International Journal of Biomedical Engineering and Technology (13 papers in press)

Regular Issues

• A Comprehensive Review on Magnetic Tissue Scaffold for Hyperthermia Treatment  
  by Debashish Gogoi, Tanyu Donarld Kongnyui, Manjesh Kumar 
  Abstract: This review explores the use of magnetic bone tissue scaffolds in hyperthermia treatment. It is a therapy that heats cancer cells to damage or destroy them while minimising harm to healthy tissues. Hyperthermia leverages the greater heat sensitivity of cancer cells, potentially enhancing treatment outcomes. Bone scaffolds, typically composed of biocompatible materials like ceramics or polymers, have emerged as promising tools for hyperthermia by incorporating magnetic nanoparticles that generate heat under an alternating magnetic field. This study aims to evaluate the current advancements in magnetic bone scaffolds for hyperthermia therapy, focusing on the materials, fabrication methods, and magnetic properties that influence their performance. The review also addresses key challenges in optimising scaffold design and offers recommendations for future research to improve therapeutic efficacy. Conclusions indicate that magnetic scaffolds have significant potential for targeted cancer treatment and bone regeneration, yet further studies are needed to enhance their clinical application. This review can guide future efforts toward optimising scaffold-based hyperthermia therapies.
  Keywords: Hyperthermia treatment; additive manufacturing; cancer; tumour; magnetic properties.
  DOI: 10.1504/IJBET.2024.10068450
   
  
• Finger Gesture Recognition by Mediapipe Algorithm and Advanced YOLOV7 Network for Deaf People  
  by Thanh-Hai Nguyen, Ba-Viet Ngo, Thanh-Long Nguyen, Chi-Cuong Vu 
  Abstract: Sign language is a challenge to be able to recognise and understand correctly its meaning. Therefore, with the deaf community, understanding sign language for daily activities is essential. For the support of the deaf community in interfacing together, researchers have represented different methods to support them in using sign languages. This article proposes using simple landmark hands with different finger shapes to detect ten letters in the American sign language (ASL) writing system. Therefore, an inference MediaPipe algorithm for extracting hand features with the advanced you only look once version 7 (YOLOv7) network was applied for sign language recognition. The article used 2000 hand images of 5 persons of different ages and genders with the YOLOv7 network to produce the high recognition performance. In particular, the average mAP accuracy reached 0.995, accuracy reached 99.4%, and recall parameter reached 99.2% using the confusion matrix. This algorithm can be developed with more letters.
  Keywords: gesture recognition; YOLOv7 network; inference – MediaPipe algorithm; hand features; sign language.
  DOI: 10.1504/IJBET.2024.10068659
   
  
• An Integrated Data-Driven Analysis-Based Deep Learning Framework for Early Autism Detection in Children to Improve Diagnostic Performance  
  by Jahanara Shaik, R. Shekhar, Chetan Shelke 
  Abstract: Autism spectrum disorder (ASD) children must be recognised early to obtain prompt care, promote development, and reduce long-term issues. This research provides a VGG16 and ResNet50-based data-driven deep learning system for early ASD screening using facial picture data. The study meticulously normalises, augments, and selects features using chi-square methods to ensure high-quality inputs and low dataset variability. Hyperparameter adjustment optimises model performance and five-fold cross-validation provides robust evaluation. VGG16 can recognise complex face characteristics with 87% accuracy for autistic classifications due to its precision and recall measures. Bio-inspired optimisation improves classification, helping ResNet50 outperform training epochs. Despite these advances, multimodal inputs are still needed for complete analysis due to the limits of facial data and the diversity of datasets. Deep learning models with feature selection can improve diagnostic precision, reduce false positives, and enable clinical real-time ASD screening. The proposed framework speeds diagnosis and is adaptable to varied healthcare circumstances. Future studies will focus on behavioural and genetic data, expandable artificial intelligence (XAI) for interpretability, and larger datasets for robustness. A scalable and effective ASD diagnosis using AI shows the transformative potential of AI in healthcare.
  Keywords: Autism Spectrum Disorder; Deep Learning; VGG16; ResNet50; Early Diagnosis; Explainable AI (XAI); Early Autism Detection.
  DOI: 10.1504/IJBET.2024.10068985
   
  
• Major Mandible Reconstruction: Design, Analysis, and Additive Manufacturing of Customised Implant and Surgical Osteotomy Guide  
  by Hari Narayan Singh, Yashwant Kumar Modi, A. Kuthe, Sanat Agrawal 
  Abstract: A patient-specific major mandible reconstruction has been presented in this article. During the process, a customized surgical osteotomy resection guide was designed to avoid overcutting and undercutting of the infected part of diseased mandible. The reconstructed bone models and customized mandible implant were 3D printed to test form and fit of the implant. Finite element analysis was used to simulate the distribution of von Mises stress and deformation in Ti6Al4V implant by subjecting it to an occlusal bite force of 300 N, replicating the forces experienced during biting. The analysis was performed for four different biting scenarios. The highest values for maximum von Mises stress and maximum deformation was observed when the biting force was applied at the incisor. Despite variations in stress and deformation, the Ti6Al4V implant was determined to be safe in all four biting scenarios.
  Keywords: Mandible reconstruction; FEM analysis; 3D printing; customized mandible implant; and additive manufacturing.
  DOI: 10.1504/IJBET.2024.10068993
   
  
• Dense Parallel Convolutional Neural Network for Lung and Colon Cancer Detection using Histopathological Images  
  by Mangore Anirudh K, Geeta S. Navale, Jawahar Sambhaji Gawade, Shailesh Pramod Bendale, Mubin Tamboli, Amol Dhumane 
  Abstract: Lung and colon cancer are deadly diseases, which cause mortality and morbidity. Here, the Dense Parallel Convolutional Neural Network (DPCNN) is proposed for detecting Lung and Colon Cancer by Histopathological Images (HI). Initially, the HI of the lung and colon are taken from the Lung and Colon Cancer Histopathological Images dataset and subjected to pre-processing. A Medav filter is used to pre-process the image. The Parallel Reverse Attention Network (PraNet) is then used to segregate the cancer region. Next, feature extraction is carried out, and features like Gray Level Run Length Matrix (GRLM) with Local Neighborhood Difference Pattern (LNDP) are extracted. After this, Lung and colon cancer detection is performed by the DPCNN approach. The DPCNN is a combination of DenseNet and Parallel Convolutional Neural Network (PCNN). The DPCNN obtained the True Negative Rate (TNR), accuracy, and True Positive Rate (TPR) of 92.044%, 92.773%, and 93.099%.
  Keywords: Dense Parallel Convolutional Neural Network; PraNet; Gray Level Run Length Matrix; Parallel Convolutional Neural Network Matrix; DenseNet.
  DOI: 10.1504/IJBET.2024.10069028
   
  
• Fusion of Multiple Time-Frequency Representation Techniques and Classifiers for ECG and PPG Signal Analysis  
  by Piyush Mahajan, Amit Kaul 
  Abstract: This study explores the fusion of multiple Time-Frequency Representation (TFR) techniques for analyzing Electrocardiogram (ECG) and Photoplethysmogram (PPG) signals. We examined 11 TFR methods for noise removal, characteristic point detection, and feature extraction. Multiple classifiers were trained to classify ECG signals into six arrhythmia classes and PPG signals for hypertension detection. The ensemble classifiers, particularly those combining Continuous Wavelet Transform (CWT), S-Transform, Wigner-Ville Distribution (WVD), and Synchrosqueezed Wavelet Transform (SSWT), achieved a testing accuracy of 96.6% for ECG signals. A combination of CWT and SSWT with the K-Nearest Neighbour (KNN) classifier achieved 81.48% accuracy on the PPG dataset. The ensemble approach using majority voting significantly enhanced classification performance, reaching 99.75% accuracy for ECG arrhythmia and 82.52% for PPG classification. This fusion of TFR techniques and ensemble classifiers demonstrates improved accuracy in signal classification tasks.
  Keywords: ECG;PPG;TFR;ML.
  DOI: 10.1504/IJBET.2024.10069147
   
  
• Hybrid Approach for Skin Lesion Analysis:- Integrating Modified U-Net Segmentation with Vision Transformers for Multi-Class Skin Cancer Detection  
  by Ramya J, Anil Kumar K.M 
  Abstract: Skin cancer is a major global health concern, where early and accurate detection is crucial for patient survival. Traditional CNN-based methods in skin lesion classification face challenges, particularly with the complexity of spatial and semantic features. To address these issues, we propose a unique hybrid deep learning approach integrating vision transformer (ViT) and a modified U-Net model. ViT processes images as tokens instead of pixels, enabling superior feature extraction and classification. Pre-processing techniques, including non-local means filtering for denoising and unsharp masking for contrast enhancement, are applied to enhance model robustness. Our hybrid approach integrates U-Net for precise segmentation, achieving metrics such as IOU of 92.46%, AUC of 97.64%, and dice coefficient of 95.96%. ViT for classification achieves exceptional accuracy, precision, recall, and F1 score, all at 99%. Using the HAM10000 dataset, our method surpasses existing techniques, demonstrating remarkable effectiveness in skin cancer detection and classification.
  Keywords: Skin Cancer Images; Multi-class classification; Vision Transformers; Region-of-Interest Segmentation; Computerized; Medical image processing.
  DOI: 10.1504/IJBET.2025.10069272
   
  
• AI-Driven Prediction of mRNA Vaccine Degradation Rates with Dropout-Enhanced Hyperparameter Optimisation  
  by Hwai Ing Soon, Azian Azamimi Abdullah, Hiromitsu Nishizaki, Mohd Yusoff Mashor, Latifah Munirah Kamarudin, Zeti-Azura Mohamed-Hussein, Zeehaida Mohamed, Wei Chern Ang 
  Abstract: Rapid mRNA vaccine degradation necessitates accurate prediction to ensure efficacy and mitigate risks. Despite challenges, mRNA vaccines affordability, high efficacy, and minimal side effects justify intensive research efforts. Multifaceted data analysis enhances prediction efficiency. This study leverages bioinformatics with label-encoding of tetra-nitrogenous bases (4-ntb-lbA) and advanced models, refined through extensive review. Hyperparameter optimisation is improved with the dropout-enhanced technique (DEet), addressing traditional shortcomings frequently suggesting suboptimal configurations. Results demonstrate that 4-ntb-lbA and DEet offer practical solutions. Specifically, 4-ntb-lbA upholds interdisciplinarity and minimises overfitting, while DEet mitigates suboptimality and accelerates convergence. These improvements are notable in the three-layered-wrapped-stacked BiLSTM (3lw-BiLSTM) model with 0.125-DEet to Bayesian optimisation with Gaussian process (BOGP) on a medium-sized subset (MSbT) at epoch 150, where training and validation losses reached 0.0015 and 0.0018, respectively, significantly reducing computational costs. This interdisciplinary approach is valuable for biotechnology and biomedicine, underscoring its contribution to efficient data analysis.
  Keywords: mRNA vaccines; degradation rates; nitrogenous bases; hyperparameter optimization; dropout; deep neural networks.
  DOI: 10.1504/IJBET.2024.10069478
   
  
• PSq-Net: Parallel SqueezeNet for Myocardial Infarction Detection using Echocardiography  
  by Shamal Bulbule, Shridevi Soma 
  Abstract: Early detection in Myocardial infarction is vital to prevent disability and fatalities through timely therapeutic interventions. However, current detection techniques are time-consuming. Hence, in this study, Parallel SqueezeNet (PSq-Net) is introduced for the myocardial infarction (MI) detection from echocardiography videos. Initially, the input video is obtained from the Hamad Medical Corporation, Tampere University, and Qatar University (HMC-QU) database. Then, pre-processing the video frames using median and Gaussian filters. Next, the left ventricular (LV) wall and endocardial boundary regions are segmented using Stack Attention U-Net (SAUN). Features are then extracted from the segmented endocardial boundary using various methods, including Local Optimal Oriented Patterns (LOOP), Local Binary Patterns (LBP), and statistical measures. Finally, myocardial infarction is detected from these features using PSq-Net, which is the integration of Parallel Convolutional Neural Network (PCNN) and SqueezeNet. The PSq-Net demonstrates a high sensitivity of 94.18%, accuracy of 93.09%, and specificity of 93.64%.
  Keywords: Stack attention U-Net; Parallel SqueezeNet; SqueezeNet; Parallel Convolutional Neural Network; median filter.
  DOI: 10.1504/IJBET.2025.10069612
   
  
• Alzheimer's Disease Recognition Based on Multimodal Image Fusion  
  by Xinjie Tao, Lisheng Wei, ShengBo Zhu 
  Abstract: To improve diagnostic accuracy for Alzheimer's Disease (AD) and enhance lesion feature extraction from single-modality imaging, a deep learning-based multimodal image fusion classification method is proposed. A novel residual network extracts features from three-dimensional images, serving as a feature extractor for both Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET). The extracted features are fused for classification, enhanced by a coordinate attention mechanism to capture spatial and channel relationships in 3D medical images. Experimental results show the fusion network achieved 91.07% accuracy in AD/Mild Cognitive Impairment (MCI)/Cognitively Normal (CN) classification, improving by 7.14% over the basic residual network, 21.43% over single-modality MRI, and 12.5% over single-modality PET. The fusion network also shows performance improvements in AD/CN, AD/MCI, and CN/MCI classification tasks, demonstrating its effectiveness.
  Keywords: Alzheimer's Disease; Multimodal feature fusion; Attention mechanism; Residual network.
  DOI: 10.1504/IJBET.2025.10069613
   
  
• Eye-blink Artifact Removal Framework for EEG Signals using DWT and Autoencoder  
  by Mohd. Faisal, Sudarsan Sahoo, Jupitara Hazarika 
  Abstract: This paper presents an automatic method for removing eye-blink artifacts from contaminated EEG signals, achieving high Signal to Artifact Ratio (SAR) and correlation coefficient (CC) with minimal computational time. In the proposed method, the autoencoder is trained to learn the underlying structure of artifacts within the contaminated frequency band of the EEG signal. Once trained, the autoencoder can effectively remove artifacts from contaminated signal by processing the contaminated signal as its input. Loss function plays important role in reconstructing the input signal. A novel loss function comprising correlation coefficient and mean square error is used to minimize the reconstruction error. The performance of the proposed method is measured by SAR and CC, achieving average values of 3.21 and 0.88, respectively, in removing artifacts from EEG signals. Also, the proposed method achieved minimum computational time in comparison to other methods which is 9 ms.
  Keywords: Brain computer Interface; electroencephalogram; eye-blink artifact; autoencoder.
  DOI: 10.1504/IJBET.2025.10070025
   
  
• The effect of Fatigue on Lower Limb Coordination Characteristics in Badminton forehand smash: a Functional Principal Component Analysis  
  by Zhou Zhanyi, Zixiang Gao, Shudong Li 
  Abstract: The study of motor coordination explores how the central nervous system controls body movements. Using functional principal component analysis (FPCA), this study examined the impact of fatigue on limb coordination synergies and activity coefficients in 23 badminton players performing forehand smashes before and after fatigue. Kinematic data revealed that pre-fatigue, three coordination synergies effectively controlled all lower limb joints, while post-fatigue, four synergies were required, reflecting increased control complexity. Fatigue significantly altered synergy control: Synergy 2’s control of the hip joint decreased, control of the right ankle increased (p < 0.05), and Synergy 3’s control of the left ankle decreased (p < 0.05). Fatigue also shifted synergy activity, enhancing Synergy 3 control during landing while reducing Synergy 1 control in the follow-through. These findings highlight how fatigue modifies coordination strategies, increasing joint control complexity, with implications for training to enhance performance stability and reduce injury risks.
  Keywords: motor control coordination; functional principal component; movement coordination synergy; badminton forehand smash.
  DOI: 10.1504/IJBET.2024.10070029
   
  
• Machine Learning based Hybrid Approach for Schizophrenia Diagnosis using Non-Stationary EEG Signals  
  by Harasees Kaur, Padmavati Khandnor, Kanu Goel 
  Abstract: A serious mental illness, Schizophrenia (SZ) affects 1% of people worldwide and is characterized by delusions, hallucinations and disorganised thought patterns. Diagnosis mostly is based on subjective interviews by a psychiatrist in which there is a high chance of human errors and bias. In this work, we have conducted a comprehensive analysis of Electroencephalogram (EEG) data using Empirical mode decomposition (EMD) algorithm which can analyse non-stationary and non-linear signals and separates them into components at different resolutions called Intrinsic mode functions (IMFs). In this work, our primary goal is to introduce a hybrid approach for IMF selection that combines four distinct parameters namely correlation, energy, statistical significance and power spectral density (PSD) distance. From the selected IMFs nine statistical features are computed and performance is evaluated using various classifiers. Among all the classifiers k-Nearest Neighbour (KNN) showed the best accuracy of 90.29% using the second IMF. These results suggest that EEG signals can effectively distinguish between Healthy Control and SZ patients and have a potential to help psychiatrists for diagnosis of SZ.
  Keywords: Electroencephalogram; Schizophrenia; Empirical Mode Decomposition; k-Nearest Neighbor; Machine learning.
  DOI: 10.1504/IJBET.2024.10070160