Mosabber Uddin Ahmed
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Mosabber Uddin Ahmed has a total of 112 co-authors. Here are the top co-authors by number of shared publications:
- Muhammad Enamul Hoque Chowdhury
Department of Electrical Engineering, Qatar University, Doha P.O. Box 2713, Qatar
9 shared publications Ahasan Atick Faisal
Qatar Univ, Dept Elect Engn, Doha 2713, Qatar
5 shared publicationsAtiqur Rahman Ahad
University of Dhaka, Dhaka, Bangladesh
4 shared publicationsDanilo P. Mandic
4 shared publications- Kawser Ahammed
Jatiya Kabi kazi Nazrul Islam University
4 shared publications - Ahasan Atick Faisal
Department of Electrical Engineering Qatar University 2713 Doha Qatar
3 shared publications
Latest Publications
- Open Access
- 17 September 2026
- Published by Elsevier in Process Safety and Environmental Protection
- Vol. 220
Physics-informed multi-task neural network for simultaneous leak detection, diameter estimation, and localization in multiphase pipelines
- 17 September 2026
- Published by Elsevier in Process Safety and Environmental Protection
- Vol. 220
Open AccessPhysics-informed multi-task neural network for simultaneous leak detection, diameter estimation, and localization in multiphase pipelines
- Open Access
- 28 January 2026
- Published by Springer Nature in International Journal of Data Science and Analytics
- Vol. 22  (1)
A novel deep learning approach with feature injection technique for automated gait detection in parkinson’s disease using multi-modal data
Parkinson’s disease (PD) is a persistent, irreversible neurodegenerative condition, with freezing of gait (FOG) being among its most incapacitating symptoms. PD patients with FOG encounter abrupt and unforeseen difficulties in initiating or maintaining movement, potentially resulting in unexpected falls and injuries. A sensor-driven FOG detection system can facilitate continuous and objective tracking of PD patients experiencing FOG and provide on-the-spot cueing support. Many recent studies have employed advanced deep learning methods to detect FOG by analyzing data from only inertial measurement unit (IMU) sensors. In our study, we have utilized a multimodal dataset that includes a gait accelerometer (ACC), electromyogram (EMG), and electroencephalogram (EEG) for FOG detection. For robust FOG detection task, we have proposed a novel deep learning-based model called Self-FOGNet, which integrates the self-organized operational neuron (Self-ONN) layer to enhance the learning of intricate patterns within the data by introducing non-homogeneity in neural networks. The robust model also promotes feature reuse and efficiently utilizes the hierarchical feature space across multiple layers, enabling effective classification. Further, we extracted several wavelet features and injected them at the last layer of the model as time domain and frequency domain features. We have used 3 different models for 3 data modalities and used an ensemble learning approach to combine the outputs of each model to get the final improved prediction. As a result of our proposed method, we achieved 98.74% accuracy and 98.68% specificity in the FOG detection task with an ensemble learning approach that outperformed the state-of-the-art approaches.
- 28 January 2026
- Published by Springer Nature in International Journal of Data Science and Analytics
- Vol. 22  (1)
Open AccessA novel deep learning approach with feature injection technique for automated gait detection in parkinson’s disease using multi-modal data
Parkinson’s disease (PD) is a persistent, irreversible neurodegenerative condition, with freezing of gait (FOG) being among its most incapacitating symptoms. PD patients with FOG encounter abrupt and unforeseen difficulties in initiating or maintaining movement, potentially resulting in unexpected falls and injuries. A sensor-driven FOG detection system can facilitate continuous and objective tracking of PD patients experiencing FOG and provide on-the-spot cueing support. Many recent studies have employed advanced deep learning methods to detect FOG by analyzing data from only inertial measurement unit (IMU) sensors. In our study, we have utilized a multimodal dataset that includes a gait accelerometer (ACC), electromyogram (EMG), and electroencephalogram (EEG) for FOG detection. For robust FOG detection task, we have proposed a novel deep learning-based model called Self-FOGNet, which integrates the self-organized operational neuron (Self-ONN) layer to enhance the learning of intricate patterns within the data by introducing non-homogeneity in neural networks. The robust model also promotes feature reuse and efficiently utilizes the hierarchical feature space across multiple layers, enabling effective classification. Further, we extracted several wavelet features and injected them at the last layer of the model as time domain and frequency domain features. We have used 3 different models for 3 data modalities and used an ensemble learning approach to combine the outputs of each model to get the final improved prediction. As a result of our proposed method, we achieved 98.74% accuracy and 98.68% specificity in the FOG detection task with an ensemble learning approach that outperformed the state-of-the-art approaches.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 169141-169160
Insights Into MRI Images: Brain Tumor Classification With CNN and PCA-Based Pattern Recognition
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 169141-169160
Open AccessInsights Into MRI Images: Brain Tumor Classification With CNN and PCA-Based Pattern Recognition
- 19 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1148-1153
Comparing Deep Learning Frameworks for Air Quality Index (AQI) Prediction in Bangladesh: Static vs. Year-Wise Forecasting Evaluation
Air quality has significant impacts on human health, ecosystems, and wildlife, making its monitoring and forecasting a critical concern. Bangladesh frequently experiences high levels of air pollution that exceed safe thresholds. This paper presents a harmonized Air Quality Index (AQI) dataset for Bangladesh (2012-2024), covering eight cities and ten monitoring stations, and a systematic comparison of static and forecasting deep learning (DL) models for AQI prediction. Building on this dataset, two complementary modeling frameworks were developed: (1) a forecasting pipeline employing residual learning with temporal lags, and (2) a static pipeline utilizing concurrent pollutant and station data. DL architectures including Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Temporal Convolutional Network (TCN), Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), and Multilayer Perceptron (MLP) were systematically evaluated under static and forecasting scenarios, with ensemble methods applied to enhance prediction accuracy. Stacking ensembles consistently outperformed individual models, achieving mean absolute error (MAE) of approximately 2.33, mean squared error (MSE) of 40.56, and coefficient of determination ($R^{2}$) of 0.98 in static evaluation, and improving from MAE of 4.1 to 1.99 in forecasting with extended temporal windows. CNN and TCN demonstrated strong performance for longer horizons, whereas recurrent models and MLP were less effective. These findings can inform more accurate AQI forecasting and support evidence-based air quality management in Bangladesh.
- 19 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1148-1153
Comparing Deep Learning Frameworks for Air Quality Index (AQI) Prediction in Bangladesh: Static vs. Year-Wise Forecasting Evaluation
Air quality has significant impacts on human health, ecosystems, and wildlife, making its monitoring and forecasting a critical concern. Bangladesh frequently experiences high levels of air pollution that exceed safe thresholds. This paper presents a harmonized Air Quality Index (AQI) dataset for Bangladesh (2012-2024), covering eight cities and ten monitoring stations, and a systematic comparison of static and forecasting deep learning (DL) models for AQI prediction. Building on this dataset, two complementary modeling frameworks were developed: (1) a forecasting pipeline employing residual learning with temporal lags, and (2) a static pipeline utilizing concurrent pollutant and station data. DL architectures including Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Temporal Convolutional Network (TCN), Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), and Multilayer Perceptron (MLP) were systematically evaluated under static and forecasting scenarios, with ensemble methods applied to enhance prediction accuracy. Stacking ensembles consistently outperformed individual models, achieving mean absolute error (MAE) of approximately 2.33, mean squared error (MSE) of 40.56, and coefficient of determination ($R^{2}$) of 0.98 in static evaluation, and improving from MAE of 4.1 to 1.99 in forecasting with extended temporal windows. CNN and TCN demonstrated strong performance for longer horizons, whereas recurrent models and MLP were less effective. These findings can inform more accurate AQI forecasting and support evidence-based air quality management in Bangladesh.
- Open Access
- 31 August 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 112
Self-DenseMobileNet: A robust framework for lung nodule classification using Self-ONN and stacking-based meta-classifier
- 31 August 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 112
Open AccessSelf-DenseMobileNet: A robust framework for lung nodule classification using Self-ONN and stacking-based meta-classifier
- 12 March 2025
- Published by Springer Nature in Natural Hazards
- Vol. 121  (9) , 10195-10224
Flood severity classification in Bangladesh: a comprehensive analysis of historical weather and water level data using machine learning approaches
- 12 March 2025
- Published by Springer Nature in Natural Hazards
- Vol. 121  (9) , 10195-10224
Flood severity classification in Bangladesh: a comprehensive analysis of historical weather and water level data using machine learning approaches
- 8 November 2024
- Published by Springer Nature
- p. 189-208
A Multibranch LSTM-CNN Model for Human Activity Recognition
- 8 November 2024
- Published by Springer Nature
- p. 189-208
A Multibranch LSTM-CNN Model for Human Activity Recognition
- Open Access
Self-DenseMobileNet: A Robust Framework for Lung Nodule Classification using Self-ONN and Stacking-based Meta-Classifier
In this study, we propose a novel and robust framework, Self-DenseMobileNet, designed to enhance the classification of nodules and non-nodules in chest radiographs (CXRs). Our approach integrates advanced image standardization and enhancement techniques to optimize the input quality, thereby improving classification accuracy. To enhance predictive accuracy and leverage the strengths of multiple models, the prediction probabilities from Self-DenseMobileNet were transformed into tabular data and used to train eight classical machine learning (ML) models; the top three performers were then combined via a stacking algorithm, creating a robust meta-classifier that integrates their collective insights for superior classification performance. To enhance the interpretability of our results, we employed class activation mapping (CAM) to visualize the decision-making process of the best-performing model. Our proposed framework demonstrated remarkable performance on internal validation data, achieving an accuracy of 99.28\% using a Meta-Random Forest Classifier. When tested on an external dataset, the framework maintained strong generalizability with an accuracy of 89.40\%. These results highlight a significant improvement in the classification of CXRs with lung nodules.
Open AccessSelf-DenseMobileNet: A Robust Framework for Lung Nodule Classification using Self-ONN and Stacking-based Meta-Classifier
In this study, we propose a novel and robust framework, Self-DenseMobileNet, designed to enhance the classification of nodules and non-nodules in chest radiographs (CXRs). Our approach integrates advanced image standardization and enhancement techniques to optimize the input quality, thereby improving classification accuracy. To enhance predictive accuracy and leverage the strengths of multiple models, the prediction probabilities from Self-DenseMobileNet were transformed into tabular data and used to train eight classical machine learning (ML) models; the top three performers were then combined via a stacking algorithm, creating a robust meta-classifier that integrates their collective insights for superior classification performance. To enhance the interpretability of our results, we employed class activation mapping (CAM) to visualize the decision-making process of the best-performing model. Our proposed framework demonstrated remarkable performance on internal validation data, achieving an accuracy of 99.28\% using a Meta-Random Forest Classifier. When tested on an external dataset, the framework maintained strong generalizability with an accuracy of 89.40\%. These results highlight a significant improvement in the classification of CXRs with lung nodules.
- 8 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Heterogeneous Ensemble Approach in Intrusion Detection using Stacking Technique
The proliferating cutting-edge technologies have given rise to massive data generation and numerous prospects of cyber-attacks on transgressors. The myriad attack landscapes in the cyber world jeopardize confidential data ranging from personal to industry level. To defend massive computer systems and preserve not only the integrity but also the confidentiality of digital data, the vitalness of intrusion detection systems (IDS) is irrefutable. The enormous and heterogeneous network traffic has rendered intrusion detection mechanisms based upon traditional machine learning techniques inadequate. Rather, a hybrid or ensemble of multiple classifiers can be a felicitous solution in elevating system performance by classifying network intrusions precisely. This study proposes a heterogeneous ensemble scheme for intrusion detection based on the stacking technique that incorporated three base estimators such as a k nearest neighbor (KNN), feedforward neural network (FNN), decision tree (DT), and one meta-classifier i.e. random forest (RF). The benchmark dataset NSL-KDD, UNSW-NB15 dataset, and CSE-CIC-IDS2018 dataset were used in the experiment and comprehensive empir-ical analyses were conducted for binary as well as multiclass classifications on various subsets of the datasets to calibrate the stacking ensemble method based on several evaluation metrics. The proposed approach manifested significant improvement in the detection rate of various network instances in multiclass classification while minimizing the false positive rate.
- 8 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Heterogeneous Ensemble Approach in Intrusion Detection using Stacking Technique
The proliferating cutting-edge technologies have given rise to massive data generation and numerous prospects of cyber-attacks on transgressors. The myriad attack landscapes in the cyber world jeopardize confidential data ranging from personal to industry level. To defend massive computer systems and preserve not only the integrity but also the confidentiality of digital data, the vitalness of intrusion detection systems (IDS) is irrefutable. The enormous and heterogeneous network traffic has rendered intrusion detection mechanisms based upon traditional machine learning techniques inadequate. Rather, a hybrid or ensemble of multiple classifiers can be a felicitous solution in elevating system performance by classifying network intrusions precisely. This study proposes a heterogeneous ensemble scheme for intrusion detection based on the stacking technique that incorporated three base estimators such as a k nearest neighbor (KNN), feedforward neural network (FNN), decision tree (DT), and one meta-classifier i.e. random forest (RF). The benchmark dataset NSL-KDD, UNSW-NB15 dataset, and CSE-CIC-IDS2018 dataset were used in the experiment and comprehensive empir-ical analyses were conducted for binary as well as multiclass classifications on various subsets of the datasets to calibrate the stacking ensemble method based on several evaluation metrics. The proposed approach manifested significant improvement in the detection rate of various network instances in multiclass classification while minimizing the false positive rate.