progression of AD through computational study.
    Keywords: Alzheimer’s disease; calcium signalling; pyramidal neurons; hippocampus.
    DOI: 10.1504/IJMEI.2022.10045261
     
  • A deep neural network-based architecture for automated detection of COVID-19 from chest X-ray images   Order a copy of this article
    by Abul Abbas Barbhuiya, Ram Kumar Karsh, Rahul Jain, Hillol Phukan 
    Abstract: The COVID-19 pandemic has a severe adverse impact on people’s health, lives, and the worldwide global economy. It is only possible to identify positive COVID-19 instances if each country carries out sufficient tests. As suggested in prior studies, X-Ray images could be used as testing samples to develop a reliable and low-cost COVID-19 testing model. This paper introduces a deep learning-based end-to-end binary classification framework, Att-Net, for automated detection of COVID-19 cases using Chest X-ray images. In this work, we have adopted pre-trained ConvNet (VGG-16) with an attention module embedded with the VGG-16 architecture, which significantly improves the model’s performance. The proposed architecture is evaluated on the COVID-Xray-5k dataset. The suggested methodology obtains a state-of-the-art sensitivity of 98.5% and specificity of 99.4%. This work also presents a detailed performance analysis in terms of accuracy, sensitivity, specificity, precision, recall, and F-score. Furthermore, we have also generated the heat maps, which reveal the most anticipated regions infected by COVID-19 while learning for prediction by the CNN to validate the proposed architecture.
    Keywords: COVID-19; pandemic; deep learning; ConvNet; machine learning; transfer learning; feature extraction; X-ray images; healthcare management; CNN.
    DOI: 10.1504/IJMEI.2022.10045268
     
  • Identifying key gait parameters in gender recognition and classification performance analysis using machine learning algorithms   Order a copy of this article
    by Neha Sathe, Anil Hiwale, Archana Ranade 
    Abstract: Different gait parameters retrieved through pressure sensors, classification spatial, statistical, temporal and demographic (SSTD) model is suggested and tested for gender recognition and classification. Combination of spatial, temporal and demographic features along with performed descriptive statistics is use to train the model. Support vector machine, logistic regression and k-nearest neighbour classification results are tested and analysed for precision and recall. Step length and stride length with weight and height provides great performance in achieving accuracy. Classification results within range of 80% to 90% for selected dataset of healthy 80 subjects were achieved. Influence of stride length in female and step length in male recognition along with single support time is observed. Contribution of weight is also recognisable in classification accuracy. Behaviour of female recognition and classification provides clear results on selected features using SSTD model while precision and recall values whereas male recognition values are on lower end.
    Keywords: spatiotemporal parameters; support vector machine; SVM; k-nearest neighbour; KNN; logistic regression; LR; gender recognition.
    DOI: 10.1504/IJMEI.2022.10046133
     
  • An actigraph data-based early diagnosis of depression using ensemble classifiers   Order a copy of this article
    by C.D. Anisha, N. Arulanand, R. Rekha 
    Abstract: Depression is one of the severe mental disorders which prevails as one of the key symptoms in unipolar and bipolar disorder. An early diagnosis of depression can lead to quicker recovery. This paper proposes an artificial intelligence (AI)-based early diagnosis system for depression using the actigraph motor data. The key contribution of the paper is the 'ensemble classifiers' which is a type of machine learning (ML) model, a subpart of AI model, which improves the diagnosis of depression state by combining the predictions of various single classifiers. The result signifies that the proposed system with ensemble classifier AI model has an accuracy of 85% which is reliable and consistent than existing systems.
    Keywords: bipolar disorder; depression; actigraph; ensemble classifiers.
    DOI: 10.1504/IJMEI.2022.10045834
     
  • Tele-health consultations during COVID times - barriers and facilitators: a multi-state exploratory study from India   Order a copy of this article
    by Sree T. Sucharita, Chakrapani Chatla, Vikhram Ramasubramanian, R. Vaishika, S.M. Balaji, R. Pradeep, Aravind Manoharan 
    Abstract: Globally, COVID-19 has drastically derailed the clinical care for outpatient services. Healthcare professionals (HCPs) used various specially designed tele-health applications (THA) to cater to the needs of patients. We attempted to explore the barriers and facilitators in utilising THAs as part of a multi-state and multi-disciplinary exploratory sequential mixed methods study. Our study comprised of initial survey of HCPs using quantitative Google Forms Survey tool followed by personal (physical and phone-based) interviews of 30 selected/consented HCPs. Of the 98 HCPs responded to the survey, ~60% of had 1-2 telehealth consultations/day during the pandemic. Approximately 61% of the consultations were limited to < 60 minutes interactions. Less than half were obtaining consent and only one third were maintaining digital log record. Most of the younger HCPs especially women have continued their services through THAs and felt that THAs will continue in the future. Besides the limitations of generalisability, the results suggest enormous scope for THAs in future towards simplifying the processes for effective healthcare service delivery at lesser cost and time.
    Keywords: COVID-19; tele-health applications; THA; healthcare service delivery; digital platforms; India.
    DOI: 10.1504/IJMEI.2022.10045849
     
  • A new segmentation method for retinal pathologies detection in optical coherence tomography images   Order a copy of this article
    by Ben Khelfallah Amel, Messadi Mahammed, Lazouni Mohammed El Amine 
    Abstract: Diabetic macular oedema (DME) and age-related macular degeneration (AMD) are the leading causes of blindness in adults. The most significant signs of these diseases are appearance of exudates and change of retinal layer structure. Screening of these diseases is very important to prevent vision loss. In this work, a new method based on a genetic k-means algorithm for lesions detection is proposed. From the selected region of interest (ROI), four textural features are extracted and used to classify these two retinal diseases against the normal subjects using the SD-OCT images. From the experimental results found, the SVM gives better results for AMD and DME recognition. The mean accuracy, sensitivity and specificity values for the macular regions classification are 99.67%, 100% and 99.51% respectively.
    Keywords: OCT images; age-related macular degeneration; AMD; diabetic macular oedema; DME; features extraction; classification; genetic algorithms.
    DOI: 10.1504/IJMEI.2022.10046139
     
  • Automated classification of cervical cells using integrated VGG-16 CNN model   Order a copy of this article
    by Rajesh Yakkundimath, Varsha S. Jadhav, Basavaraj S. Anami 
    Abstract: The most popular method for early cervical cancer screening and detection is the Pap-smear. Automatic analysis of Pap-smear images using computer technology will help in the accurate classification of cervical cancer cells. In this paper, a deep learning approach based on VGG-16 convolutional neural network (CNN) model integrated with support vector machine (SVM) classifier is proposed to identify and classify the cervical cells. A deep convolutional generative adversial network (DCGAN) framework is employed to generate the required synthetic Pap-smear images. The best average classification result of 96.24% is achieved on the held-out dataset comprising 16,124 images belonging to five classes of cervical cells.
    Keywords: cervical cancer; Pap-smear images; data augmentation; classification.
    DOI: 10.1504/IJMEI.2022.10046152
     
  • Feature importance analysis for a highly unbalanced multiple myeloma data classification   Order a copy of this article
    by Rima Guilal, Nesma Settouti, Gonzalo Martínez-Munoz, Mohammed Amine Chikh 
    Abstract: Multiple myeloma (MM) is a hematological cancer associated with abnormal plasma cell proliferation. Its diagnostic process is long because it is very difficult to discover it at an early stage. This paper presents an approach to aid in MM diagnosis and staging. Tree-based ensemble learning methods are used to measure the features importance in models constructed for predicting MM stages. Comparative analysis showed that random forest outperformed other algorithms with an accuracy of over 97%; however, XGBoost gives a ranking of features considered most prognostic for MM staging. A discussion of results with specialists in hematology supported and validated our proposed study.
    Keywords: blood cancers; multiple myeloma; prognostic factors; ensemble learning; feature importance; unbalanced data; grid search.
    DOI: 10.1504/IJMEI.2022.10046878
     
  • Effectiveness of machine learning for mental health: observing the mental state of Bangladeshi people   Order a copy of this article
    by Sayda Umma Hamida, Narayan Ranjan Chakraborty 
    Abstract: Analysing and finding the most used AI applications in the mental health sector and advising appropriate directions for advanced research is the intention of this research. With this purpose, authors commenced a systematic review by analysing selected 31 articles and found several neuroimaging and recognising technologies in real life for checking brain abnormalities. Besides, it revealed from the study that bot is the most used AI assistant in digital care. However, the authors surveyed the young people (aged between 19-29) of Bangladesh to identify mental disorders like as- anxiety, depression, and PTSD. The authors used Python to analyse the dataset, find correlations, and applied machine learning classification algorithms (e.g., decision tree, support vector machine, random forest) to measure the accuracy. The researchers explained a few threats of mental instability in their findings and offered several directions for future research using virtual and real-life AI technologies.
    Keywords: AI; mental-health; anxiety; depression; PTSD; Chatbot.
    DOI: 10.1504/IJMEI.2022.10046879
     
  • New bio-inspired approach for deep learning techniques applied to neonatal seizures   Order a copy of this article
    by Mohamed Akram Khelili, Sihem Slatnia, Okba Kazar, Seyedali Mirjalili, Samir Bourekkache, Guadalupe Ortiz, Yizhang Jiang 
    Abstract: Neonatal seizures are a common emergency in the neonatal intensive care unit and their detection using electroencephalography (EEG) recording is one of the biggest challenges that neurologists face. Even though using artificial intelligence methods such as deep learning for computer vision can help to solve these problems, time consumption, complexity, and overfitting or underfitting of the model still limit the application of deep learning. In order to produce a real-time system that can detect neonatal seizures using EEG and solve the problem of the lack of availability of neurologists, a convolution neural network-based marine predator algorithm system is proposed.
    Keywords: neonatal seizures; electroencephalography; EEG; artificial intelligence; deep learning; convolution neural network; CNN; parallel metaheuristic optimisation; marine predators algorithm; MPA; genetic algorithm.
    DOI: 10.1504/IJMEI.2022.10046880
     
  • Neoteric machine learning approaches to diagnose the state of carotid artery   Order a copy of this article
    by Hariharan Anantharaman, Navendu Chaudhary, Vimal Raj 
    Abstract: Develop machine learning models based on the information, data elements and images captured on a carotid ultrasound. Initiated with capture, collation and compilation of comprehensive carotid ultrasound reports of patients. Next analyse, extract, clean and compile data for the development of model. In this neoteric approach, a set of supervised algorithms and image-deep learning algorithms are implemented, different models built and tested. Performance of all the models is par excellence with a majority delivering accuracy over 75%. All the models, based on the varied machine learning algorithms, delivered acceptable and consistent accuracy - few models have even surpassed and delivered superior accuracy.
    Keywords: classifier; supervised learning; carotid ultrasound; stenosis; machine learning; deep learning.
    DOI: 10.1504/IJMEI.2022.10046881
     
  • Cloud-based electronic health record sharing and access controlling blockchain architecture using data de-identification method   Order a copy of this article
    by Munshi Rejwan Ala Muid, Afrin Jubaida, Md. Mehedi Hasan Onik, Hamim Hamid 
    Abstract: Electronic health record (EHRs) demands the highest privacy and security as the accuracy of medical research and service largely depend on the integrity of it. Intruders can tamper sensitive protected health information (PHI) or personally identifiable information (PII) purposefully. This study aims to use blockchain technology to build a complete EHR sharing and tracking system, such that even after access is granted to researchers or medical personnel, EHRs remain unaltered and the identities of patients are hidden. This work uses a traditional cloud storage system to store information while saving the indexes of EHRs in a chain to ensure PHI integrity. Additionally, the data sharing is secured by applying a randomly selected data de-identification method that guarantees integrity and trackability of user data anonymously. Finally, this work demonstrates a complete architectural implementation (using Hyperledger and a private blockchain network) of the proposed EHR sharing and access controlling by using two levels of chain (two distinct ledgers) that ensures an efficient sharing platform of PHI.
    Keywords: electronic health records; EHR; blockchain; private cloud; data de-identification; security; access control.
    DOI: 10.1504/IJMEI.2022.10046882
     
  • An energy efficient reconfigurable architecture for multi-lead ECG signal compression   Order a copy of this article
    by Vinod Arunachalam, N. Kumareshan 
    Abstract: The most used non-invasive diagnostic technique for a wide range of cardiac disorders is an ECG, which records the heart’s electrical activity over time. Compressed ECG signals are a necessary part of most electronic health systems to store and transmit data across long distances. The field programmable gate array (FPGA), a high-speed parallel compute unit, and customisable software capabilities are available with reconfigurable architecture. Consequently, this architecture is suitable for devices like ECGs, which require precise real-time computing for multi-channel signal processing. The Xilinx Zynq 7.000 SoC development board used in this work has an FPGA-based reconfigurable signal processing unit. When compressing data, the method uses fast fourier transformation (FFT). It is possible to achieve a 90% compression rate with this system running in real-time and with minimal to no signal distortion. This method is also the only one in the industry that can reduce high-frequencynoise.
    Keywords: ECG; reconfigurable architecture; health system.
    DOI: 10.1504/IJMEI.2022.10047060
     
  • Biotechnical neural network system for predicting cardiovascular health state using processing of bio-signals   Order a copy of this article
    by Sergei Filist, Riad Taha Al-Kasasbeh, Olga Vladimirovna Shatalova, Mohammad Hjouj Btoush, Manafaddin Namazov, Ashraf Adel Shaqadan, Mahdi Alshamasin, Nikolay Korenevskiy, Saleh Aloqeili, Maxim Borisovich Myasnyankin 
    Abstract: In this study, for the early diagnosis of cardiovascular diseases, a multimodal classifier is built, in which three groups of heterogeneous data are used. The data is classified by autonomous intelligent agents with subsequent aggregation of their solutions at the next hierarchical level of classification. As one of the lower-level classifiers, a classifier is used, built on descriptors obtained on the basis of monitoring and analysing the evolution of the amplitudes of the harmonics of the 0.1 Hz systemic rhythm. The presented architecture of the multimodal classifier showed an increase in the accuracy of the diagnostic efficiency by 11%.
    Keywords: system rhythms; signal demodulation; electrocardiosignal; spectral analysis; neural networks.
    DOI: 10.1504/IJMEI.2022.10047451
     
  • A review on applications of near infrared spectroscopy technique for neonatal monitors   Order a copy of this article
    by Pooja Gohel, Vijay Dave 
    Abstract: In current scenario, to avoid infant mortality especially preterm infants with unknown cause is a big challenge for neonatal hospital personnel. In spite of observation with currently available monitoring devices, there has been remarkable infant death happening at the hospital site. So, need for a device arises that monitor major vital parameters reflecting assessment of important organs such as brain and heart on a continuous basis. Near infrared spectroscopy (NIRS) is the recent technique that measures vitals such as regional cerebral oxygenation (rScO2), tissue perfusion, and related demand-supply of blood at specific organ in continuous, non-invasive, real-time quantitative manner. Many articles have been published showing wide applications of NIRS technique to compare the recorded rScO2 signal under various diseases conditions. This paper gives review of NIRS based measurement of vitals for infants, its applications and effectiveness.
    Keywords: near infrared spectroscopy; NIRS; preterm infant; tissue perfusion; neonatal monitors; regional cerebral oxygenation.
    DOI: 10.1504/IJMEI.2022.10047493
     
  • Chest x-ray image analysis for pneumonitis disease classification   Order a copy of this article
    by Ruchika Arora, Indu Saini, Neetu Sood 
    Abstract: Computer-aided health system increase doctors diagnosing capability and drastically reduces patients’ death. This paper introduces an algorithm with combinational approach of convolution neural network (CNN) and gated recurrent unit (GRU) for pneumonia detection on low-cost chest X-ray (CXR) images. This model practices potential of multiple GRUs with CNN and fuses spatial and label information of CXR images for pixel-level classification. The proposed CNN+GRU model is experimented on pneumonia CXR image dataset available at Kaggle, which consists of 5,216, 624 train and test images respectively. The proposed model achieves 99.74% and 98.37% accuracy on training and testing dataset respectively.
    Keywords: artificial intelligence; disease detection; image classification; convolution neural network; chest images; gated recurrent units; GRU; pneumonia disease; Kaggle; lung diseases; medical image diagnosis.
    DOI: 10.1504/IJMEI.2022.10047778
     
  • FoodKnight: a mobile educational game and analyses for obesity awareness of children   Order a copy of this article
    by Thomas Baily, Fadi Thabtah, Marcus Wright, Duy Anh Tran 
    Abstract: One of the main contributing factors towards child obesity is deficient education regarding food choices. In most cases, children will choose energy-dense foods over those with more nutritional value, without understanding the consequences of their decisions. To help overcome this problem, this paper proposes and assesses the efficacy of an educational gaming application called FoodKnight. Games have the ability to engage children more than traditional teaching schooling methods. FoodKnight incorporates stealth learning to disguise the teaching of healthy food choices while playing a game, while simultaneously incorporating a step-counter to encourage the user to be active. Survey results received from 38 participants regarding the initial prototype have been positive, with minor issues being addressed in an upcoming update. FoodKnight has been shown to be an effective tool for subconsciously conditioning children to make healthier choices.
    Keywords: educational gaming application; food choices; medical informatics; mobile development; stealth learning; software development.
    DOI: 10.1504/IJMEI.2022.10048016
     
  • A novel method to detect and classify Alzheimer using naive Bayes classification algorithm   Order a copy of this article
    by C. Ravichandran, T. Senthil Kumar, K. Balaji, Kavitha Sandanam, Chandrasekaran Sivakumaran 
    Abstract: Digital medical imaging technology has been more accessible to the general people in recent years. This study describes a software strategy for detecting brain irregularities to diagnose Alzheimer’s disease. The suggested method creates a 3D model of the brain using MRI slices. This method is more precise and reliable. In this research, we proposed two strategies. To extract radiological data from MRI images, image processing and machine learning were utilised, and a deep learning approach was applied to analyse the condition of Alzheimer’s disease. The algorithm normalises and removes the skull from the MRI images in the first phase. Using a modified K-Means approach, the image is split into white matter (WM), grey matter (GM), and black holes (BH). The classifier is trained using the training data to predict the test data. The characteristics are defined using naive Bayes to create a classification model.
    Keywords: Alzheimer’s disease; magnetisation resonance image; MRI; mild cognitive impairment; magnetic resonance imaging; deep learning; residual neural network.
    DOI: 10.1504/IJMEI.2022.10048130
     
  • Machine learning based wearable sensor module for human fall detection - a fully functional solution   Order a copy of this article
    by Juluru Anudeep, Shriram K. Vasudevan, T.S. Murugesh 
    Abstract: This paper addresses a potential concern faced by the majority of aged ones left unaccompanied at home. Any untoward fall owing to poor health conditions or slippery floors that are more prevalent among the elders happen, it usually goes unnoticed and the aged ones are deprived of the potentially life saving 'golden hour' treatment. To mitigate such sort of problems faced by the elderly, we have designed a wrist-wearable fall detection system that employs a machine learning model for movement tracking and to detect a fall just in case. During a fall, an automated call is generated to the emergency services as well as to a caretaker through a GSM module. Two datasets are collected, trained and tested on seven different machine learning models, and the results presented.
    Keywords: machine learning; fall detection; wearable device; medical applications.
    DOI: 10.1504/IJMEI.2022.10048147
     
  • Healthcare services: applications, trend, and challenges   Order a copy of this article
    by Abderrazak Sebaa, Nabil Djebari, Abdelkamel Tari 
    Abstract: Recently, a large amount of heterogeneous health-related data and services are generated daily. Therefore, managing these medical flows of data and services requires complex and costly techniques. Moreover, the big service paradigm has received growing attention in various disciplines. This study aims to review and investigate the impact of service computing on health and medical sector. It illustrates the big service challenges, applications and describes how healthcare will benefit from service computing advances. A number of potential research opportunities related to the big service computing paradigm and underlying issues that require longer-term work are also discussed in this paper.
    Keywords: big service; big data; web service; medical services.
    DOI: 10.1504/IJMEI.2022.10048151
     
  • Brain image compression and reconstruction system using deep learning   Order a copy of this article
    by S. Seenuvasamurthi, S. Ashok, B. Shankarlal, A. Mohamed Abbas, Ashok Vajravelu 
    Abstract: New perspectives on brain structure and function can only be gained through the rapid advancement of brain imaging technology. Throughout history, this has been the case. It is common practise in medicine to employ image processing in the early stages of diagnosis and treatment. In classification and segmentation tasks, deep neural networks (DNNs) have so far proven to be exceptional. Functional ultrasound (fUS) is a novel imaging technique that enables the observation of neuronal activity across the brain in awake, ambulatory rats. To achieve adequate blood flow sensitivity in the brain microvasculature, fUS relies on lengthy ultrasonic data collecting at high frame rates, placing a load on the sampling and processing hardware. Parallel MRI is introduced in broad terms, with an emphasis on the classical understanding of image space and k-space-based techniques.
    Keywords: accelerated MRI; parallel imaging; iterative image reconstruction; numerical optimisation; machine learning; deep learning.
    DOI: 10.1504/IJMEI.2022.10048343
     
  • Brain cancer analysis using deep learning architecture on MRI brain image   Order a copy of this article
    by B. Kannan, S. Karthigai Lakshmi 
    Abstract: Brain cancer diagnosis in the medical images sector without human involvement is a huge complex one. The brain tumour tissue can be detected from the whole brain are extremely difficult. Multi-sequence MRI technology is not standardised in brain tumour segmentation clinical practice and hence, a flexible segmentation process is required which uses all of the available MRI data optimally. The proposed algorithm provides a precise and robust segmentation of tumours, which helps in diagnosis, therapy planning, and risk factor detection. SVM classification and convolutional neural network classification are applied and analysed. The proposed system obtains more accurate predictions.
    Keywords: image processing; MRI images; brain tumour; SVM classification; watershed image segmentation techniques; deep learning.
    DOI: 10.1504/IJMEI.2022.10048344
     
  • Detection of Parkinson's disease using CNN   Order a copy of this article
    by M. Kamesh, C. Augustine, D. Sarathy, S. Leopauline, Sheshang D. Degadwala 
    Abstract: Parkinson’s disease can be diagnosed using computer-assisted diagnosis systems based on brain imaging, with the ultimate goal of finding patterns that characterise the disease. In this case, convolutional neural networks (CNNs) have proven to be extremely beneficial. Neurological disease Parkinson’s disease (PD) is characterised by a decrease in the brain’s dopamine-producing neurons. Patients with Parkinson’s disease have difficulty producing speech due to a lack of coordination in the muscles that control breathing, phonation, articulation, and prosody, among other things. Speech analysis can be used by clinicians to objectively assess the severity of Parkinson’s disease in a non-invasive manner. In the LSTM layer, the output is then analysed for important temporal feature relationships. Existing state-of-the-art CNN models are compared to the proposed DenseNet-LSTM model. Training accuracy is 93.75%, testing accuracy is 90%, and validation accuracy is 93.87%, according to the suggested model.
    Keywords: Parkinson’s disease; Parkinson’s speech; empirical mode decomposition; convolutional neural networks; CNNs; Parkinson’s disease; machine learning; meta-heuristics.
    DOI: 10.1504/IJMEI.2022.10048478
     
  • A review on automated detection and classification of diseases in retinal images   Order a copy of this article
    by M.C. Padma, Esra'a Mahmoud Jamil Al Sariera, Thamer Mitib Ahmad Al Sariera 
    Abstract: Hypertensive retinopathy (HR) and glaucoma are the most diseases that cause blindness. Early detection of these diseases is very important for ophthalmologists in clinical diagnostic and successful treatment. The identification and diagnosis of glaucoma and HR require segmentation of the normal objects inside retina such as blood vessels and optic disc (OD). This article describes publicly available retinal datasets and an overview of the state-of-the-art for segmenting normal objects in the retina such as blood vessels and the OD, as well as ways for detecting pathologies that affect normal objects in the retina such as glaucoma and HR. The purpose of this study is to develop a professional structure that will familiarise the researcher with the most up-to-date blood vessels and OD segmentation techniques and the classification of HR and glaucoma diseases. Furthermore, we compared the dataset, evaluation metrics, pre-and post-processing steps, segmentation and classification methods and induced results of these approaches.
    Keywords: hypertensive retinopathy; HR; glaucoma; blood vessels; optic disc; OD.
    DOI: 10.1504/IJMEI.2022.10048781
     
  • Machine learning-based CAD system for thyroid tumour characterisation using ultrasound images   Order a copy of this article
    by Niranjan Yadav, Rajeshwar Dass, Jitendra Virmani 
    Abstract: The main objective of this paper is to develop an efficient computer-aided diagnosis (CADx) system for the characterisation of thyroid tumours using US images. An efficient CADx system is designed to address speckle noise’s effect on thyroid tumour classification in this work. The analysis has been carried out with 820 thyroid ultrasound images. The CADx system designs were examined based on original and despeckled images to compare the texture and morphological features. The extracted features have been used to design five CADx systems, namely: 1) PCA-KNN; 2) PCA-PNN; 3) PCA-SVM; 4) SFS-SVM; 5) LS-ANFC algorithms. The results illustrate that a CADx system based on the PCA-SVM algorithm with (LBP + ZRM) features yields optimal performance for the characterisation of thyroid tumour ultrasound images.
    Keywords: thyroid ultrasound images; local binary pattern; LBP; Zernike features; support vector machine; sequential feature selection; SFS.
    DOI: 10.1504/IJMEI.2022.10049164
     
  • Analysis, design and implementation of electronic navigation device for visually impaired persons   Order a copy of this article
    by Abiodun O. Ogunsanya, Omini E. Okoi 
    Abstract: In this paper, we describe an electronic navigation system, a wearable device intended for obstacle detection and navigation purposes for visually impaired persons. Three ultrasonic sensors, a passive infrared sensor, a programmed microcontroller, GPS module, memory card, audio stimuli feedback, and vibro-tactile feedback comprise the inexpensive, commodity hardware, were connected in a closed circuit. They were used to detect barriers at a distance of up to 1 m. The passive infrared sensor was utilised to detect both animate and inanimate impediments. Initial results of the experiment demonstrate that using the wearable navigation system, the percentage collision rate was reduce that is, 90.1% reduction, 50% reduction in response time, and 98% of variance in distance detection using regression analysis compared to conventional guide care. The device shows dependable response to obstacle detection with minimal error. This demonstrates that wearable navigation device is safe and has the potential to improve quality of life of visually impaired persons.
    Keywords: visually impaired; obstacle detection; wearable navigation device; convectional guide care; microcontroller.
    DOI: 10.1504/IJMEI.2022.10049168
     
  • Deep learning approaches in detection of COVID-19 using X-ray images   Order a copy of this article
    by Shibili Said, Mredhula Lokanathan 
    Abstract: We all have seen how badly COVID-19 had affected our lives. The utilisation of deep learning in analysing covid is a promising research area. This paper brief on the utilisation of deep learning in analysing covid X-ray. A Convolutional Neural Network (CNN), a deep learning algorithm is utilised here, to study the detection of covid X-rays. The literature shows several successful deep learning models in detection of COVID-19. In this paper, we have implemented shallow layered simple CNN and deep layered CoroNet (which is taken from literature) using three different dataset. This paper would give a brief picture of deep learning in covid classification.
    Keywords: deep learning; covid X-rays; convolutional neural network; CNN.
    DOI: 10.1504/IJMEI.2022.10049172
     
  • Analogy of cervical malignancy through Inception V3 and Xception network of CNN   Order a copy of this article
    by K. Hemalatha, N. Kasthuri, N.S. Kavitha, T. Jamuna, K. Kanchana 
    Abstract: Cancer is a disease formed by the abnormal growth of cells and if it not treated in its early stage it spreads to other parts of body. There are more than hundred types of cancer available in the world. An effective method of testing is needed to diagnose the presence or absence of disease, monitor progress of the cancer, and evaluate the treatment’s effectiveness. Cervical cancer ranks fourth among the cancers afflicting women worldwide. To overcome the aforementioned problem deep learning techniques are used for automatically diagnosing the disease from Pap smear images. The proposed model is evaluated on SIPAKMED dataset. CNN-based architectures such as Inception V3 and Xception were used to classify the cervical cells and their accuracy was ascertained. The performance measures such as precision, recall and sensitivity are calculated. The obtained results concluded that CNN pre-trained model Xception achieved the higher classification rate of 95.99%.
    Keywords: cervical cancer; SIPAKMED; Xception; Inception V3; deep learning.
    DOI: 10.1504/IJMEI.2022.10049258
     
  • Use of a neuro-fuzzy technique to predict complete Rockall score in patients with upper gastrointestinal bleeding   Order a copy of this article
    by Zhaleh Ghobadi, Somayeh Saraf Esmaili 
    Abstract: This study, using a neuro-fuzzy technique tried to provide an intelligence model for predicting the complete Rockall score of the patients with upper gastrointestinal bleeding. The data related to 350 patients with upper gastrointestinal bleeding who were hospitalised in the public hospital in Iran have been used. The 30% of the data were randomly selected and used as test data, and the others were also applied as training data. Clinical Rockall and Blatchford scores as inputs and complete Rockall score as the output were considered. To evaluate the model performance, conventional criteria were calculated. The accuracy of the model was compared with the accuracy of the artificial neural network tool. According to the results, proposed model has higher accuracy and efficiency than the artificial neural network and it seems to be used as a useful tool. Accuracy, cheap, high speed and being new are some benefits of the presented model.
    Keywords: upper gastrointestinal bleeding; Rockall; Blatchford; artificial intelligence; neuro-fuzzy system; artificial neural network.
    DOI: 10.1504/IJMEI.2022.10049259
     
  • Auditory model system to recognise Alzheimer's diseases: speech signal analysis   Order a copy of this article
    by Ravi Kumar, R. Prabha, B. Kannan, T.J. Nagalakshmi, Sheshang D. Degadwala 
    Abstract: Alzheimer’s disease is a catch-all phrase for a variety of illnesses, including numerous neurodegenerative disorders. A century ago, neurosyphilis-caused dementia was the most frequent cause of dementia in developed nations, but Alzheimer’s disease (AD) is today the most common cause. Dementia might still be difficult to diagnose because of a number of barriers. Early symptoms overlap with other disorders, and the potential of several, or mixed, aetiologies are just a few of the factors that contribute to a wide range of possible outcomes. Because language diminishes along with neurodegeneration in Alzheimer’s disease, it is a significant source of clinical information. As a result, data on speech and language have been intensively investigated in relation to its diagnosis. Text characteristics extracted from the transcripts are used to detect AD using an SVM classifier. However, the accuracy of automatic assessment falls as WER rises, the two are very weakly associated (0.31).
    Keywords: Alzheimer’s disease; auditory model system; wavelet transform; gammatone filter.
    DOI: 10.1504/IJMEI.2022.10049260
     
  • Brain image enhancement and segmentation using anatomically constrained neural networks   Order a copy of this article
    by P.S. Arthy, A. Kavitha 
    Abstract: Brain image segmentation is one of the processes that take the most time and is the most complicated to do in a therapeutic scenario. The essential principles and features of medical image segmentation based on deep learning are presented. MRI-based medical image classification issues are addressed in this study using a histogram and time-frequency differential deep (HTF-DD) technique. The following are the stages of the proposed approach’s construction. An unsupervised training procedure is used to build a deep convolutional neural network (CNN), which then outputs standardised improved pre-processed features for data extraction. Secondly, a set of time-frequency characteristics is derived from medical images using the time signal and the frequency signal. The last step is to develop an effective model based on differential deep learning for classifying objects. Multi-modal brain data sets and public standards are used to illustrate the applicability of our methodology.
    Keywords: neural network; image processing; brain image enhancement; segmentation; anatomically constrained neural networks; ACNNs; histogram and time-frequency differential deep; HTF-DD; convolutional neural network; CNN.
    DOI: 10.1504/IJMEI.2022.10049261
     
  • Detection and classification of arrhythmia disorders using machine learning algorithm   Order a copy of this article
    by P. Ramani, S. Sugumaran, Manoharan Nivethitha Devi, T.J. Nagalakshmi, G. Annapoorani 
    Abstract: A recent study by the United Nations Agency (World Health Organization) reported that 17.9 million people died due to cardiovascular disease, and it is increasing exponentially. Furthermore, it was also reported that it was highly difficult to recognise the sickness and dictate the relevant care in a timely manner. For analysis, a user data file for cardiopathy prediction that contains parameters such as gender, age, kind of pain, force per unit area, hyperglycaemia, and so on has been considered. The approach entails determining the correlations between the numerous properties of the data file using regular processing techniques and then treating the attributes appropriately to forecast the likelihood of cardiopathy. This article endeavours at probing methodised data-mining techniques such as NB classifier, random forest classification, decision tree in addition to support vector machine. These machine learning approaches require less time to anticipate sickness with a high degree of accuracy. The proposed algorithm provides 91.2% recognition rate than SVM and decision tree classifier.
    Keywords: support vector machine; SVM; NB classifier; random forest; RF; arrhythmia disorders; decision tree.
    DOI: 10.1504/IJMEI.2022.10049587
     
  • Feature analysis for fundus image classification of multi-retinal diseases   Order a copy of this article
    by Widhia K.Z. Oktoeberza 
    Abstract: Retinal diseases affect the vital eye tissue, which can reduce eye vision and even cause blindness if left undiagnosed and untreated. Some retinal diseases can be prevented and even treated properly to return the lost vision by conducting early detection. A scheme to classify multi-retinal diseases is proposed in this study; specifically diabetic retinopathy (DR), age-related macular degeneration (ARMD), and media haze (MH). The process is starting by extracting some features consisting of statistical and texture features, which are undergone in 500 fundus images taken from the RFMiD dataset. Thereupon, these features were classified based on the MLP classifier. Based on that classification process, the accuracy rate of DR, ARMD, and MH classifications achieved 84.2%, 93.2%, and 89.4%, respectively. These achievements show that the proposed scheme effectively classifies multi-retinal diseases and has the potential to assist ophthalmologists in early detecting the appearance of retinal diseases for preventing the worst effect.
    Keywords: retinal diseases; feature analysis; image classification; fundus images.
    DOI: 10.1504/IJMEI.2022.10049694
     
  • A CNN and spatial tract-based statistics-based approach for the diagnosis of Alzheimer’s disease   Order a copy of this article
    by Latifa Houria, Noureddine Belkhamsa, Assia Cherfa, Yazid Cherfa 
    Abstract: The prevalence of Alzheimer’s disease (AD) is growing dramatically each year, making it critical to find efficient strategies to detect the disease early on and prevent its progression. In this work, we present an approach combining tract-based spatial statistics (TBSS) and convolutional neural network (CNN) to classify the AD and mild cognitive impairment (MCI) from a cognitively normal (CN) subject using the diffusion tensor imaging (DTI). The TBSS was applied to generate the WM skeleton from the two DTI maps fractional anisotropy (FA) and mode of anisotropy (MO). The CNN is trained first on FA-Skeleton and MO-Skeleton and then fine-tuning on FA and MO relevant slices. This combinatory procedure achieved a higher result and represents a powerful diagnostic tool.
    Keywords: diffusion tensor imaging; DTI; Alzheimer’s disease; convolutional neural network; CNN; tract-based spatial statistics; TBSS; fractional anisotropy; mode of anisotropy.
    DOI: 10.1504/IJMEI.2022.10049862
     
  • Classification of brain MRI using hypercolumn technique with convolutional neural network   Order a copy of this article
    by K. Kavin Kumar 
    Abstract: The purpose of this study is to classify brain tumours. When abnormal cells grow within the brain, a brain tumour develops. Cancerous (malignant) tumours and benign (non-cancerous) tumours are the two basic forms of tumours. In this regards an algorithm is developed to classify the tumour and non-tumour. The main focus is given to the hypercolumn implementation along with VGG-16 and ALEX-NET. The datasets were taken from the Kaggle and real brain data from Johnson’s MRI and it consists of tumour and non-tumour. The obtained result identifies whether the person is having a tumour or not. The developed algorithm is having an accuracy of 94.5% for VGG-16 with hypercolumn and 91.2% for ALEXNET with hypercolumn.
    Keywords: tumour; non-tumour; hypercolumn; ALEXNET; VGG16; MRI; malignant; benign.
    DOI: 10.1504/IJMEI.2022.10049911
     
  • Automatic knee anterior cruciate ligament torn diagnosis using CNN-XGBoost   Order a copy of this article
    by Kamel H. Rahouma, Ahmed Salama Abdel Latif, Kadry Ali Ezzat 
    Abstract: The knee joint is very important for everyone as it helps us in movements, which is essential for everyone. One of the most diseases that injure the knee is the anterior cruciate ligament (ACL). This work has developed a computer aided diagnosis (CAD) system for examining the given knee magnetic resonance imaging (MRI) image and automatically determining if there is a torn in ACL or not. The region growing based segmentation algorithm is used to get the region of interest (ROI) from MRI image, e.g., extract ACL region from knee image then CNN-XGBoost model is used for knee ACL classification. The model is divided into two main parts: the first part extract the feature uses CNN and the second part using XGBoost for feature classification. The designed model gives us an accuracy of 91%.
    Keywords: deep learning; CNN; XGBoost; knee ACL.
    DOI: 10.1504/IJMEI.2022.10049912
     
  • Analysis and trends of COVID-19 in Italy   Order a copy of this article
    by Cinzia Muriana, Giovan Battista Vizzini 
    Abstract: SARS-CoV-2 is impacting the public health-system worldwide and requires policies to address the demand for additional capacity. Monitoring its spread allows the identification of alarm signals useful for scaling up resources and reacting to the pandemic. In Italy, starting with the identification of the first patient, the Protezione Civile has published a range of indicators as open data, which has supported the country’s government in discovering trends and in setting-up targeted measures for preventing the spread of the virus and controlling the speed of transmission. This paper analyses these indicators from February 2020 to June 2021 and provides insights for healthcare managers.
    Keywords: SARS-CoV-2 outbreak; SARS-CoV-2 predictors; COVID-19 response; outbreak monitoring; Italy.
    DOI: 10.1504/IJMEI.2022.10049913
     
  • Design of automated computer aided diagnosis system to predict diabetic retinopathy based on EfficientNet   Order a copy of this article
    by S. Sathiya Devi, K. Vignesh, V. Raguvaran 
    Abstract: The World Health Organization (WHO) recognises that diabetic retinopathy (DR) is one of the rising healthcare problems in the world which leads to vision loss if left untreated. In this paper, an automated computer aided diagnosis (CAD) system for DR is implemented based on EfficientNet. Initially, image pre-processing is performed by smoothing it with median filter and converting into grey scale image. Then size, colour and shape normalisation are carried out. To increase the volume and to solve the data imbalance issue flipping, rotation, zooming and distortion operations are performed. The CAD system diagnoses the severity levels of DR from fundus images by exploring feature extraction based on EfficientNet B0 model and classification with XGBoost classifier. It is experimented with Indian Diabetic Retinopathy Image Dataset (IDRiD) and experimental result reveals that, the combination of EfficientNet B0 and XGBoost produces better classification accuracy when compared with other convolutional neural network (CNN) models.
    Keywords: diabetic retinopathy; EfficientNet B0; XGBoost classifier; computer aided diagnosis system; healthcare; convolutional neural network; CNN.
    DOI: 10.1504/IJMEI.2022.10050478
     
  • Solving moderated mediation model using interval data on fuzzy elevation   Order a copy of this article
    by A. Akilbasha, R. Vanitha, D. Kalpanapriya 
    Abstract: A variable that grounds mediation between dependent and independent variables. These mediator variables take vital parts in the analysis of data which entails numerous variables, exclusively when the dependent and independent variables are affected by other variables. Thus, mediation analysis requires all areas which need regression analysis, especially in psychology, industry, teaching, production area, etc. Mediation has been projected in multiple studies. On the other hand, at times it is more logical to articulate the fuzzy theory data when the variables are not purely distinct. Normally, to articulate the attitude of a persona bad, moderate, good use fuzzy numbers than using crisp numbers. This paper explains the model of fuzzy moderated mediation using interval data. Also, to find the total, direct and indirect effects psychological data have been applied when the mediator, moderator, and confounding variable subsist.
    Keywords: mediation; moderation; moderated mediation; fuzzy mediation; interval data.
    DOI: 10.1504/IJMEI.2022.10049914
     
  • Cardiac arrhythmia classification of imbalanced data using convolutional autoencoder and LSTM techniques   Order a copy of this article
    by Rekha Rajagopal, V. Shyam Kumar 
    Abstract: Cardiovascular diseases (CVD) can be identified by medical professionals with the help of electrocardiogram (ECG) signals. The ECG signals shows the heart rhythm and any irregularity in heart rhythm is called arrhythmia. The arrhythmias can be broadly classified into five categories: 1) class N; 2) class S; 3) class V; 4) class F; 5) class Q. The proposed research work automatically categorises the ECG beats into one of the five classes using long short-term memory (LSTM). The ECG waveform is divided into individual ECG beats and is provided as input to the convolutional autoencoders. The compressed representation of the encoder is used as features for further classification by LSTM. The class imbalance problem in the dataset is overcome using ADASYN technique. The proposed research work gives an overall accuracy of 99.12%.
    Keywords: arrhythmia; long short-term memory; LSTM; autoencoder; ADASYN; deep learning; disease classification; convolutional neural network; CNN.
    DOI: 10.1504/IJMEI.2022.10049967
     
  • Epilepsy detection and classification based on the contour maps of brain MR images   Order a copy of this article
    by H.S. VinayKumar 
    Abstract: A new method of Epilepsy detection and classification of the degree of affliction is presented. The proposed method is based on the information generated by the filled contour maps of the brain MRI. The result is independent of the size and orientation of the target image. Distribution of multiple intensity levels in an MR image is generated by the proposed contour map segmentation technique. Disease afflicted regions are segmented using this technique. The epilepsy diagnosis is carried out by classifying MR images as healthy and diseased. The diseased ones are further classified into three sub-classes of affliction, namely, mild, medium and severe. The classification is based on the area of the top level segment relative to the total area of the non-skull region. In the proposed method, the classification error is found to be between 5% to 10%.
    Keywords: contour map segmentation; skull-stripping; top level segment; area of non-skull region; filled contour map; contour level list.
    DOI: 10.1504/IJMEI.2022.10050095
     
  • A secure health monitoring system based on fog to cloud computing   Order a copy of this article
    by Hafida Saidi, Nabila Labraoui, Ado Adamou Abba Ari 
    Abstract: Nowadays, the elderly can receive care in their home and enable physicians to follow their diseases in real-time. However, these technologies suffer from several issues like security and privacy-preserving data challenges. In this paper, we proposed a HIPAA-compliant framework that enables security and privacy-preserving medical data based on fog-to-cloud (F2C) computing. Our aims are to define a system that solves the privacy and security issues with remote elderly monitoring. The F2C infrastructure is used to provide better security of medical data and allow a real-time diagnosis of the elderly. Furthermore, F2C combines the benefits of cloud and fog computing such as providing permanent storage, reducing computation load and data transmission delay, and enhancing the security challenges. Simulation results suggest that F2C technology delivers better performance in terms of latency, cost, and energy consumption.
    Keywords: elderly healthcare; wearable sensors; fog to cloud computing; AES-ECC encryption; internet of medical things; IoMT.
    DOI: 10.1504/IJMEI.2022.10050253
     
  • An advanced manta ray search optimisation with likelihood marginal classification mechanism for ECG arrhythmia detection   Order a copy of this article
    by Budidi Udaya Kumar, Jaidhan Beera John, Bharath Kumar Reddy Soma Venkata, Ummadisetty Sreenivasulu 
    Abstract: The proposed work objects to develop an advanced optimisation-based classification system for identifying the arrhythmia disease with increased accuracy and reduced error rate. The optimal features are selected using the manta ray search optimisation (MRSO) technique. The likelihood marginal classification (LMC) approach is employed to predict the classified label as whether normal or disease affected. During analysis, the different types of performance measures are used to evaluate the results of the proposed approach. Also, the obtained values are compared with the existing techniques for showing the effectiveness of the proposed model.
    Keywords: electrocardiogram; ECG; cardiovascular disease; manta ray search optimisation; MRSO; likelihood marginal classification; LMC.
    DOI: 10.1504/IJMEI.2022.10050386
     
  • Reconstruction of a 3D medical image from pre-processed 2D DICOM slices : clinical application   Order a copy of this article
    by Kamal Halloum, Hamid Ez-Zahraouy 
    Abstract: This paper presents two techniques, histogram equalisation (HE) and contrast limited adaptative histogram equalisation (CLAHE), for contrast enhancement of 2D magnetic resonance imaging (MRI) slices in order to reconstruct them as 3D images. These powerful techniques allow the maximum possible contrast to be established without ultimately affecting medical interpretation, based on a comparative study of three 3D images constructed by the slices: real, enhanced only by HE and those processed by CLAHE after HE. We have experimentally demonstrated that the CLAHE technique is effective in improving contrast and reducing noise amplification especially for medical images.
    Keywords: histogram equalisation; CLAHE; 3D medical image; DICOM image; image processing.
    DOI: 10.1504/IJMEI.2022.10050387
     
  • Detection of adverse drug reactions from online health communities’ data: a case study of anti-epileptic drugs   Order a copy of this article
    by Anwar Ali Yahya 
    Abstract: This paper investigates the problem of detecting adverse drug reactions of anti-epileptic drugs from patients’ reviews in online health communities. A lexicon-based methodology is proposed and applied to a dataset of patients’ reviews collected from two online health communities. The dataset is cleaned and the adverse reactions of anti-epileptic drugs are extracted with the aid of consumer health vocabulary and a lexicon of adverse drug reactions. A proportional reporting ratio is then applied to quantify the correlation between each drug and adverse reactions and thus identify the adverse reactions of each drug. The results are validated quantitatively against a database of adverse drug reactions, called side effect resource, and qualitatively against the extant knowledge related to the common adverse reactions and drug-drug similarities of anti-epileptic drugs. The validation results provide evidences on the effectiveness of the proposed methodology and the validity of online health communities’ data for adverse drug reactions detection.
    Keywords: adverse drug reaction detection; pharmacovigilance; anti-epileptic drugs; AEDs; data mining; online health communities.
    DOI: 10.1504/IJMEI.2022.10050418
     
  • Segmentation of the human spinal cord using U-Net architecture   Order a copy of this article
    by S. Kumarganesh, Muzammil Hussain, H. Shaheen, S. Anthoniraj, M. Somaskandan, C. Sivakumaran 
    Abstract: When attempting to assess spinal cord atrophy caused by a variety of disorders, the first step that must be taken is to segment the spinal cord contour. A tumour of the spinal cord is an abnormal development of cells that may occur anywhere in or around the spinal cord. The process of locating tumours in the spinal cord is a very important one. It is difficult to identify the tumour with MRI due to the irregular form of the spinal cord. The model begins by locating the spinal cord, after which it creates the bounding box coordinates. Our technique is validated using four separate clinical datasets. The results of the experiments that used a unique segmentation strategy that was dependent on MRI images reveal that the algorithm that was presented for the system delivers a higher level of accuracy when compared to the other algorithms that are already in use.
    Keywords: spinal cord segmentation; bounding box; U-Net architecture; deep learning.
    DOI: 10.1504/IJMEI.2022.10050461
     
  • Deep learning-based malignancy prediction in thyroid nodules   Order a copy of this article
    by L. Mohana Sundari, M.S. Maharajan, T. Senthil Kumar, Leo John Baptist Andrews 
    Abstract: Although the vast majority of thyroid nodules are non-cancerous, determining whether or not a nodule is cancerous may be a difficult and time-consuming process that often involves unnecessary surgical events. In addition, we discussed the process of developing a model that might anticipate the presence of cancer in thyroid nodules by including a number of the core demographic and ultrasound parameters. A combined sensitivity and specificity score was used to assess the diagnostic performance, and their accuracy was compared to that of radiologists. The comparison between model prediction and expert evaluation reveals the benefit of our approach over human judgement in predicting thyroid nodule malignancy. The results of the experiments show that the suggested algorithm performs better. Nodules of TI-RADS category 4 were used. The area under the receiver operating characteristic curve in the validation dataset was 0.92 (with accuracy of 0.70, sensitivity of 0.81 and specificity of 0.58).
    Keywords: thyroid; AI; deep learning; medical imaging; deep learning; CNNs; preoperative diagnosis.
    DOI: 10.1504/IJMEI.2022.10050829
     
  • Empirical wavelet decomposition of photoplethysmographic signal for hypertension risk stratification and detection of diabetes mellitus using machine learning techniques   Order a copy of this article
    by Muzaffar Khan, Bikesh Kumar Singh, Neelamshobha Nirala 
    Abstract: Hypertension (HT) is a leading risk factor for cardiovascular disease (CVD), and the overlap of diabetes mellitus (DM) with hypertension can lead to severe complications. Presently, the diagnostic method for detecting hypertension and DM is unsuitable for large-scale screening. The proposed model uses a statistical feature extracted by decomposing the PPG signal into a sub-band signal using Empirical wavelet transform (EWT), a comparative study conducted between various soft and hard computing classification models. The highest accuracy achieved by sequential neural network for the three categories, namely normal (NT) vs. prehypertension (PHT), NT vs. hypertension type 1 (HT-I), NT vs. hypertension type 2 (HT-II) in terms of F1 scores is 78.9%, 91.2% and 94%, respectively and F1 score of 97.9% for detection of DM-II patients. We conclude that soft computing techniques such as deep learning neural networks have shown superior performance compared to hard computing techniques. Furthermore, features selected using a hybrid feature selection technique were found to improve the classifier’s performance. The main advantage of the proposed model that uses a decomposition technique is found to be more immune to noisy PPG signals, overcoming the limitation of the morphological-based model.
    Keywords: hypertension; diabetes mellitus; photoplethysmographic; empirical wavelet transform; Hilbert transform ensemble classifier; deep learning neural network.
    DOI: 10.1504/IJMEI.2022.10050838
     
  • Deep neural networks for medical image segmentation: geodesic distance transform   Order a copy of this article
    by P. Jenifer Darling Rosita, W. Stalin Jacob, R. Kalpana, T. Cynthia Anbuselvi 
    Abstract: The segmentation of medical images aids in managing the dose of medication, as well as the dosage of exposure to radiation, limiting the development of diseases like tumours, and monitoring the progression of diseases like cancer. The process of segmentation involves the division of a picture into distinct regions that each includes fragments of pixels that have similar characteristics. The regions should have a strong link to the items or elements of interest depicted in the picture in order to be expressive and useful for image analysis and interpretation. A proposal is made for an interactive framework that takes a deep learning approach. P-Net is the first stage of the framework, and it is used to produce an initial automated segmentation. The second stage is where the framework is implemented. This interaction is included in the input of the R-Net.
    Keywords: image segmentation; deep learning; brain tumour; datasets.
    DOI: 10.1504/IJMEI.2022.10050839
     
  • Detection of diabetic retinopathy severity from fundus images: DCNN   Order a copy of this article
    by T.Senthil Kumar, R. Muthalagu, L. Mohana Sundari, M. Nalini 
    Abstract: Diabetes retinopathy is a frequent diabetic complication that damages the retina and, if left untreated, may lead to blindness. The exponential rise in the number of diabetics throughout the globe has resulted in an equivalent rise in the number of diabetic retinopathy (DR) patients, one of the most serious consequences of diabetes. The goal of this research is to develop a hybrid solution approach for identifying diabetic retinopathy using retinal fundus pictures. The process of retinal vascular segmentation is critical for detecting a variety of eye disorders, such as the effects of diabetes on the eyes, also known as diabetic retinopathy. Morphologically based operations were used for the autoextraction of retinal blood vessels. Wavelet decomposition and back propagation neural networks were used to extract retinal vascular characteristics and evaluate the dataset that was used for this article. Morphologically based operations were also used for autoextraction of retinal blood vessels.
    Keywords: diabetic retinopathy; fundus images; retina; deep learning; image processing.
    DOI: 10.1504/IJMEI.2022.10050840
     
  • Detection and classification of Alzheimers using super-resolution algorithm and convolutional neural network   Order a copy of this article
    by T. Senthil Kumar, Ashok Vajravelu, R. Muthalagu, P. Sri Latha 
    Abstract: On the basis of data obtained from brain imaging, a number of different machine learning (ML) methods may be used to categorise Alzheimer’s disease (AD). Convolutional neural networks have been suggested for the classification of Alzheimer’s disease based on anatomical MRI in more than 30 different studies. Since the frameworks and implementation details of many researchers aren’t available to the public, it makes them difficult to replicate. This article extracts the green channel initially, which is further enhanced by using super resolution algorithm. Convolutional neural network is applied to the contrast-enhanced image. We used CNN and T1-weighted MRI to broaden open-source solution for Alzheimer’s disease categorisation. Preprocessing, classification, and evaluation techniques for deep learning are included in the framework as well as tools for converting ADNI, AIBL, and OASIS data to the standard of the BIDS format. By combining deep learning with radionics, the accuracy of Alzheimer’s disease diagnosis is increased.
    Keywords: image classification; convolutional neural network; CNN; Alzheimer’s disease classification; magnetic resonance imaging; MRI.
    DOI: 10.1504/IJMEI.2022.10050896
     
  • Breast cancer diagnosis by hybrid fuzzy CNN network   Order a copy of this article
    by W. Stalin Jacob, P. Jenifer Darling Rosita, M. Sri Geetha, P. Jagadeesh, Sivakumaran Chandrasekaran 
    Abstract: Breast cancer is a common gynaecological ailment that affects women all over the world. Early identification of this disease has been shown to be extremely beneficial in terms of therapy. Mammographic pictures are analysed in this article utilising image processing methods and a pipeline structure to see whether they contain malignant tumours, which are subsequently categorised. The SVM classifier is used for classification, and it is fed by the characteristics that have been picked. It is supported by a number of kernel functions. This differs from standard machine learning classification and optimisation strategies, and it is shown in a unique manner. The outcomes of the actualised computer-aided diagnostic (CAD) learning process are analysed in order to determine whether or not it was successful. The BCDR-F03 dataset is evaluated, as well as the: 1) local mammographic dataset; 2) colony optimisation-based multi-layer perceptron (ACO-MLP) dataset.
    Keywords: breast cancer; deep learning; convolution neural network; CNN; prediction; benign and malignant; computer-aided diagnostic; CAD.
    DOI: 10.1504/IJMEI.2022.10050897
     
  • Robust liver segmentation using marker controlled watershed transform   Order a copy of this article
    by Mohammad Anwarul Siddique, Shailendra Kumar Singh, Moin Hasan 
    Abstract: The liver is the body’s largest organ, and it is largely responsible for metabolism and detoxification. In computer vision-based biomedical image analysis, liver segmentation is a critical step in detecting liver cancer. Due to the complicated structure of abdominal computed tomography (CT) images, noise, and textural differences across the image, liver segmentation is a key task that results in under-segmentation and over-segmentation. This paper uses a marker-based watershed transform to segment the liver in abdominal CT images. The double stage Gaussian filter with texture and contrast enhancement (DSGFTCE) is used to improve image quality at the pre-processing stage. The performance of the proposed segmentation is assessed using various performance evaluation metrics such as dice score (DS), volume overlapping error (VOE), Jacquard index (JI) and relative volume difference on LiTS dataset. The performance comparison with previous state of arts shows that proposed liver segmentation scheme provides better results (DS = 0.968, VOE = 0.089, JI = 0.9379, RVD = 0.09) compared with existing techniques.
    Keywords: liver segmentation; contrast enhancement; texture smoothening; watershed transform; Gaussian filtering; computer tomography.
    DOI: 10.1504/IJMEI.2022.10050979
     
  • Trimester analysis for detecting abnormalities in foetal growth based on femur length using LabVIEW   Order a copy of this article
    by D. Ruban Thomas, V. Vasanth, R. Kishore, B. Kishore Kumar 
    Abstract: Foetal development characteristics are used to determine the effects of various maternal traits and identify people who are more likely to have adverse outcomes. Foetal development features were measured during each trimester of pregnancy and were impacted by several mothers socio-demographic and lifestyle factors. Ultrasound measures of the foetus are crucial during pregnancy. The stability of foetal growth features can be assessed using tracking analyses throughout pregnancy. The length of the femur, the long bone in the thigh, is one of the markers used to monitor foetal growth and health. Previous research concentrated on birth outcomes compared to the foetal development, during pregnancy. In this work, the femur length is measured using LabVIEW. An ultrasound finding of short femur length may signal the need for additional testing to rule out illnesses. Femur length is one of the several factors for determining a babys health.
    Keywords: Laboratory Virtual Instrument Engineering Workbench; LabVIEW; virtual instruments; VI; trimester; ultrasound; ultrasonography; foetal growth; femur length; pregnancy.
    DOI: 10.1504/IJMEI.2022.10051334
     
  • Analyses of non-steroidal anti-inflammatory drug induced GI bleed related hospitalisation discharges from 2016 to 2018 in the USA   Order a copy of this article
    by Chancey Sweeney, Shankar Srinivasan, Dinesh P. Mital, Riddhi Vyas 
    Abstract: This study determines the predictors of gastrointestinal (GI) bleeds induced by NSAIDs and their relation to length of hospital stay and total charges. A descriptive analysis showed the highest incidence as non-Hispanic White, older adult patients with variation in household income. Risk factors include Helicobacter pylori, alcohol, and smoking. The overall prediction of study outcomes showed associations between independent and dependent variables such as age, race, and gender, in relation to length of hospital stay and total charges. A high influence was detected with interaction of predictors such as health and socio-demographic characteristics, like advanced age, gender, race, and smoking.
    Keywords: non-steroidal anti-inflammatory drug; NSAID; gastrointestinal bleed; NSAID induced; GI bleed; hospitalisation discharges.
    DOI: 10.1504/IJMEI.2022.10051360
     
  • A deep analysis of chronic kidney disease for early detection using machine learning classifiers   Order a copy of this article
    by Saurabh Pal 
    Abstract: All patients who have suffered by chronic kidney disease (CKD) are not easily identified due to symptoms on early stage. The main objective of this research paper is to develop a CKD prediction model which can give better accuracy as compared to other studies. In this paper, we have organised CKD datasets from UCI machine learning repository. Most significant CKD features are eliminated by recursive feature elimination techniques. We have trained base classifiers RF, K-NN, MLP and SVM on 70% disease dataset and test on 30% dataset. Bagging and voting ensemble methods are used to enhance the prediction model. The proposed bagging ensemble model outperformed the other classifiers by achieving 97% accuracy.
    Keywords: chronic kidney disease; CKD; recursive feature elimination; RFE; random forest; K-nearest neighbour; multilayer perception; support vector machine; voting and bagging ensemble classifier.
    DOI: 10.1504/IJMEI.2022.10051462
     
  • Developments in technology assisted gait analysis in post-knee arthroplasty   Order a copy of this article
    by Sumit Raghav, Anshika Singh, Shashwat Pathak, Suresh Mani, Mukul Kumar 
    Abstract: Early monitoring in knee arthroplasty is a critical issue to tackle deviation from expected healing, patient satisfaction and ensuring quality of life. There are various methods suggested and implemented over years with varied degree of performance. This paper presents a relevant review of technology assisted gait analysis in knee arthroplasty. The systematic search revealed 272 studies, of which 13 were added retrospectively through reference screening of the included articles. After title and abstract screening, only 20 studies were included in this review. This review paper provides a comprehensive overview of applications of technology assisted gait analysis to monitor and quantify the status of waking. There is moderate-quality of evidence showed technology-assisted; in particular, sensor-based technology, motion sensors and motion analysis results in a statistically significant improvement in monitoring of gait parameters.
    Keywords: gait analysis; knee arthroplasty; gait parameters.
    DOI: 10.1504/IJMEI.2022.10051736
     
  • The new development in the lower limb exoskeleton - a review   Order a copy of this article
    by Jun-Yi Ge, Maysam F. Abbod, Jiann-Shing Shieh 
    Abstract: The lower limb exoskeleton is a human machine interaction system which is designed to help users perform movements involving the lower limb by providing assistive torque. This technology has been proposed and developed in recent years. It is a comprehensive application of biomechanics, automatic control, information science and so on. In this review, we summarised the characteristics of exoskeletons that have emerged in the last twenty years. And at the end of the paper, we propose some possible solutions of the existing exoskeleton problems. This may be a guideline for further research on lower limb exoskeleton.
    Keywords: lower limb; exoskeleton; biomechanics; automatic control.
    DOI: 10.1504/IJMEI.2022.10051870
     
  • mUNET: glioma segmentation with delineation of tumour sub-regions using optimised architecture of UNET   Order a copy of this article
    by Sonal Gore, Jayant Jagtap 
    Abstract: Glioma tumour is aggressive due to infiltrative nature, that exhibits unpredictable shape, size, morphology. Therefore, its manual segmentation from MRI remains challenging task. Study presents automated segmentation using T2-FLAIR data of BRATS challenge by proposing modified UNET (mUNET) with optimised architecture to segment tumour in enhancing, necrosis and edema sub-regions. Work was evaluated using six test cases by proposing slight modifications in UNET. Experimentation has gained comparable results with slightly better accuracy of 99.39% and loss of 0.0054, as compared to original UNET. Moreover, test cases were capable to complete segmentation at speed of 0.713 milliseconds per image.
    Keywords: brain tumour; glioma; MRI; FLAIR; segmentation; deep learning; convolutional neural network; modified UNET; mUNET: BRATS data; tumour subregions.
    DOI: 10.1504/IJMEI.2022.10051871
     
  • Machine learning-based framework for early prediction of diabetes   Order a copy of this article
    by Salliah Bhat Shafi, Venkatesan Selvam, Gufran Ahmad Ansari 
    Abstract: The recent advancements in technology have changed the landscape of healthcare. With changes in lifestyle and rise in living standard diabetes remains leading cause of death globally. Prediction of diabetes using machine learning algorithms (MLA) for early prediction is need of the hour. However, it is still in its nascent stage. The goal of this study is to employ significant features of machine learning algorithms to the prediction of diabetic and to get the best results which are compared to clinical outcomes. Using predictive analysis, the suggested strategy focuses on choosing the features that aid in the early detection of diabetes. The result shows that the support vector machine (SVM) algorithm has the highest accuracy of 99.349%as compared to naive Bayes which is 98.95%. In order to improve classification and accuracy, this research also normalises the selection of suitable features in the data.
    Keywords: machine learning; naive Bayes; support vector machine; SVM; prediction; diabetes; patients.
    DOI: 10.1504/IJMEI.2022.10052157
     
  • Implementing machine learning techniques to predict bipolar disorder   Order a copy of this article
    by Nisha Agnihotri, Sanjeev Kumar Prasad 
    Abstract: Bipolar disorder (BD) is a mental and psychiatric disorder which is characterised by alternate mode swings between mania and depression is very common these days. The classification, modelling and characterisation and diagnosis of these mental disorders are important in medical research. An unexpected and unexplored area in BD is to judge the non-verbal behaviour of person accurately. Therefore, this paper address the challenges of detecting BD state by machine learning (ML) techniques to test the non-verbal behaviours activities like various facial expressions, voice recordings and body gestures of mentally ill and controlled persons in a whole spectrum. ML techniques can potentially provide new horizons in diagnosing and treating in mental healthcare. Further, this paper aims to present commonly used algorithms such as decision trees (DT), support vector mechanism (SVM), logistics regression (LR), K-nearest neighbours (KNN), etc. and describe their properties and performances which could act as a guide to select appropriate models. The study shows that people with controlled state behaves significantly different as compared to BD patients in their interpersonal accuracy (IPA). This develops a new training program to improve better understanding and psychosocial functionality in their rehabilitation.
    Keywords: machine learning; mood disorder; anxiety depression; bipolar disorder-I and II; Python; interpersonal accuracy.
    DOI: 10.1504/IJMEI.2022.10052205
     
  • Prediction of epileptic seizure using deep learning architectures   Order a copy of this article
    by Vajravelu Ashok, J. Anitha, Isabel De La Torre Díez, D. Jude Hemanth 
    Abstract: Neurological illnesses such as epilepsy are among the most frequent. Epileptic sufferers’ lives are greatly impacted by early warnings of impending seizures. Using electroencephalogram signals, this research aims to create an epileptic seizure prediction algorithm that can automatically identify an epileptic seizure. Early seizure prediction using EEG data is now possible thanks to the latest machine learning algorithms. An average AUC of 0.74 is achieved by the new technique, compared to 0.72 for the state-of-the-art approach, a 3.25-fold increase in computing time. In-depth knowledge of seizure detection, classification, and potential future research areas may be gained through this presentation’s cutting-edge methodologies and concepts. Predicting seizure activity might benefit from a modified atom search optimisation-based deep recurrent neural network. Numerous hidden layers are used by the deep recurrent neural network (DRNN) classifier to predict seizure activity.
    Keywords: electroencephalogram; EEG; signal processing; epilepsy; CHBMIT dataset.
    DOI: 10.1504/IJMEI.2022.10052827
     
  • Segmentation of retinal features in colour fundus images   Order a copy of this article
    by N. Bino, P.A. Haris, O. Sheeba 
    Abstract: Retinal diseases could be diagnosed by the variations in size, shape and texture of features in retinal images. Colour fundus images are widely used for the diagnosis of retinal diseases like diabetic retinopathy (DR), glaucoma, age-related macular degeneration (AMD), etc. Image processing enhances the input image, which segments and highlights regions of interest and quantify it. In this paper, we have segmented optic disc and blood vessels in colour fundus images using an algorithm coded in MATLAB. The input images and their corresponding ground truth images were obtained from public datasets available in Kaggle. The validation of the segmented images is done for parameters like accuracy, sensitivity, precision, F-measure, Mathews correlation coefficient, Dice coefficient, Jaccard index and specificity. The algorithm could effectively segment images, with an average accuracy of 99.74% in segmenting optic disc and 94.74% that of blood vessels.
    Keywords: retinal image segmentation; colour fundus imaging; optic disc; retinal blood vessels.
    DOI: 10.1504/IJMEI.2022.10052912
     
  • Hybrid multi-kernel SVM algorithm for microaneurysm recognition in colour fundus images   Order a copy of this article
    by S.B. Mohan, B. Kannan, D. Ravikumar, C. Sivakumaran 
    Abstract: Diabetic retinopathy (DR) is a condition that results in impaired vision and is triggered by diabetic. These alterations in the retinal vessels may be traced back to hyperglycemia. Microaneurysms (MAs) seem to be the initial disease symptoms of DR, and a prompt identification of microaneurysms may assist in the detection of DR in the preclinical phase. Optical coherence imaging system, often known as OCT, is a non-invasive imaging technology that offers a cross-sectional picture of the retinal. We construct new probability foundations for support vector machines (SVMs) using informational differences and the Fisher score. The findings that were achieved were as follows: the accuracy criterion had a score of 96.32%, the sensitivity criterion had a score of 97.34%, the specificity criterion had a score of 95.42%, and the precision criterion had a score of 95.27%.
    Keywords: retinal pictures; scaled Dirichlet combination; support vector machines; SVMs; microaneurysm images.
    DOI: 10.1504/IJMEI.2023.10053229
     
  • Application of deep learning to segment pelvis bones   Order a copy of this article
    by N. Thamaraikannan, M. Saravanan, N.K. Anushkannan, S. Ramesh, C. Sivakumaran 
    Abstract: The proper identification and localisation of pelvic bone metastases begins with precise segmentation of the pelvic bones. Existing pelvic bone segmentation algorithms are generally manual or semi-automatic, and they exhibit low accuracy when dealing with picture appearance changes caused by multi-site domain shifts, etc. This paper presents a strategy for segmenting normal pelvic bone characteristics in multiparametric magnetic resonance imaging (mpMRI) using a 3D U-Net based on deep learning. Extensive testing on our dataset indicates the usefulness of our automated technique. The 3D U-Net network, based on deep learning, offers patients accurate identification and segmentation of pelvic bone metastases.
    Keywords: U-Net; pelvis bone; segmentation; biomedical signal processing.
    DOI: 10.1504/IJMEI.2023.10053901
     
  • Psychological distress and its association with socioeconomic and health behaviour factors   Order a copy of this article
    by Xiangdong An, Hui-chuan Chen, Monty Taylor 
    Abstract: The authors assess the relationships among psychological distress, socioeconomic status, and health behaviour factors. Data from the Canadian Community Health Survey (CCHS) dataset was analysed for this study using hierarchical regression and ANOVA. Results show that daily energy expenditure is positively related with psychological distress. The authors also find that daily fruit consumption, daily green salad consumption, and monthly frequency of leisure activities are negatively associated with psychological distress. Furthermore, the results indicate that people with personal incomes of less than $20,000 perceive significantly higher distress than people with higher incomes. Among five occupation groups, the sales and services group perceives significantly higher distress compared to the other four groups. The four other occupation groups as categorised by CCHS do not perceive significantly different distress levels.
    Keywords: psychological distress; gender; occupation; income; health behaviour; fruit consumption; green salad consumption; leisure activities; energy expenditure.
    DOI: 10.1504/IJMEI.2023.10053931
     
  • A novel method to predict stroke using deep learning approach   Order a copy of this article
    by Swagata Sarkar, A. Jayashree, R. Thirumurugan, C. Sivakumaran 
    Abstract: Strokes remain the main cause of major impairment that lasts for lengthy period of time and ranks as the third greatest cause of mortality overall in United States. The ability to accurately anticipate a stroke is very useful for facilitating earlier intervention as well as treatments. Several studies have concentrated on determining the chance of a heart attack; on the other hand, only few have investigated the possibility of a brain attack. Various machine learning techniques are being created to estimate the chance of a cerebral infarction. We present a data-driven classifiers deep neural Networks (ResNet) for diagnosing strokes predicated on 12-leads ECG information. The quality of the model, which has been fine-tuned to perfection, allows us to achieve a training correctness of 99.99% and an accuracy rate of 85.82%. The findings imply that ECG is a viable adjunct tool for diagnosing stroke.
    Keywords: machine learning; stroke prediction; CNN; deep learning; ResNet.
    DOI: 10.1504/IJMEI.2023.10053942
     
  • Face recognition-based security system for automated teller machine using deep face model   Order a copy of this article
    by K. Kavin Kumar, S. Mythili, S. Prabhu Kumar 
    Abstract: The purpose of this study is to show about using OpenCV and deep learning techniques to design and implement a face recognition-based ATM security system. Face recognition only provides service to the user if the user is authentic or if the user has been validated by an authentic ATM card user. Users are authenticated by comparing the persons video taken in front of the ATM. The proposed method identifies the right persons by comparing the blink of eyes. If someone enter the ATM with photocopy of another person, checking the blink of eyes the validation is made. If the user is authentic, to strengthen the models accuracy, the current image is being utilised. A web link is sent to the registered mobile number that owns the ATM card, to verify the access of the illegitimate user to his/her account only then the user is considered as a legitimate user. Histogram algorithms and deep learning approaches are used by the system to identify persons. To process the image and detect the faces in the image, this system use the OpenCV package. Face recognition is achieved through the use of an open CV and deep learning.
    Keywords: OpenCV; blink of eyes; VGG Face model and deep face.
    DOI: 10.1504/IJMEI.2023.10054109
     
  • Detection and classification of lung cancer using deep neural network   Order a copy of this article
    by S. Babu Kumar, M. Vinoth Kumar 
    Abstract: Lung cancers hold a critical spot among the reasons for most cancer deaths among humans. The better way to maximise the survival rate is the detection of cancer at the earliest. But existing traditional techniques are time-consuming and error-prone. This study is a significant and efficient method for the detection and classification of lung cancer into large cell carcinomas, small cell, adenocarcinoma, squamous cell carcinomas, or benign respectively. In the proposed technique, a novel algorithm is implemented to generate the Image patches from whole slide histopathological images. Then, histogram normalisation is carried out to remove noise and enhance the image. Then a novel extended Mobius transformation technique is used for image augmentation. Finally, Dense EfficientNetB7 is trained to extract the features for the detection and classification of lung cancer. The performance of the proposed technique is proved more efficient and par with histologists attaining an accuracy of 98.87%.
    Keywords: lung cancer; histopathology; deep convolution neural network; DCNN; EfficientNetB7.
    DOI: 10.1504/IJMEI.2023.10054175
     
  • A review of the relationship between flow-volume curve and obstructive sleep apnea   Order a copy of this article
    by Seval Bulut Eris, Cahit Bilgin, Ömer Eriş, Mehmet Recep Bozkurt 
    Abstract: This study presents a systematic review of studies in the literature evaluating the relationship between the flow-volume curve and obstructive sleep apnea (OSA). When the literature was examined, it was observed that there are few studies in this field, and only statistical analyses were made using the ready parameters given by the pulmonary function test (PFT). New biomarkers can be discovered with characteristic and statistical features extracted from the flow-volume curve to diagnose OSA. Machine learning-based systems can be developed using biomarkers. Rules-based automatic diagnosis systems can enable faster OSA diagnosis. In addition, it can eliminate the disadvantages arising from using a sleep laboratory to diagnose OSA.
    Keywords: obstructive sleep apnea; OSA; flow-volume curve; flow-volume loop; saw-tooth pattern; pulmonary function test; PFT; machine learning.
    DOI: 10.1504/IJMEI.2023.10054791
     
  • Design and analysis of microfluidic channel with W-shaped electrodes for separation of bio-particles using dielectrophoresis technique   Order a copy of this article
    by Muktimani Brahma, R. Kumar, Trigunesh Narzary 
    Abstract: A unique microfluidic device with W-shaped electrodes is designed to separate the micro-bio-particles. The microfluidic device efficiently separates the red blood cells (RBCs) and platelets using the dielectrophoresis technique. The separation process has been analysed by using COMSOL Multiphysics software and found that the electrode with the minimum gap of 25 um acquired optimum results at an applied voltage of 8 V. The proposed device gave a high separation efficiency of 89.5% and a separation purity of 99.63%. The enhanced microfluidic device gave better results compared to other recent works and is exceptionally suitable for separation of micro-bio-particles.
    Keywords: dielectrophoresis; red blood cells; RBCs; platelets; separation; microchannel; electrodes; flow rate; electric field; non-uniform; force.
    DOI: 10.1504/IJMEI.2023.10054875
     
  • Diffusion tensor imaging for Alzheimer’s disease classification using a bag of features and majority voting   Order a copy of this article
    by Latifa Houria, Noureddine Belkhamsa, Assia Cherfa, Yazid Cherfa 
    Abstract: Alzheimer’s disease (AD) is a neurodegenerative disease and the most common cause of dementia. Thus, various neuroimaging-based methods were proposed to detect this disease at its early stage, called mild cognitive impairment (MCI). We developed a novel approach combining Diffusion tensor imaging (DTI)-indices and ensemble learning to classify AD. A bag of features (BoF) is used to retrieve the locale features, and a support vector machine (SVM) is applied for classification. The majority voting technique is used to combine the final predicted labels. The proposed method achieves an accuracy of 94.0%, 97.0%, and 95.9% to classify CN vs. MCI, CN vs. AD, and MCI vs. AD, respectively.
    Keywords: diffusion tensor imaging; Alzheimer’s disease; bag of features; BoF; support vector machine; SVM; majority voting.
    DOI: 10.1504/IJMEI.2023.10054939
     
  • An intelligent method for predicting cardiac disease based on PSO-convolutional neural network   Order a copy of this article
    by M. Balamurugan, P. Blessed Prince 
    Abstract: Cardiovascular disease (also known as CVD) is one of the primary contributors to both morbidity and death. The present state of the art in artificial intelligence plays a significant part in the process of aiding medical professionals in the diagnosis of a variety of disorders. A hybrid framework is proposed for the diagnosis of cardiovascular illnesses by analysing medical voice data. Eleven datasets comprising 14,416 numerical characteristics may be produced by using the method that has been suggested. From the datasets that are produced as a consequence, numerical and graphical characteristics are extracted. In the third layer, numerical data is provided to five separate machine learning (ML) techniques and graphical characteristics are transmitted to convolutional neural networks (CNNs), with transfer learning utilised to choose the best suited configurations. From a benchmark dataset, the PSO-CNN technique that was presented obtains an accuracy of 96.78%.
    Keywords: cardiovascular; PSO-CNN; optimisation; machine learning.
    DOI: 10.1504/IJMEI.2023.10054940
     
  • Data mining approach for nutrition score calculation of geriatric patients   Order a copy of this article
    by Vaishali P. Suryawanshi, Rashmi S. Phalnikar 
    Abstract: Due to lack of uniformity in terms of understanding nutrition status, comorbidity types, patient mobility and BMI readings mal-nutrition screening forms cannot be used to correctly comprehend each patients nutritional status. The goal of this study is to use a data mining approach to understand nutrition screening forms in order to determine the malnutrition status of hospitalised geriatric patients. The evaluation measures are used to indicate the accuracy of the classification algorithms. The research work on nutrition screening form, which is designed using classification algorithms, will help physicians to understand their patients nutritional status and help in avoiding malnutrition in them.
    Keywords: malnutrition screening forms; classification; data mining; evaluation measures.
    DOI: 10.1504/IJMEI.2023.10055068
     
  • Integer wavelet transform based data hiding scheme for medical image protection   Order a copy of this article
    by Moad Med Sayad, Zermi Narima, Khaldi Amine, Kafi Med Redouane 
    Abstract: A crucial security and protection challenge today in telemedicine is how to ensure the confidentiality of sensitive data sent over the internet and limit access to specified designated information. To keep such details private, we propose in this work a robust and blind medical image watermarking approach that combines integer wavelet transform (IWT) and singular value decomposition (SVD). Low embedding capacity is a potential drawback of modern integer wavelet transform (IWT)-based watermarking systems. A secure high capacity watermarking scheme based on IWT is proposed in this work to overcome this specific limitation. According to experiment results on imperceptibility and robustness, the proposed solution efficiently maintains a significant quality of watermarked images and the watermark is resistant to the most commonly used attacks in watermarking. The integration of information about the patient in his medical image will avoid any confusion between two images.
    Keywords: medical image; digital watermarking; blind watermarking; integer wavelet transform; IWT; singular value decomposition; SVD; QR code.
    DOI: 10.1504/IJMEI.2023.10055240
     
  • Segnet and U-Net based brain tumour segmentation   Order a copy of this article
    by R. Ashwini, Swagata Sarkar, C. Pandi, S. Rajalakshmi 
    Abstract: The process of separating individual brain tumours in diagnostic pictures is an essential component of therapeutic therapy. The manual segmentation process takes a lot of time and requires a lot of work, while the current automated segmentation techniques have problems such as a large number of parameters and a lack of accuracy. We created a completely automated technique for the segmentation of brain tumours by applying deep learning. The system was tested on 285 examples of brain tumours using multi-parametric magnet resonant images taken from either the BraTS2018 data set. The quantitative study of brain tumours is helpful in gaining a better knowledge of the features of the tumour as well as in developing more effective treatment strategies. Through the use of this technology, it was possible to get mean dice values of 0.9213 for the overall tumour and 0.8729 for the tumour core.
    Keywords: brain tumour; BraTS database; computer intelligence; Densenet; U-Net.
    DOI: 10.1504/IJMEI.2023.10055242
     
  • Myocardial infarction detection in late gadolinium enhancement cardiac MRI   Order a copy of this article
    by Sarra Dali Youcef, Mahammed Messadi 
    Abstract: Cardiac magnetic resonance imaging (MRI) has become the most used technique for assessing myocardial viability. Myocardial segmentation is a fundamental step in the detection of myocardial infarction (MI) on late gadolinium enhancement (LGE) images. In this paper, we provide a system for automated myocardial infarct detection. The myocardial segmentation is applied to cine images and then transferred to LGE images, to subsequently detect myocardial infarction. We tested our approach on the sunnybrook cardiac database. The proposed method shows remarkable accuracy. We obtained a dice similarity coefficient of 0.92 and an average perpendicular distance of 1.75 (mm) between automated and manual segmentation.
    Keywords: left ventricle; cine MR images; late gadolinium enhanced MR images; myocardial infarct; viability.
    DOI: 10.1504/IJMEI.2023.10055269
     
  • Medical image registration and automatic hippocampus segmentation through convolutional neural network   Order a copy of this article
    by S. Durga Prasad, K.S.N. Murthy, B. Kannan, C. Sivakumaran 
    Abstract: Alzheimer’s disease (AD) is a brain degenerative ailment that progresses and is irreversible. Mild cognitive impairment, known as MCI, is a clinical indicator that AD may eventually develop. In order to effectively treat and prevent AD, an accurate diagnosis of the illness’s early stages is required. AD often manifests its symptoms first in the hippocampus. Deep machine learning was used in this project with the intention of achieving its aim of segmenting a specific region. The suggested method’s performance was compared to manual segmentation using similarity measures. The performance of a CNN that segments the hippocampus directly is inferior to that of any contouring technique and the findings were 96% accurate. The quantitative results are improved by the application of stringent corrections to the data, although the gap is still rather large. The suggested technique is promising and may be expanded in AD prognosis by predicting hippocampal volume changes in the early stages of the illness.
    Keywords: Alzheimer’s disease; hippocampus; magnetic resonance imaging; convolutional neural network images; TBI; U-net.
    DOI: 10.1504/IJMEI.2023.10055511
     
  • CNN-based detection of cervical spinal cord injury   Order a copy of this article
    by G. Chandraiah, K. Mekala Devi, A. Mohamed Abbas, S. Rajalakshmi 
    Abstract: Magnetic resonance imaging (MRI) has the ability to infer alterations on a microscopic level in lesions that are present on the spinal cord. For the purpose of locating lesions brought on by cervical illnesses, our deep neural network using MRI was suggested. The segmentation of spine pictures, as well as their identification, diagnosis, as well as quantitative assessment, have all seen significant applications of the deep learning technology. The most cutting-edge approach to machine learning using medical imaging data is called convolutional neural networks (CNNs), which are powered by deep learning. The proposed network produces segmentation results that are in high degree of agreement with the real world situation. The suggested network produces outstanding results on testing. These findings are based upon that pixel level. The machine learning network that was suggested is both efficient and reliable for doing completely autonomous segmentation of the problematic area.
    Keywords: analysis of the spine province; magnetic resonance imaging; MRI; spinal cord; convolutional neural networks; CNNs.
    DOI: 10.1504/IJMEI.2023.10055578
     
  • Pneumonia detection using modified VGG 19 net architecture: application of deep convolutional neural network   Order a copy of this article
    by K. Sivakumar, P. Vinitha Baby, Lourdu Jennifer J.R., C. Sivakumaran 
    Abstract: Pneumonia is a somewhat common condition that may be brought on by a variety of micro-organisms, including bacteria, viruses, and fungus. The use of a deep convolutional neural network fed with X-ray images to detect lung pneumonia has been proposed. Collecting data, enhancing images, performing an adaptive and precise ROI evaluation, extracting features, and diagnosing diseases are all part of the framework’s scope. The suggested deep CNN models were trained using 12,000 pictures from the Pneumonia chest X-ray Dataset, which contained both infected and uninfected chest X-rays. The publicly available new CNN models were trained, and the results were compared with ensemble approach. On the validation set, the suggested technique demonstrated its superiority with an AUC of 95.21 and a sensitivity of 97.76. The proposed technique had a 90.71% success rate in properly identifying images of a patient’s chest as normal, viral pneumonia, or bacterial pneumonia.
    Keywords: pneumonia; deep learning; machine learning; CNN.
    DOI: 10.1504/IJMEI.2023.10055646
     
  • Effects of various phases and errors on partial volume estimation in the analysis of PV correction in ASL MRI   Order a copy of this article
    by A. Shyna, Amma C. Ushadevi, Ansamma John, C. Kesavadas, Bejoy Thomas 
    Abstract: Arterial spin labelling (ASL) MRI is widely used to measure cerebral blood flow (CBF). The low resolution of ASL images causes partial volume (PV) effect that causes an erroneous estimation of CBF values. A two-phase study is conducted to investigate the impact of various factors on the PV correction technique using linear regression. The effects of different stages in the PV estimation process are examined in the first phase, and the impact of noise and blurring on PV estimates on the second phase. Experiments are conducted on simulated PCASL images generated from ADNI and the results are validated using RMSE score and ROI analysis.
    Keywords: arterial spin labelling MRI; partial volume effect; cerebral blood flow; CBF; linear regression.
    DOI: 10.1504/IJMEI.2023.10055831
     
  • Motor imagery classification based upon wavelet packet decomposition and binary particle swarm optimisation   Order a copy of this article
    by Rohtash Dhiman, Pawan 
    Abstract: Motor imagery (MI)-electroencephalogram (EEG) signals are most convenient basis for brain computer interface (BCI) systems. This investigation aims to develop a novel method for feature extraction and selection that improves classification accuracy in the BCI system. Wavelet packet decomposition (WPD) and approximate entropy (ApEn) are used for feature extraction for EEG signals. Binary particle swarm optimisation (BPSO) technique is used for feature selection. Artificial neural network (ANN) is employed for classifying MI activities with mean accuracy of 87.29%. The proposed methodology can be useful in clinical applications for motor-disabled subjects connecting outside environment with sequences of MI activities.
    Keywords: approximate entropy; ApEn; artificial neural network; ANN; binary particle swarm optimisation; BPSO; brain computer interface; BCI; electroencephalogram; EEG; motor imagery; MI; wavelet packet decomposition; WPD.
    DOI: 10.1504/IJMEI.2023.10056105
     
  • Radial basis function network equipped with an ensemble-based Lasso ridge model in diagnosis of breast cancer   Order a copy of this article
    by Pooja J. Shah, Trupti P. Shah 
    Abstract: The information that is kept in the form of medical records is of tremendous assistance when it comes to the process of putting together medical decision support systems. The goal of this study is to provide a description of two distinct methods for the development of a medical diagnostic tool that is driven by data in the detection of breast cancer. The radial basis function network (RBFN) architecture with Lasso and ridge regularisation, as well as ensemble learning, are both methods that have been proposed as potential solutions. Following the implementation of the proposed networks on the Wisconsin Breast Cancer (WBC) dataset, comparative analysis is carried out.
    Keywords: radial basis function network; RBFN; Lasso and ridge regularisation; LR; ensemble learning; breast cancer; BC.
    DOI: 10.1504/IJMEI.2023.10056107
     
  • A new hybrid approach for automated leakage detection in fundus fluorescein angiography images   Order a copy of this article
    by Bikesh K. Singh, Suraj Chandrakar, Ankur Khare, Ravi Khatri, Deepak Chandravanshi, Sumit Kumar Banchhor 
    Abstract: The existing methods for quantitative investigation of leakage in fluorescein angiography (FA) images involve manual delineation and lack annotated datasets and high accuracy. The proposed hybrid approach can automatically detect leakage in FA images with high accuracy. The hybrid approach combines the fuzzy c-means (FCM) with distance regularised level set evolution (DRLSE). Two expert radiologists test the accuracy of the proposed system. We have also tested inter-and intra-observer variability between the two experts. The study observed high accuracy of 0.98 +- 0.02 and 0.98 +- 0.03 with experts 1 and 2, respectively. Further, the accuracy of manual inter-observer (expert 1 and expert 2) and intra-observer (expert 1) were observed as 0.95 += 0.05 and 0.98 +- 0.02, respectively. The proposed fully automated hybrid approach can accurately detect and quantify the leakage in FA images and thus has great potential value in diabetic retinopathy (DR).
    Keywords: diabetic retinopathy; DR; fluorescein angiography; FA; leakage; segmentation; inter- and intra-observer variability; fuzzy c-means; FCM; distance regularised level set evolution; DRLSE.
    DOI: 10.1504/IJMEI.2023.10056108
     
  • Remotely operated infant incubator   Order a copy of this article
    by Maham Sarvat, Suhaib Masroor, Jawaid Shabbir, Zohra Jabeen, Bilal Ahmad 
    Abstract: A neonatal incubator is a sealed enclosure where a child may be placed for treatment in a safe and healthy environment. The system can provide a heater, a fan, a water container for applying moisture, a control valve to regulate oxygen supply, and access ports for nursing care. In this paper, a novel approach is presented to create a cost effective wireless portable baby incubator. It is shown that the control and operation of the incubator is performed wirelessly from the nursing counter, located within the range of 30 m via an Android device. The device will provide information of all the installed functions on a single screen. Moreover, it requires only a single staff member to monitor them, and give them necessary instructions via the same Android device. To validate the efficiency of the proposed design, the incubator is simulated 30 times, and the acquired results are tested statistically by ANOVA.
    Keywords: infant incubator; remote monitoring and control; wireless communication; smart electro-medical devices.
    DOI: 10.1504/IJMEI.2023.10056665
     
  • Efficacy and bioequivalence study of Zithromax and Azithral a drug for the treatment COVID-19   Order a copy of this article
    by Hattab Youcef 
    Abstract: Generic drugs are increasingly marketed around the world to the detriment of their quality. The objective of this work was therefore to carry out a bioequivalence study between the originator Zithromax a drug for the treatment COVID-19 and its generic Azithral, this bioequivalence study is based on the quality control of generic and originator azithromycin tablets, which is carried out after validation of the azithromycin assay method, and essentially on pharmaco-technical tests (mass uniformity, breakability, disintegration, dissolution in vitro). By referring mainly to different pharmacopoeias, the tests have shown that the quality of generic tablets is acceptable but seems to be less satisfactory than that of originator tablets.
    Keywords: bioequivalence study; generic; azithromycin; COVID-19.
    DOI: 10.1504/IJMEI.2023.10056861
     
  • Multi perceptron network based model to classify the multi stages of Alzheimer’s disease using clinical data   Order a copy of this article
    by Chevvenahalli Rangegowda Nagarathna, Mohanchandra Kusuma 
    Abstract: Alzheimer’s is a neurodegenerative disease that diminishes the capability of individuals to perform their daily activities. It is an irreversible disease. Once it is started it cannot be cured, but early detection of the disease can help to slow down the progression of the disease by starting the treatment early. In this experiment, we use clinical study data available at the Alzheimer’s disease neuroimaging initiative (ADNI) dataset to detect the different stages of Alzheimer’s and forecast the duration required for conversion from mild cognitive impairment to Alzheimer’s disease (AD) and also classification of the disease is done. The clinical signs of AD are rendered by age, patient education details, the progression rate of disease, and cognitive information. Various machine learning techniques like multi perceptron networks, random forest, SVM, and decision tree classifiers are used to do binary and multi-classification of AD, late mild cognitive impairment (LMCI), early cognitive impairment (EMCI), and cognitive control (CN). The multi perceptron network shows the best performance by giving an average accuracy of 99.24% for binary classification and 93% for multi class classification. The proposed model enables early detection and also predicts the different stages of Alzheimer’s using the clinical dataset.
    Keywords: Alzheimer’s; dementia; MMSE; multi-layer perceptron; random forest; SVM; late mild cognitive impairment; LMCI; early cognitive impairment; EMCI.
    DOI: 10.1504/IJMEI.2023.10056957
     
  • Diagnosis of the COVID-19 with deep learning   Order a copy of this article
    by Moulud Demouche, Abderrahmane Baadache, Djamil Aissani, Taha Zerrouki 
    Abstract: Since the beginning of 2020, the COVID-19 virus has spread widely throughout the world. RT-PCR is used to diagnose the disease, but radiologists studied advances in chest CT scans to produce high-quality images. The purpose of this research is to develop a deep learning system for detecting the COVID-19 virus using chest computed tomography. We proposed three convolutional neural network (CNN) approaches to developed three models, and applied them to a publicly available COVID-19 screening chest CT scan dataset of 2,482 images. Our models have shown that they work by getting a high F1-score of 98.5% for the InceptionResNetV2 model.
    Keywords: COVID-19; convolutional neural network; CNN; deep learning; Xception; InceptionResNetV.
    DOI: 10.1504/IJMEI.2023.10057090
     
  • Deep learning for smart home security systems   Order a copy of this article
    by N. Ashokkumar 
    Abstract: Smart home security systems helps to monitor the homes from remote place. These systems were created with the goal of mitigating risks and, eventually, leading to the consolidation of a sense of security. We get off with an overview of the deep learning approach and an analysis of its significance in the evolution of machine learning. The most basic prerequisite for choosing the most appropriate way to build a smart home is face recognition. Kinect is used to generate the point cloud image of a human body in this technique. The data that has been received will be processed by the Arduino microcontroller, which will ultimately result in the intelligent management of various domestic electrical equipments. The experimental findings demonstrate that by analysing and processing human point cloud photos, the proposed algorithm is able to effectively recognise human attitude and operate domestic appliances and other operations.
    Keywords: smart home; machine learning; deep learning; DL; machine learning; ML; denial of service; DoS; intrusion detection system; IDS; internet of things; IoT.
    DOI: 10.1504/IJMEI.2023.10057160
     
  • Information domain approach to investigate the cardio-vascular, cardio-respiratory and vasculo-respiratory causal coupling to study gender and age-based changes   Order a copy of this article
    by Kirti Singh, Indu Saini, Neetu Sood 
    Abstract: This study presents a framework to quantify the coupling changes in cardiovascular, cardiorespiratory, and vasculorespiratory complexity using information domain approaches based on compensated transfer entropy (cTE). The dataset used for this research consists of the group of healthy young and old subjects. The proposed technique delivers significant coupling changes in healthy dataset based on age and gender in contribution of heart rate (HR), blood pressure (BP), and respiration (RESP). For validation of statistically significant values, the paired t-test is used. This study provides additional value to the prognostic and diagnostic approach in biomedical science.
    Keywords: heart rate variability; HRV; respiration; RESP; blood pressure; BP; RR interval; transfer entropy; TE.
    DOI: 10.1504/IJMEI.2023.10057240
     
  • Parkinson disease detection using ResNet-50: a CNN architecture   Order a copy of this article
    by N.K. Anushkannan, S. Nirmalkumar, L. Chithra, K. Manoharan 
    Abstract: A person’s motor and cognitive symptoms from Parkinson’s disease (PD) can change dramatically over time. Some of the signs of Parkinson’s disease are shared with more common illnesses including ageing and essential tremor, making diagnosis challenging. Lots of study has gone into figuring out the best ways to diagnose this illness. Using deep learning, recursive neural networks (RNNs), and convolutional neural networks (CNNs) that can differentiate between healthy and PD patients, this research hopes to automate the Parkinson’s disease (PD) diagnostic procedure. As such, this study intends to analyse numerous imaging and movement datasets to determine which Parkinson’s disease test is best in terms of its capacity to differentiate between individuals, as different datasets may emphasise different elements of the illness (notably cube and spiral pentagon datasets). Furthermore, this study will be utilised to compare imaging and time series datasets for their diagnostic utility in Parkinson’s disease.
    Keywords: Parkinson disease; deep learning; signal processing; convolutional neural network; CNN; ResNet-50.
    DOI: 10.1504/IJMEI.2023.10057278
     
  • Parkinsons detection based on combined ResNet architecture and LSTM   Order a copy of this article
    by M. Gokuldhev, Anjani Kumar, K. Kiruthika, B.R. Tapas Bapu 
    Abstract: In this research, a novel approach to the clinical condition of neurodegenerative disorders like Parkinson’s is presented. The proposed method uses a mix of deep networks that have been pre-trained as well as long-term and short-term memory (LSTM). A new model called PD-ResNet is constructed and based on the residual network (ResNet) architecture to understand the variations between people with Parkinson disease and healthy controls. In order to execute adoption of the obtained learnt representations across data originating from various medical contexts, a new loss functionality is presented as well as used in the development of the deep neural networks (DNNs). Experiments conducted on the clinic gait dataset demonstrate that our suggested model has good performance, with a correctness of 95.51%, an accuracy of 94.44%, a recalls of 96.59%, a sensitivity of 94.44%, as well as a F1 measure of 95.50%.
    Keywords: Parkinson illness; ResNet architecture; deep neural networks; DNNs; long short-term memory.
    DOI: 10.1504/IJMEI.2023.10057280
     
  • Forward neck posture on cervical pain among university students: effect of smartphone addiction   Order a copy of this article
    by Selvaraj Antoniraj, Hafizah Che Hassan, K. Baleswamy 
    Abstract: These days, smartphones are used for much more than just making and receiving phone calls and text messages; they can also be used to view and share media, surf the web, and send and receive electronic mail. A cross-sectional analysis of college students between the ages of 17 and 30, who were using smartphones for prolonged period of time. The survey was conducted through an online questionnaire. Simple random sampling was used to get 404 samples, 216 of which were men and 188 of which were women. There is a positive and highly significant (p < 0.01) association between forward neck posture and cervical pain in college students. Using a smartphone excessively is associated with increased risk of cervical pain (p < 0.01). The findings of this study reveal that the extended and extensive use of smartphones by university undergraduates is a key contributing cause of their neck pain.
    Keywords: cervical pain; neck posture; smartphone addiction; students.
    DOI: 10.1504/IJMEI.2023.10057778
     
  • An optimised machine learning approach using Intel oneAPI for detecting cardiovascular system failure   Order a copy of this article
    by Akshay Bhuvaneswari Ramakrishnan, A. Srilakshmi, Shriram K. Vasudevan 
    Abstract: Heart failure is a frequent illness that might result in circumstances that could be fatal. Early heart illness identification is essential for prompt treatment and better patient outcomes. In this study, we offer a multi-machine learning model, trained with Intel oneAPI, technique for the early diagnosis of heart illness. To choose the most pertinent patient characteristics that will be used as input for our machine learning models, we employ a genetic algorithm (GA). To forecast the risk of cardiac disease, we optimise the models using a genetic algorithm and Intel oneAPI. Our findings demonstrate the high accuracy of our approach, with linear discriminant analysis optimised with GA producing the most accurate model with a 92.34% accuracy rate.
    Keywords: heart failure; machine learning; genetic algorithm; optimisation; oneAPI.
    DOI: 10.1504/IJMEI.2023.10057866
     
  • Increasing the accuracy of oscillometric blood pressure measurement   Order a copy of this article
    by Y.M. Blessy, K. Rajalakshmi, R. Raj Anandh, Vajravelu Ashok 
    Abstract: The presence of unregulated high blood pressure (BP) is linked to an elevated chance of health problems, some of which may cause harm to a range of organs across the body. Among the most dangerous as well as widespread conditions is hypertension, which refers to increased blood pressures. Unfortunately, correct blood pressure readings need the use of many medical instruments. The proposed system compares the systolic and diastolic objectives of blood-pressure readings, which are obtained using analogue devices and those obtained using an electronically controlled blood pressure reading tool. The systolic measurement alone had a mean variance of 3,265 mmHg, while the diastolic measurement had a mean variance of 6,165 mmHg. The discrepancy is then included into these processes of calibrating the automated blood pressure measuring device. The accuracy of the output of the calibration reading tool is between 99.47% and 99.82%.
    Keywords: blood pressure measurement; non-invasive; micro controller.
    DOI: 10.1504/IJMEI.2023.10057971
     
  • Integrating dialysis machine functions using system design approach   Order a copy of this article
    by Vicky B. Sardar, Atul Sajgure, Neela R. Rajhans 
    Abstract: In new product development (NPD) process, system design thinking contributes significantly towards the effectiveness of the product. For effective system design solution, incremental improvements to the existing system at each element level should meet the user expectations. A detailed view of the inputs and outputs occurring at each component level of a hemodialysis machine is presented in this study. This research work provides suggestions to address the functions of existing dialysis machine for home use. The suggestions are based on systems approach. Risk mitigation plan is devised for preventing the critical accidents while designing the machine for home hemodialysis using Swiss Cheese approach.
    Keywords: new product development; NPD; system design; home hemodialysis; HHD; Swiss Cheese approach.
    DOI: 10.1504/IJMEI.2023.10057977
     
  • Classification of cardiac arrhythmia disease using deep learning auto encoder algorithm   Order a copy of this article
    by B. Nithyasundari 
    Abstract: An electrocardiogram (ECG) is a very important diagnostic tool for figuring out what is wrong with a person’s heartbeat. Arrhythmia has been put into groups in many different ways. Due to the fact that electrocardiogram (ECG) data are in a state of continual change, it might be challenging to employ standard handmade approaches. We offer a framework for deep learning that does not need a great deal of supervision and is geared toward the detection of arrhythmias (WSDL-AD). Three distinct ECG waveforms are selected from the arrhythmia database maintained by MIT and BIH in order to evaluate the proposed methodology. The primary objective of this research is to discover a deep learning method, with the end aim of classifying the three distinct heart ailments that have been selected. The highest rate of correct recognition that can be achieved is 98.51%, while the accuracy of testing is around 92%.
    Keywords: MIT-BIH dataset; electrocardiogram; ECG; arrythmia disease; deep learning.
    DOI: 10.1504/IJMEI.2023.10058004
     
  • Breast cancer prediction using whale optimisation algorithm and ANFIS classifier   Order a copy of this article
    by M. Sri Geetha, S. Navaneethan, G. Divya, B. Geethavani 
    Abstract: Breast cancer is one of the most prevalent forms of the disease that affects women and is responsible for billions of deaths globally. The adaptive clustering is for the purpose of image segmentation, followed by an adaptive neuro-fuzzy inference system (ANFIS) for the categorisation of images. In order to optimise the parameters, a deep learning-based method known as the whale optimisation algorithm (WOA) has been suggested. The breast cancer database for the state of Wisconsin comprised information of patients who had a known diagnosis. The ANFIS classifiers were provided with a training set of records of this kind in order to learn how to discriminate between new cases in the domain. In order to identify breast cancer, the ANFIS classifier was fed nine characteristics that define breast cancer indications. The suggested ANFIS model included both the adaptability of neural networks and the qualitative approach of fuzzy logic.
    Keywords: adaptive neuro-fuzzy inference system; ANFIS; whale optimisation algorithm; WOA.
    DOI: 10.1504/IJMEI.2023.10058260
     
  • Human emotion classification enabled by EEG signal analysis and machine learning   Order a copy of this article
    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
     
  • Brain tumour segmentation using deep learning method - inception U-net   Order a copy of this article
    by S.B. Mohan, B. Ramesh, G. Gurumoorthy, B. Kannan 
    Abstract: The objective of the procedure of segmentation for brain tumours is to provide an accurate delineation of the regions that are affected by the tumour. In recent years, deep learning algorithms have shown performance that is seen to be promising in the field of addressing a variety of difficulties pertaining to computer vision. With the assistance of a deep learning model that is supported by DenseNet and commencement, a highly autonomous process for managing the work of glioma division in pre-usable X-ray examinations has been developed. This method can handle the task of handling the task of segmentation of gliomas. The incorporation of inception modules resulted in a substantial improvement (p < 0.001) in the segmentation performance across the board including all glioblastoma sub-districts. The algorithm attained a high degree of accuracy while maintaining a high level of performance throughout all three of the BraTS 2018 datasets.
    Keywords: segmentation; brain tumour; computer vision; inception; DenseNet architecture.
    DOI: 10.1504/IJMEI.2023.10058262
     
  • Hybrid brain computer interface structures to control drone in 3D   Order a copy of this article
    by B. Divya, S. Bharathi, G.R. Mahendra Babu, K. Rajalakshmi, Nalini Mohan 
    Abstract: In order to overcome the shortcomings of traditional single BCI systems, a novel method of brain-computer interfacing (BCI) has been proposed; this method is known as a hybrid BCI. Although several tests using hybrid BCIs have yielded promising results, much more research and development is still needed in this area. As a corollary, we also did a thorough analysis of the most recent studies focusing on the usefulness of BCIs. To determine the weight each factor should have in the overall evaluation of BCI usability, we zeroed in on the tasks and measurements involved. We reported on the satisfaction-based usability features of BCI and mixed BCI systems efficacy, and efficiency, and made some suggestions for further research. Usability testing can reveal problems with the HCI and ergonomics of BCI and hybrid BCI systems, as well as provide suggestions for future study in the area.
    Keywords: brain-computer interfacing; BCI; drone control; 3D images.
    DOI: 10.1504/IJMEI.2023.10058455
     
  • Integration of matching smart phone data with long-term continuous glucose monitoring data pictures of food intake   Order a copy of this article
    by A. Rajasekar, J. Megala, Swagata Sarkar, C. Sivakumaran 
    Abstract: The treatment of diabetes relies heavily on careful glucose monitoring as an important component. Nevertheless, further information on the context is required in order to completely comprehend and interpret glucose levels in a meaningful manner. Continuous glucose monitoring (CGM), automated analysis of new glucose variables, and visualisation of CGM data via the ambulatory glucose profile have all contributed to significant advancements in precise glucose monitoring over the past five years. This made significant improvements in glucose monitoring. There are considerable challenges to its adoption, such as the fact that it is inconvenient and does not give immediate and frequent feedback. However, intermittent self-monitored blood glucose (SMBG) can offer extra information that may be used to make choices about therapy. Systems that indicate greater results compared to standard SMBG alone include those that give instant feedback to patients as well as decision support tools for both patients and physicians.
    Keywords: glucose monitoring; biomedical; smartphone; diabetes.
    DOI: 10.1504/IJMEI.2023.10058522
     
  • Revolutionising organ donation and transplantation for a better future: a blockchain based approach   Order a copy of this article
    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
     
  • Anomaly detection architecture for smart hospitals based on machine learning, time series, and image recognition analysis: survey   Order a copy of this article
    by Somaya Haiba, Tomader Mazri 
    Abstract: Smart hospital networks are considered the most sensitive networks for anomalies; any tiny existence might produce very different dangerous scales. The usual anomaly detections dedicated to this kind of network are not able to analyse all the different categories and proprieties of the generated data, because the majority of them rely only on time series analysis which is not able to cover all the circulated pieces of information. For that, in this paper, we will survey a proposed anomaly detection architecture that can dominate all the data categories that exist inside the e-health network using image recognition as well as time-series analysis.
    Keywords: E-healthcare monitoring network; IoMT; smart hospitals; E-health anomaly; anomaly detection; machine learning; time-series analysis; IoT security; ImageGray analysis; medical data; Cybersecurity.
    DOI: 10.1504/IJMEI.2023.10058832
     
  • A novel deep learning approach for b-value optimisation in intravoxel incoherent motion magnetic resonance imaging on simulated data   Order a copy of this article
    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
     
  • Tri-partition-based b-value optimisation for intravoxel incoherent motion magnetic resonance imaging of brain   Order a copy of this article
    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
     
  • Multi-task deep neural network models for learning COVID-19 disease representations from multimodal data   Order a copy of this article
    by Veena Mayya, K. Karthik, Krishnananda Prabhu Karadka, S. Sowmya Kamath 
    Abstract: Over the continued course of the COVID-19 pandemic, a significant volume of expert-written diagnosis reports has been accumulated that capture a multitude of symptoms and observations on diagnosed COVID-19 cases, along with expert-validated chest X-ray scans. The utility of rich, latent information embedded in such unstructured expert-written diagnosis reports and its importance as a source of valuable disease-specific information has been explored to a very limited extent. In this work, a convolutional attention-based dense (CAD) neural model for COVID-19 prediction is proposed. The model is trained on the rich disease-specific parameters extracted from chest X-ray images and expert-written diagnostic text reports to support an evidence-based diagnosis. Scalability is ensured by incorporating content based learning models for automatically generating diagnosis reports of identified COVID-19 cases, reducing radiologists' cognitive burden. Experimental evaluation showed that multimodal patient data plays a vital role in diagnosing early-stage cases, thus helping hasten the diagnosis process.
    Keywords: COVID-19 diagnosis; clinical decision support systems; multi-task learning; healthcare analytics; deep convolution neural networks.
    DOI: 10.1504/IJMEI.2021.10043617
     
  • Analysis on effect of resampling techniques on cardiac arrhythmia classification using convolutional neural networks   Order a copy of this article
    by Rekha Rajagopal, V. Shyam Kumar 
    Abstract: Cardiac arrhythmia is a condition in which the heart beats at a faster rate or a slower rate instead of a regular rhythm. Medical professionals identify the category of cardiac arrhythmia by manually viewing the electrocardiogram (ECG) signal which is more time-consuming. There are possibilities of incorrectly identifying the arrhythmia categories due to practical difficulties in manually assessing the slight variations in the amplitude of ECG signals. This research work focuses on automatically categorising the heartbeats as normal beat, supraventricular beat, ventricular beat, fusion beat, and unknown beat using convolutional neural network. The class imbalance problem that arises because of a few numbers of heartbeats in certain arrhythmia categories is resolved using techniques such as synthetic minority oversampling technique (SMOTE), borderline SMOTE, SVM-SMOTE, and adaptive synthetic sampling (ADASYN). The proposed model demonstrates an average accuracy of 97.76% in classifying arrhythmia classes using ADASYN technique. This model can help medical professionals in accurately diagnosing the arrhythmia classes.
    Keywords: cardiovascular disease; electrocardiogram; ECG; arrhythmia; convolutional neural network; CNN; class imbalance.
    DOI: 10.1504/IJMEI.2022.10044232
     
  • Developing a biotechnical scheme using fuzzy logic model for classification of severity of pyelonephritis   Order a copy of this article
    by Nikolay Korenevskiy, Seregin Stanislav Petrovich, Riad Taha Al-Kasasbeh, Ayman Ahmad Alqaralleh, Sofia Nikolaevna Rodionova, Ashraf Adel Shaqadan, Ilyash Maxim Yurievich, Mahdi Salman Alshamasin 
    Abstract: The aim of the work is to develop fuzzy logic model to process health data involving oxidative indicators in patients with pyelonephritis to predict the severity level of pyelonephritis as severe and purulent forms. A 13 immunity and oxidative health indicators (lipid peroxidation) are used for classification of disease severity. A control sample of patient's is analysed to develop class's and experienced physicians are consulted to modify considered class limits. The fuzzy logic model gives high accuracy in diagnosis of serous and purulent pyelonephritis in patients with urolithiasis. Verification of the results of the operation of the obtained decision rules on the control sample showed that the proposed method's diagnostic efficiency reaches 93%, which is acceptable for use in medical practice.
    Keywords: pyelonephritis; serious and purulent form; fuzzy logic; mathematical models; differential diagnosis; oxidative status.
    DOI: 10.1504/IJMEI.2021.10041822
     
  • Identification and classification of schizophrenic speech using convolutional neural network for medical healthcare   Order a copy of this article
    by Akshita Abrol, Nisha Kapoor, Parveen Kumar Lehana 
    Abstract: Schizophrenia is a brain disorder that significantly affects the quality of life of affected individuals. One of its prominent symptoms is the induction of changes in the acoustics of the patients. In the absence of definite methods for its diagnosis, speech analysis can help in the preliminary screening of the patients. In this paper, an automated method using deep learning for differentiating between individuals with schizophrenia and psychosis from healthy individuals is suggested. Using convolutional neural networks with speech spectrograms as input, a classification accuracy of 87.01% has been obtained for levels of schizophrenia and 95.26% for differentiating between schizophrenic and healthy speech.
    Keywords: schizophrenia; convolutional neural network; CNN; deep learning; spectrograms.
    DOI: 10.1504/IJMEI.2021.10041905
     
  • Efficient breast cancer detection using novel intensity partitioning-based clustering algorithm and multi-dimensional LSTM cyclic neural network   Order a copy of this article
    by Gul Shaira Banu Jahangeer, T. Dhiliphan Rajkumar 
    Abstract: Recently, early detection of breast cancer is significant to reduce the mortality rate, especially in women. Hence, the study aims to classify breast cancer from digital database for screening mammography (DDSM) dataset using partition and intensity based segmentation algorithm and modified convolutional neural network-long short-term memory (CNN-LSTM) classifier. Initially, pre-processing is performed using Gaussian filtering by taking the mammogram image. Then, it is segmented using a novel intensity partitioning-based clustering algorithm (IPCA). Further, feature extraction is performed and finally, classification is implemented using a novel multi-dimensional LSTM cyclic neural network (MLSTM-CNN). The analysis is performed to evaluate the efficiency of the proposed system and the outcomes explored its efficacy in breast cancer detection.
    Keywords: intensity partitioning; segmentation; long short-term memory; LSTM; cyclic neural network; CNN; breast cancer classification; digital database for screening mammography; DDSM; intensity partitioning-based clustering algorithm; IPCA.
    DOI: 10.1504/IJMEI.2021.10043507
     
  • Comparing the performance of machine learning techniques for low back pain diagnosis   Order a copy of this article
    by Hamid Bouraghi, Sorayya Rezayi, Soheila Saeedi, Rasoul Salimi, Meysam Jahani, Sajjad Abdolmaleki 
    Abstract: Low back pain is a global health problem that is a major cause of disability in developing and developed countries. Machine learning and data mining algorithms can be used to help diagnose this disease. This study aimed to determine the performance of different machine learning algorithms. Nine machine learning techniques, including support vector machine, decision tree, Naive Bayes, K-nearest neighbours, neural network, random forest, deep learning, auto-MLP, and rule induction, were used to modelling. This study revealed that the highest accuracy was related to the random forest (83.55%) and support vector machine (82.26%) classifiers. As a result, machine learning algorithms have good accuracy in low back pain diagnosis.
    Keywords: machine learning; data mining; low back pain; LBP; diagnose.
    DOI: 10.1504/IJMEI.2021.10043054
     
  • Detecting obstructive sleep apnea by extracting multimodal HRV features using ensemble subspace discriminant classifier   Order a copy of this article
    by Nivedita Singh, R.H. Talwekar 
    Abstract: Obstructive sleep apnea disorder is very peculiar sleep disorder which is triggered because of rapid and repeated transition of breathing. Hypopnea is also known as partial blockage of respiration during sleep. Polysomnography is gold standard to detect OSA but it is very expensive and complex which motivates us to detect OSA through multimodal heart rate variability (HRV) feature analysis using single channel ECG. The comparison among three classifiers SVM, weighted KNN and ensemble subspace discriminant are investigated for OSA detection. The accuracy obtained by the ESD classifier is 100%. True positive rate (TPR) and the true negative rate (TNR) have been attained 100% which is best suitable classifier for our experiment.
    Keywords: obstructive sleep apnea; ensemble subspace discriminant; heart rate variability; multimodal features.
    DOI: 10.1504/IJMEI.2022.10044042
     
  • Global sensitivity analysis implications on the design of PF-04475270 model predictive control system   Order a copy of this article
    by Omer Hamid 
    Abstract: In this study, we adopted Pharmacokinetic-Pharmacodynamic (PKPD) of Pfizer ophthalmic drug 5-{3-[(2S)-2-{(3R)-3-hydroxy-4-[3-(trifluoromethyl) phenyl] butyl}-5-oxopyrrolidin-1-yl] propyl} thiophene-2-carboxylate (PF-04475270) from literature and performed a series of global sensitivity analysis (GSA) to identify the most influential PKPD parameters of PF-04475270. We used the sensitivity analysis for everybody (SAFE) toolbox GSA with two different well-established GSA methods. Our GSA analyses successfully identified the aqueous chamber transfer rates as high-ranking parameters in the comparison with all the other 11 different model parameters. Our data with prioritised parameters suggest the need for probing the aqueous drug concentration for future design of model predictive control (MPC) release system of PF-04475270.
    Keywords: global sensitivity analysis; GSA; model parameters uncertainties; predictive control; Iris-Ciliary body; pharmacokinetics; pharmacodynamics; sensitivity analysis for everybody; SAFE; model predictive control; MPC.
    DOI: 10.1504/IJMEI.2021.10043868