Mohammad Jahangir Alam
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Mohammad Jahangir Alam has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Mushfiqur Rahman
Department of Computer Science and Engineering, Daffodil International University, Dhaka, Bangladesh
3 shared publications - Alif Sheakh
Department of Computer Science and Engineering, Daffodil International University, Dhaka, Bangladesh
2 shared publications - Mst. Sazia Tahosin
Department of Computer Science and Engineering, Daffodil International University, Dhaka, Bangladesh
2 shared publications - Al Habib Islam
Daffodil International University
1 shared publication - Ananda Sutradhar
Department of Computer Science and Engineering, Daffodil International University, Dhaka, Bangladesh
1 shared publication - Anupam Kumar Bairagi
Khulna University
1 shared publication
Latest Publications
- Open Access
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 30 (2) , 1272-1285
FedVGM: Enhancing Federated Learning Performance on Multi-Dataset Medical Images With XAI
Advances in deep learning have transformed medical imaging, yet progress is hindered by data privacy regulations and fragmented datasets across institutions. To address these challenges, we propose FedVGM, a privacy-preserving federated learning framework for multi-modal medical image analysis. FedVGM integrates four imaging modalities, including brain MRI, breast ultrasound, chest X-ray, and lung CT, across 14 diagnostic classes without centralizing patient data. Using transfer learning and an ensemble of VGG16 and MobileNetV2, FedVGM achieves 97.7% $\pm$ 0.01 accuracy on the combined dataset and 91.9–99.1% across individual modalities. We evaluated three aggregation strategies and demonstrated median aggregation to be the most effective. To ensure clinical interpretability, we apply explainable AI techniques and validate results through performance metrics, statistical analysis, and k-fold cross-validation. FedVGM offers a robust, scalable solution for collaborative medical diagnostics, supporting clinical deployment while preserving data privacy.
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 30 (2) , 1272-1285
Open AccessFedVGM: Enhancing Federated Learning Performance on Multi-Dataset Medical Images With XAI
Advances in deep learning have transformed medical imaging, yet progress is hindered by data privacy regulations and fragmented datasets across institutions. To address these challenges, we propose FedVGM, a privacy-preserving federated learning framework for multi-modal medical image analysis. FedVGM integrates four imaging modalities, including brain MRI, breast ultrasound, chest X-ray, and lung CT, across 14 diagnostic classes without centralizing patient data. Using transfer learning and an ensemble of VGG16 and MobileNetV2, FedVGM achieves 97.7% $\pm$ 0.01 accuracy on the combined dataset and 91.9–99.1% across individual modalities. We evaluated three aggregation strategies and demonstrated median aggregation to be the most effective. To ensure clinical interpretability, we apply explainable AI techniques and validate results through performance metrics, statistical analysis, and k-fold cross-validation. FedVGM offers a robust, scalable solution for collaborative medical diagnostics, supporting clinical deployment while preserving data privacy.
- 28 January 2025
- Published by Springer Nature in Neural Computing and Applications
- Vol. 37 (14) , 8479-8507
Optimizing medical image analysis through MViTX on multiple datasets with explainable AI
- 28 January 2025
- Published by Springer Nature in Neural Computing and Applications
- Vol. 37 (14) , 8479-8507
Optimizing medical image analysis through MViTX on multiple datasets with explainable AI
- Open Access
- 1 November 2023
- Published by Institute of Advanced Engineering and Science in Indonesian Journal of Electrical Engineering and Computer Science
- Vol. 32 (2) , 1115-1122
An early warning system of heart failure mortality with combined machine learning methods
Heart failure (HF) is currently the leading cause of morbidity and mortality worldwide. Identifying the risk of mortality at the early stages is crucial to reducing the mortality rate. However, the traditional methods for exploring the signs of mortality are difficult and time-consuming. Whereas, machine learning (ML) methods are superior in reducing HF’s mortality rate by providing early warnings. This study presents a novel ML classifier called imperial boost-stacked (IBS) that can serve as an effective early warning system for predicting HF mortality. Initially, we performed an efficient data balancing technique named synthetic minority oversampling technique with edited nearest neighbors (SMOTE-ENN) to mitigate the imbalance problem. Next, two well-known feature selection techniques, the extra tree (ET) and information gain (IG), are applied to reduce the data dimensions and select the most significant features. Following that, the prepared feature sets are trained with our proposed IBS classifier. Simultaneously leveraging the advantages of Boosting, Stacking, and multiple robust methods, it significantly correlates with the intricate patterns of clinical data of HF patients. Finally, the robust outcomes of 92.75% accuracy over existing studies reveal that our proposed study can effectively warn the HF mortality at early stages and reduce the burden on the healthcare sector.
- 1 November 2023
- Published by Institute of Advanced Engineering and Science in Indonesian Journal of Electrical Engineering and Computer Science
- Vol. 32 (2) , 1115-1122
Open AccessAn early warning system of heart failure mortality with combined machine learning methods
Heart failure (HF) is currently the leading cause of morbidity and mortality worldwide. Identifying the risk of mortality at the early stages is crucial to reducing the mortality rate. However, the traditional methods for exploring the signs of mortality are difficult and time-consuming. Whereas, machine learning (ML) methods are superior in reducing HF’s mortality rate by providing early warnings. This study presents a novel ML classifier called imperial boost-stacked (IBS) that can serve as an effective early warning system for predicting HF mortality. Initially, we performed an efficient data balancing technique named synthetic minority oversampling technique with edited nearest neighbors (SMOTE-ENN) to mitigate the imbalance problem. Next, two well-known feature selection techniques, the extra tree (ET) and information gain (IG), are applied to reduce the data dimensions and select the most significant features. Following that, the prepared feature sets are trained with our proposed IBS classifier. Simultaneously leveraging the advantages of Boosting, Stacking, and multiple robust methods, it significantly correlates with the intricate patterns of clinical data of HF patients. Finally, the robust outcomes of 92.75% accuracy over existing studies reveal that our proposed study can effectively warn the HF mortality at early stages and reduce the burden on the healthcare sector.
- Open Access
- 1 July 2023
- Published by Institute of Advanced Engineering and Science in Indonesian Journal of Electrical Engineering and Computer Science
- Vol. 31 (1) , 321-328
Skin disease detection employing transfer learning approacha fine-tune visual geometry group-19
Your skin may become damaged by skin diseases and conditions. These illnesses can cause skin changes such as rashes, inflammation, itching, and other skin changes. While some skin conditions may run in families, others may result from a person’s way of life. Skin conditions may be treated with pills, creams, ointments, changes in diet, and lifestyle modifications. Deep learning algorithms for computer vision applications have advanced quickly thanks to a significant amount of data for training the model and advancements in evaluation of proposed that can provide stronger simplifications. Undesired skin disease regions are eliminated, quality is raised, and the disease is tinted by discarding artifacts, decrease noise, and improving the image. Three augmentation techniques have raised the quantity of skin disease images. The five transfer learning models and various convolutional neural network (CNN) architectures analyzed the augmentation dataset. Visual geometry group-19 (VGG-19) offers the highest level of accuracy. Following the segmentation of the dermoscopic images, the affected skin cells' features are extracted using a feature extraction technique. The retrieved features are stratified using a CNN classifier, that is focused in deep learning. The best outcomes were obtained using the hyper-tuned VGG-19, which had test and validation accuracy of 99.21% and 99.25%, including both.
- 1 July 2023
- Published by Institute of Advanced Engineering and Science in Indonesian Journal of Electrical Engineering and Computer Science
- Vol. 31 (1) , 321-328
Open AccessSkin disease detection employing transfer learning approacha fine-tune visual geometry group-19
Your skin may become damaged by skin diseases and conditions. These illnesses can cause skin changes such as rashes, inflammation, itching, and other skin changes. While some skin conditions may run in families, others may result from a person’s way of life. Skin conditions may be treated with pills, creams, ointments, changes in diet, and lifestyle modifications. Deep learning algorithms for computer vision applications have advanced quickly thanks to a significant amount of data for training the model and advancements in evaluation of proposed that can provide stronger simplifications. Undesired skin disease regions are eliminated, quality is raised, and the disease is tinted by discarding artifacts, decrease noise, and improving the image. Three augmentation techniques have raised the quantity of skin disease images. The five transfer learning models and various convolutional neural network (CNN) architectures analyzed the augmentation dataset. Visual geometry group-19 (VGG-19) offers the highest level of accuracy. Following the segmentation of the dermoscopic images, the affected skin cells' features are extracted using a feature extraction technique. The retrieved features are stratified using a CNN classifier, that is focused in deep learning. The best outcomes were obtained using the hyper-tuned VGG-19, which had test and validation accuracy of 99.21% and 99.25%, including both.
- 11 June 2023
- Published by Springer Nature
- p. 376-385
IoT and ML Based Approach for Highway Monitoring and Streetlamp Controlling
- 11 June 2023
- Published by Springer Nature
- p. 376-385
IoT and ML Based Approach for Highway Monitoring and Streetlamp Controlling
- 25 May 2023
- Published by Springer Nature
- p. 237-248
A Machine Learning-Based Traditional and Ensemble Technique for Predicting Breast Cancer
- 25 May 2023
- Published by Springer Nature
- p. 237-248
A Machine Learning-Based Traditional and Ensemble Technique for Predicting Breast Cancer