Isaac Acquah
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Isaac Acquah has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Benjamin Appiah Yeboah
Kwame Nkrumah University of Science and Technology
6 shared publications - Mawusi Gbemavor-Assonhe
Kwame Nkrumah University of Science and Technology
3 shared publications - Michael Asiedu Asare
Kwame Nkrumah University of Science and Technology
3 shared publications - Emmanuel Owusu
Biomedical Engineering Program, Department of Computer Engineering, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana
2 shared publications - Prince Ebenezer Adjei
Department of Computer Engineering Kwame Nkrumah University of Science and Technology Kumasi Ghana
2 shared publications - Albert Dede
Kwame Nkrumah University of Science and Technology
1 shared publication
Latest Publications
- Open Access
- 10 August 2026
- Published by SAGE Publications in Biomedical Engineering and Computational Biology
- Vol. 17
Attention-Enhanced Hybrid Bidirectional LSTM and Temporal Convolutional Network for Early Detection of Lung Cancer in Low-Dose CT Scans
Objectives: We developed and evaluated a lightweight, interpretable, and computationally efficient hybrid deep learning model for multiclass classification of early lung cancer in low-dose CT scans, potentially deployable in resource-limited healthcare environments. Methods: A novel hybrid architecture was developed that integrates Bidirectional Long Short-Term Memory (Bi-LSTM) networks, Temporal Convolutional Networks (TCNs), Efficient Channel Attention (ECA) blocks, and Local Interpretable Model-Agnostic Explanations (LIME). The model employed depthwise separable convolutions to reduce computational complexity. A multi-stream feature extraction framework was implemented to enhance interpretability and capture spatial-temporal patterns. The model was trained and validated on the IQ-OTH/NCCD dataset (version 2), containing 3,609 CT scan slices from 110 patients (1,097 original, 2,512 augmented), across three classes: normal, benign, and malignant. The dataset was split into training (60%), validation (30%), and testing (10%) subsets. Training was conducted using the AdamW optimizer for 16 epochs. Results: The model achieved 98.06% accuracy, 98.15% precision, 98.06% recall, and 98.04% F1-score, with a 99.88% AUC, a model size of 3.33 MB, and 279,561 parameters. Conclusion: The lightweight model achieves high diagnostic accuracy with computational efficiency, SHAP-based and LIME-based interpretability methods, enabling potential suitability for deployment in resource-constrained clinical settings.
- 10 August 2026
- Published by SAGE Publications in Biomedical Engineering and Computational Biology
- Vol. 17
Open AccessAttention-Enhanced Hybrid Bidirectional LSTM and Temporal Convolutional Network for Early Detection of Lung Cancer in Low-Dose CT Scans
Objectives: We developed and evaluated a lightweight, interpretable, and computationally efficient hybrid deep learning model for multiclass classification of early lung cancer in low-dose CT scans, potentially deployable in resource-limited healthcare environments. Methods: A novel hybrid architecture was developed that integrates Bidirectional Long Short-Term Memory (Bi-LSTM) networks, Temporal Convolutional Networks (TCNs), Efficient Channel Attention (ECA) blocks, and Local Interpretable Model-Agnostic Explanations (LIME). The model employed depthwise separable convolutions to reduce computational complexity. A multi-stream feature extraction framework was implemented to enhance interpretability and capture spatial-temporal patterns. The model was trained and validated on the IQ-OTH/NCCD dataset (version 2), containing 3,609 CT scan slices from 110 patients (1,097 original, 2,512 augmented), across three classes: normal, benign, and malignant. The dataset was split into training (60%), validation (30%), and testing (10%) subsets. Training was conducted using the AdamW optimizer for 16 epochs. Results: The model achieved 98.06% accuracy, 98.15% precision, 98.06% recall, and 98.04% F1-score, with a 99.88% AUC, a model size of 3.33 MB, and 279,561 parameters. Conclusion: The lightweight model achieves high diagnostic accuracy with computational efficiency, SHAP-based and LIME-based interpretability methods, enabling potential suitability for deployment in resource-constrained clinical settings.
- Open Access
- 1 April 2026
- Published by SAGE Publications in Health Informatics Journal
- Vol. 32 (2)
Evaluation of medical device calibration knowledge, perceived importance, and implementation support among healthcare professionals in Ghana
Objectives: Medical device calibration underpins accuracy, precision, and reliability, which are critical to optimal device performance and quality healthcare delivery. Despite its importance, little empirical evidence exists on calibration knowledge among healthcare professionals in sub-Saharan Africa. This study assessed such knowledge in Ghana by examining familiarity with calibration principles, procedures, device-specific practices, institutional support, and factors influencing competence. Methods: A cross-sectional survey of 425 healthcare workers (technicians/clinical engineers, nurses, midwives, and doctors) employed a structured questionnaire. Statistical analyses ( t -tests, ANOVA, and correlation analysis) were conducted to identify differences and associations in calibration knowledge and practices. Results: Analyses revealed significant knowledge gaps. Technicians/clinical engineers outperformed other groups, while early-career professionals (1–3 years’ experience) scored higher than mid-career counterparts (7–10 years), highlighting shortcomings in continuous professional development. Formal training was also associated with higher scores. Conclusion: These findings highlight the urgent need for structured, role-specific training and ongoing professional education to close calibration knowledge gaps, safeguard patient safety, and optimise device performance. The results provide evidence for policy reforms and workforce development strategies to strengthen medical device management and healthcare quality.
- 1 April 2026
- Published by SAGE Publications in Health Informatics Journal
- Vol. 32 (2)
Open AccessEvaluation of medical device calibration knowledge, perceived importance, and implementation support among healthcare professionals in Ghana
Objectives: Medical device calibration underpins accuracy, precision, and reliability, which are critical to optimal device performance and quality healthcare delivery. Despite its importance, little empirical evidence exists on calibration knowledge among healthcare professionals in sub-Saharan Africa. This study assessed such knowledge in Ghana by examining familiarity with calibration principles, procedures, device-specific practices, institutional support, and factors influencing competence. Methods: A cross-sectional survey of 425 healthcare workers (technicians/clinical engineers, nurses, midwives, and doctors) employed a structured questionnaire. Statistical analyses ( t -tests, ANOVA, and correlation analysis) were conducted to identify differences and associations in calibration knowledge and practices. Results: Analyses revealed significant knowledge gaps. Technicians/clinical engineers outperformed other groups, while early-career professionals (1–3 years’ experience) scored higher than mid-career counterparts (7–10 years), highlighting shortcomings in continuous professional development. Formal training was also associated with higher scores. Conclusion: These findings highlight the urgent need for structured, role-specific training and ongoing professional education to close calibration knowledge gaps, safeguard patient safety, and optimise device performance. The results provide evidence for policy reforms and workforce development strategies to strengthen medical device management and healthcare quality.
- Open Access
- 23 February 2026
- Published by SAGE Publications in Cancer Informatics
- Vol. 25
Resilient Sinkhorn-Based Optimal Transport Late Fusion Framework for Breast Cancer Diagnosis
Objective: This research aims to develop and evaluate a clinically deployable multimodal deep learning framework for breast cancer diagnosis that maintains robustness, even when clinical data are asynchronous, unpaired, or incomplete, effectively addressing real-world challenges related to data heterogeneity and fragmented clinical workflows. Methods: In this retrospective study, a multimodal deep learning architecture was developed that integrates histopathological images with structured clinical risk factors. Custom models were developed and independently trained for each modality, and late fusion was achieved via a dynamically reweighted Sinkhorn-based fusion layer. Model performance was evaluated using precision–recall Area Under Curve (PR-AUC), recall, F 1 score, and Brier score under complete and partial modality availability scenarios. Robustness and clinical utility were further assessed through statistical significance testing and decision curve analysis (DCA). Additionally, we employed a Sinkhorn cost matrix to enhance interpretability. Results: The proposed Sinkhorn fusion model outperformed all baseline methods, achieving the highest recall (0.96), PR-AUC (0.775), F 1 score (0.828), and the best calibration (Brier score ≈ 0.19). Notably, it maintained perfect recall (1.00) under a 50% simulated modality dropout, despite a significant drop in PR-AUC (20% vs 0%: t = –20.35, P < .0001; 50% vs 0%: t = 88.60, P < .0001), portraying a strong overall robustness to information missingness. Under internally controlled conditions, DCA demonstrated superior clinical utility across thresholds of 0.2 to 0.7. Conclusions: The model’s ability to accommodate unpaired and incomplete clinical inputs while maintaining both calibration and sensitivity makes it particularly well-suited for deployment in asynchronous and resource-constrained settings. Its consistent performance under clinical uncertainty and minimal preprocessing requirements represents a significant advancement toward equitable, reliable, and scalable AI-assisted breast cancer screening. To our knowledge, this is the first paper to model breast cancer late fusion as an optimal transport problem.
- 23 February 2026
- Published by SAGE Publications in Cancer Informatics
- Vol. 25
Open AccessResilient Sinkhorn-Based Optimal Transport Late Fusion Framework for Breast Cancer Diagnosis
Objective: This research aims to develop and evaluate a clinically deployable multimodal deep learning framework for breast cancer diagnosis that maintains robustness, even when clinical data are asynchronous, unpaired, or incomplete, effectively addressing real-world challenges related to data heterogeneity and fragmented clinical workflows. Methods: In this retrospective study, a multimodal deep learning architecture was developed that integrates histopathological images with structured clinical risk factors. Custom models were developed and independently trained for each modality, and late fusion was achieved via a dynamically reweighted Sinkhorn-based fusion layer. Model performance was evaluated using precision–recall Area Under Curve (PR-AUC), recall, F 1 score, and Brier score under complete and partial modality availability scenarios. Robustness and clinical utility were further assessed through statistical significance testing and decision curve analysis (DCA). Additionally, we employed a Sinkhorn cost matrix to enhance interpretability. Results: The proposed Sinkhorn fusion model outperformed all baseline methods, achieving the highest recall (0.96), PR-AUC (0.775), F 1 score (0.828), and the best calibration (Brier score ≈ 0.19). Notably, it maintained perfect recall (1.00) under a 50% simulated modality dropout, despite a significant drop in PR-AUC (20% vs 0%: t = –20.35, P < .0001; 50% vs 0%: t = 88.60, P < .0001), portraying a strong overall robustness to information missingness. Under internally controlled conditions, DCA demonstrated superior clinical utility across thresholds of 0.2 to 0.7. Conclusions: The model’s ability to accommodate unpaired and incomplete clinical inputs while maintaining both calibration and sensitivity makes it particularly well-suited for deployment in asynchronous and resource-constrained settings. Its consistent performance under clinical uncertainty and minimal preprocessing requirements represents a significant advancement toward equitable, reliable, and scalable AI-assisted breast cancer screening. To our knowledge, this is the first paper to model breast cancer late fusion as an optimal transport problem.
- Open Access
- 7 February 2026
- Published by Oxygen Alliance in Journal of Biomedical Engineering Systems and Technologies for LMICs
- Vol. 1 (1)
Early Cardiovascular Disease Detection: An Improved Pan-Tompkins Algorithm for QRS Detection in Electrocardiogram
Cardiovascular diseases (CVDs) are the leading cause of death globally, with low- and middle-income countries (LMICs) bearing the greatest burden. Electrocardiograms (ECGs), which reflect the heart's electrical activity, are an essential tool in diagnosing CVD. The QRS complex is the most prominent wave in an ECG signal and is used for evaluating the overall heart function of an individual. The Pan-Tompkins algorithm is widely used for the detection of QRS complexes. It is, however, susceptible to baseline wander noise, has decreased sensitivity for diverse ECG morphologies, and exhibits real-time delay, leading to its suboptimal performance in QRS complex detection. This study presents an improved PanTompkins approach that combines the strengths of the Pan-Tompkins algorithm with a Recurrent Neural Network (RNN) to deliver more accurate and efficient QRS detection. The proposed algorithm achieved precision, recall, and F1-scores above 96% on Lead II of the MIT-BIH Arrhythmia Database. Overall, false positive and false negative rates were below 0.05%, calculated across the selected segments from all patient records. Execution time was reduced by 4% relative to the original Pan–Tompkins algorithm on identical ECG segments, directly lowering latency and improving real-time performance. A band-pass filter of 6–16 Hz was used, which improved robustness against baseline wander, effectively reducing noise. The algorithm demonstrated enhanced resilience to morphological variability in ECG signals, ensuring more reliable detection across diverse patterns. By integrating this AI-driven algorithm into low-cost, portable ECG devices, there is strong potential to support early detection of CVDs, particularly in underserved areas. This work contributes to a practical, scalable solution that can help strengthen digital health infrastructure and improve clinical outcomes in LMICs.
- 7 February 2026
- Published by Oxygen Alliance in Journal of Biomedical Engineering Systems and Technologies for LMICs
- Vol. 1 (1)
Open AccessEarly Cardiovascular Disease Detection: An Improved Pan-Tompkins Algorithm for QRS Detection in Electrocardiogram
Cardiovascular diseases (CVDs) are the leading cause of death globally, with low- and middle-income countries (LMICs) bearing the greatest burden. Electrocardiograms (ECGs), which reflect the heart's electrical activity, are an essential tool in diagnosing CVD. The QRS complex is the most prominent wave in an ECG signal and is used for evaluating the overall heart function of an individual. The Pan-Tompkins algorithm is widely used for the detection of QRS complexes. It is, however, susceptible to baseline wander noise, has decreased sensitivity for diverse ECG morphologies, and exhibits real-time delay, leading to its suboptimal performance in QRS complex detection. This study presents an improved PanTompkins approach that combines the strengths of the Pan-Tompkins algorithm with a Recurrent Neural Network (RNN) to deliver more accurate and efficient QRS detection. The proposed algorithm achieved precision, recall, and F1-scores above 96% on Lead II of the MIT-BIH Arrhythmia Database. Overall, false positive and false negative rates were below 0.05%, calculated across the selected segments from all patient records. Execution time was reduced by 4% relative to the original Pan–Tompkins algorithm on identical ECG segments, directly lowering latency and improving real-time performance. A band-pass filter of 6–16 Hz was used, which improved robustness against baseline wander, effectively reducing noise. The algorithm demonstrated enhanced resilience to morphological variability in ECG signals, ensuring more reliable detection across diverse patterns. By integrating this AI-driven algorithm into low-cost, portable ECG devices, there is strong potential to support early detection of CVDs, particularly in underserved areas. This work contributes to a practical, scalable solution that can help strengthen digital health infrastructure and improve clinical outcomes in LMICs.
- Open Access
- 27 September 2025
- Published by SAGE Publications in Health Services Insights
- Vol. 18
Confidence in Medical Device Calibration Knowledge and Skills Among Healthcare Professionals in Ghana: A Cross-Sectional Analysis
Background: Calibration of medical devices is vital for diagnostic reliability, quality healthcare delivery and patient safety. While awareness is critical, the confidence of healthcare professionals (HCPs) in calibration-related knowledge significantly influences adherence to best practices in calibration-related tasks. However, this area is underexplored in sub-Saharan Africa. Objectives: This study aimed to assess the confidence levels of HCPs in Ghana regarding their calibration-related knowledge and competencies. It examined how these vary by professional role, training status, and years of experience. Methods: A cross-sectional survey was conducted among 461 HCPs in Ghana, including doctors, nurses, midwives and technicians/clinical engineers, using a structured questionnaire covering five thematic areas: general understanding of calibration knowledge, knowledge of calibration procedures, device-specific knowledge, training and institutional support, and the perceived impact on patient care. Results: Statistical analyses revealed variations in sentiment scores across profession, experience levels and training status. Technicians/clinical engineers consistently reported higher confidence, while doctors, nurses and midwives scored significantly lower. Additionally, respondents with prior calibration training, regardless of experience level, showed substantially greater confidence than their untrained counterparts. Conclusion: These findings demonstrate the need for targeted, hands-on calibration training and continuous professional development to improve confidence and competencies in calibration practices. Such initiatives are essential for effective device management and overall healthcare system performance. This study offers evidence to guide policy improvements and capacity-building efforts to strengthen calibration knowledge and practices in Ghana’s healthcare sector.
- 27 September 2025
- Published by SAGE Publications in Health Services Insights
- Vol. 18
Open AccessConfidence in Medical Device Calibration Knowledge and Skills Among Healthcare Professionals in Ghana: A Cross-Sectional Analysis
Background: Calibration of medical devices is vital for diagnostic reliability, quality healthcare delivery and patient safety. While awareness is critical, the confidence of healthcare professionals (HCPs) in calibration-related knowledge significantly influences adherence to best practices in calibration-related tasks. However, this area is underexplored in sub-Saharan Africa. Objectives: This study aimed to assess the confidence levels of HCPs in Ghana regarding their calibration-related knowledge and competencies. It examined how these vary by professional role, training status, and years of experience. Methods: A cross-sectional survey was conducted among 461 HCPs in Ghana, including doctors, nurses, midwives and technicians/clinical engineers, using a structured questionnaire covering five thematic areas: general understanding of calibration knowledge, knowledge of calibration procedures, device-specific knowledge, training and institutional support, and the perceived impact on patient care. Results: Statistical analyses revealed variations in sentiment scores across profession, experience levels and training status. Technicians/clinical engineers consistently reported higher confidence, while doctors, nurses and midwives scored significantly lower. Additionally, respondents with prior calibration training, regardless of experience level, showed substantially greater confidence than their untrained counterparts. Conclusion: These findings demonstrate the need for targeted, hands-on calibration training and continuous professional development to improve confidence and competencies in calibration practices. Such initiatives are essential for effective device management and overall healthcare system performance. This study offers evidence to guide policy improvements and capacity-building efforts to strengthen calibration knowledge and practices in Ghana’s healthcare sector.
- 27 May 2025
- Published by IOP Publishing in Biomedical Physics & Engineering Express
- Vol. 11 (4) , 042001
Rational design of alternative natural-based coupling media for diagnostic ultrasound imaging: a review
Ultrasound imaging is an indispensable diagnostic and screening tool in healthcare, renowned for its non-invasive nature, real-time visualization, and use of non-ionizing radiation. It plays a vital role in obstetrics and gynaecology by significantly reducing maternal mortality and enhancing patient care. In addition to its use in obstetrics, ultrasound is used to guide biopsies and to evaluate various diseases such as liver cirrhosis, thyroid disorders, and kidney stones. Point-of-care ultrasonography has proven to be increasingly beneficial in low-resource settings. However, the availability and cost of commercial ultrasound gels pose significant challenges. Alternative natural-based gels formulated from locally sourced materials have emerged as viable substitutes. This review critically examines alternative natural-based ultrasound gels, focusing on their physicochemical properties, formulation procedures, and the limitations associated with their use in diagnostic imaging. Furthermore, it presents a rational design approach that methodically selects the ingredients based on their properties and interactions to formulate these gels for imaging applications. This offers a promising pathway for indigenous manufacturers to develop gels that meet ideal performance criteria, ensuring better imaging outcomes and wider acceptability in clinical practice.
- 27 May 2025
- Published by IOP Publishing in Biomedical Physics & Engineering Express
- Vol. 11 (4) , 042001
Rational design of alternative natural-based coupling media for diagnostic ultrasound imaging: a review
Ultrasound imaging is an indispensable diagnostic and screening tool in healthcare, renowned for its non-invasive nature, real-time visualization, and use of non-ionizing radiation. It plays a vital role in obstetrics and gynaecology by significantly reducing maternal mortality and enhancing patient care. In addition to its use in obstetrics, ultrasound is used to guide biopsies and to evaluate various diseases such as liver cirrhosis, thyroid disorders, and kidney stones. Point-of-care ultrasonography has proven to be increasingly beneficial in low-resource settings. However, the availability and cost of commercial ultrasound gels pose significant challenges. Alternative natural-based gels formulated from locally sourced materials have emerged as viable substitutes. This review critically examines alternative natural-based ultrasound gels, focusing on their physicochemical properties, formulation procedures, and the limitations associated with their use in diagnostic imaging. Furthermore, it presents a rational design approach that methodically selects the ingredients based on their properties and interactions to formulate these gels for imaging applications. This offers a promising pathway for indigenous manufacturers to develop gels that meet ideal performance criteria, ensuring better imaging outcomes and wider acceptability in clinical practice.
- Open Access
- 1 May 2025
- Published by SAGE Publications in Digital Health
- Vol. 11
LiteFallNet: A lightweight deep learning model for efficient real-time fall detection
Objective: This study introduces LiteFallNet, a lightweight and interpretable deep learning model for real-time fall detection using only inertial sensor data. It aims to overcome key limitations in current systems, including high computational demands, latency, and privacy concerns, while delivering accurate and reliable performance. Methods: LiteFallNet integrates a Gated Recurrent Unit (GRU) layer, a Temporal Convolutional Network (TCN) block, depthwise separable convolutions, and a Squeeze-and-Excitation (SE) block to efficiently extract temporal features from tri-axial accelerometer, gyroscope, and magnetometer signals. The model was trained and evaluated on the FallAllD and the UMAFall datasets. To enhance transparency, one-dimensional gradient-weighted class activation mapping (1D Grad-CAM) and local interpretable model-agnostic explanations (LIME) were used to interpret how the model made its predictions. Results: The model on the FallAllD dataset achieved an accuracy of 97.81%, a recall of 98.55%, and an F1-score of 97.88%, with an area under the receiver operating characteristic curve of 99.33%. With a size of just 0.312 MB and an inference time of 7.07 ms, LiteFallNet combines strong performance with efficiency. These attributes make it highly suitable for deployment in real-time, resource-constrained environments. Conclusion: LiteFallNet offers a privacy-preserving and real-time solution for fall detection. Its accuracy, transparency, and lightweight design make it suitable for smart homes, eldercare facilities, and wearable health technologies.
- 1 May 2025
- Published by SAGE Publications in Digital Health
- Vol. 11
Open AccessLiteFallNet: A lightweight deep learning model for efficient real-time fall detection
Objective: This study introduces LiteFallNet, a lightweight and interpretable deep learning model for real-time fall detection using only inertial sensor data. It aims to overcome key limitations in current systems, including high computational demands, latency, and privacy concerns, while delivering accurate and reliable performance. Methods: LiteFallNet integrates a Gated Recurrent Unit (GRU) layer, a Temporal Convolutional Network (TCN) block, depthwise separable convolutions, and a Squeeze-and-Excitation (SE) block to efficiently extract temporal features from tri-axial accelerometer, gyroscope, and magnetometer signals. The model was trained and evaluated on the FallAllD and the UMAFall datasets. To enhance transparency, one-dimensional gradient-weighted class activation mapping (1D Grad-CAM) and local interpretable model-agnostic explanations (LIME) were used to interpret how the model made its predictions. Results: The model on the FallAllD dataset achieved an accuracy of 97.81%, a recall of 98.55%, and an F1-score of 97.88%, with an area under the receiver operating characteristic curve of 99.33%. With a size of just 0.312 MB and an inference time of 7.07 ms, LiteFallNet combines strong performance with efficiency. These attributes make it highly suitable for deployment in real-time, resource-constrained environments. Conclusion: LiteFallNet offers a privacy-preserving and real-time solution for fall detection. Its accuracy, transparency, and lightweight design make it suitable for smart homes, eldercare facilities, and wearable health technologies.
- Open Access
- 18 February 2025
- Published by Wiley in Engineering Reports
- Vol. 7 (2)
Wavelet‐Based Feature Extraction for Efficient High‐Resolution Image Classification
Convolutional neural networks (CNNs) typically compress high‐resolution images to minimize computational requirements. However, this can lead to loss of information and reduced accuracy in classification tasks. This paper introduces WaveNet, a novel approach for processing high‐resolution images using wavelet‐domain inputs in CNNs. We address the challenge of maintaining classification accuracy with high‐resolution inputs while minimizing computational complexity. Our method employs wavelet packet transform (WPT) for image pre‐processing, extracting detailed multi‐scale and directional information from high‐resolution images. We propose a wavelet‐adaptive efficient channel attention (WAECA) module to dynamically select the most informative wavelet subbands. Popular CNN architectures like ResNet‐50 and MobileNetV2 are adapted by replacing their initial convolutional layers with wavelet‐transformed inputs, enabling direct learning in the wavelet domain. Experiments conducted on the Caltech‐256 and ALOT datasets demonstrate that WaveNet improves classification accuracy while reducing computational complexity. For instance, our wavelet‐enhanced ResNet‐50 achieves a Top‐1 accuracy of 72.47% on Caltech‐256, outperforming the baseline (70.65%) while reducing FLOPs from 16.52G to 3.98G. Similar improvements are observed across different architectures and datasets. We also evaluate various wavelet filters and ResNet backbones, finding that the bior1.1 filter and ResNet‐50 provide optimal performance. This work presents a practical solution for developing more accurate and efficient models for high‐resolution inputs without extensive computational resources or complex architectural modifications.
- 18 February 2025
- Published by Wiley in Engineering Reports
- Vol. 7 (2)
Open AccessWavelet‐Based Feature Extraction for Efficient High‐Resolution Image Classification
Convolutional neural networks (CNNs) typically compress high‐resolution images to minimize computational requirements. However, this can lead to loss of information and reduced accuracy in classification tasks. This paper introduces WaveNet, a novel approach for processing high‐resolution images using wavelet‐domain inputs in CNNs. We address the challenge of maintaining classification accuracy with high‐resolution inputs while minimizing computational complexity. Our method employs wavelet packet transform (WPT) for image pre‐processing, extracting detailed multi‐scale and directional information from high‐resolution images. We propose a wavelet‐adaptive efficient channel attention (WAECA) module to dynamically select the most informative wavelet subbands. Popular CNN architectures like ResNet‐50 and MobileNetV2 are adapted by replacing their initial convolutional layers with wavelet‐transformed inputs, enabling direct learning in the wavelet domain. Experiments conducted on the Caltech‐256 and ALOT datasets demonstrate that WaveNet improves classification accuracy while reducing computational complexity. For instance, our wavelet‐enhanced ResNet‐50 achieves a Top‐1 accuracy of 72.47% on Caltech‐256, outperforming the baseline (70.65%) while reducing FLOPs from 16.52G to 3.98G. Similar improvements are observed across different architectures and datasets. We also evaluate various wavelet filters and ResNet backbones, finding that the bior1.1 filter and ResNet‐50 provide optimal performance. This work presents a practical solution for developing more accurate and efficient models for high‐resolution inputs without extensive computational resources or complex architectural modifications.
- Open Access
- 1 January 2025
- Published by SAGE Publications in Science Progress
- Vol. 108 (1)
A hybrid long short-term memory-convolutional neural network multi-stream deep learning model with Convolutional Block Attention Module incorporated for monkeypox detection
Background: Monkeypox (mpox) is a zoonotic infectious disease caused by the mpox virus and characterized by painful body lesions, fever, headaches, and exhaustion. Since the report of the first human case of mpox in Africa, there have been multiple outbreaks, even in nonendemic regions of the world. The emergence and re-emergence of mpox highlight the critical need for early detection, which has spurred research into applying deep learning to improve diagnostic capabilities. Objective: This research aims to develop a robust hybrid long short-term memory (LSTM)-convolutional neural network (CNN) model with a Convolutional Block Attention Module (CBAM) to provide a potential tool for the early detection of mpox. Methods: A hybrid LSTM-CNN multi-stream deep learning model with CBAM was developed and trained using the Mpox Skin Lesion Dataset Version 2.0 (MSLD v2.0). We employed LSTM layers for preliminary feature extraction, CNN layers for further feature extraction, and CBAM for feature conditioning. The model was evaluated with standard metrics, and gradient-weighted class activation maps (Grad-CAM) and local interpretable model-agnostic explanations (LIME) were used for interpretability. Results: The model achieved an F1-score, recall, and precision of 94%, an area under the curve of 95.04%, and an accuracy of 94%, demonstrating competitive performance compared to the state-of-the-art models. This robust performance highlights the reliability of our model. LIME and Grad-CAM offered insights into the model's decision-making process. Conclusion: The hybrid LSTM-CNN multi-stream deep learning model with CBAM successfully detects mpox, providing a promising early detection tool that can be integrated into web and mobile platforms for convenient and widespread use.
- 1 January 2025
- Published by SAGE Publications in Science Progress
- Vol. 108 (1)
Open AccessA hybrid long short-term memory-convolutional neural network multi-stream deep learning model with Convolutional Block Attention Module incorporated for monkeypox detection
Background: Monkeypox (mpox) is a zoonotic infectious disease caused by the mpox virus and characterized by painful body lesions, fever, headaches, and exhaustion. Since the report of the first human case of mpox in Africa, there have been multiple outbreaks, even in nonendemic regions of the world. The emergence and re-emergence of mpox highlight the critical need for early detection, which has spurred research into applying deep learning to improve diagnostic capabilities. Objective: This research aims to develop a robust hybrid long short-term memory (LSTM)-convolutional neural network (CNN) model with a Convolutional Block Attention Module (CBAM) to provide a potential tool for the early detection of mpox. Methods: A hybrid LSTM-CNN multi-stream deep learning model with CBAM was developed and trained using the Mpox Skin Lesion Dataset Version 2.0 (MSLD v2.0). We employed LSTM layers for preliminary feature extraction, CNN layers for further feature extraction, and CBAM for feature conditioning. The model was evaluated with standard metrics, and gradient-weighted class activation maps (Grad-CAM) and local interpretable model-agnostic explanations (LIME) were used for interpretability. Results: The model achieved an F1-score, recall, and precision of 94%, an area under the curve of 95.04%, and an accuracy of 94%, demonstrating competitive performance compared to the state-of-the-art models. This robust performance highlights the reliability of our model. LIME and Grad-CAM offered insights into the model's decision-making process. Conclusion: The hybrid LSTM-CNN multi-stream deep learning model with CBAM successfully detects mpox, providing a promising early detection tool that can be integrated into web and mobile platforms for convenient and widespread use.