Ahmed Bouatmane
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Ahmed Bouatmane has a total of 5 co-authors. Here are the top co-authors by number of shared publications:
Abdelaziz Daaif
IACS Laboratory, ENSET Mohammedia, Hassan II University of Casablanca, Mohammedia, Morocco
2 shared publications- Abdelmajid Bousselham
IACS Laboratory, ENSET Mohammedia, Hassan II University of Casablanca
2 shared publications Omar Bouattane
EEIS Laboratory, ENSET Mohammedia, Hassan II University of Casablanca, Mohammedia, Morocco
2 shared publications- Bouchra Bouihi
Laboratory 2IACS, ENSET Mohammedia, Hassan II University of Casablanca, Morocco
1 shared publication Mohamed Youssfi
IACS Laboratory, ENSET Mohammedia, Hassan II University of Casablanca, Mohammedia, Morocco
1 shared publication
Latest Publications
- Open Access
- 19 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 204337-204354
A Hybrid Diffusion-VAE for High-Fidelity Tissue Doppler Imaging Augmentation in Cardiotoxicity Detection
Early detection of chemotherapy-induced cardiotoxicity remains a critical challenge, hindered by the scarcity of large and balanced datasets combining Tissue Doppler Imaging (TDI) and clinical information. This study introduces a conditional Diffusion-VAE hybrid that generates high-fidelity synthetic TDI images conditioned on clinical vectors, enabling robust data augmentation for cardiotoxicity prediction. A skip-connected Convolutional VAE (ConvVAE) first encodes each 2D TDI scan into a structured 256 × 10 × 24 latent tensor. After freezing the encoder, a cross-attention–conditioned U-Net iteratively denoises these latents under clinical guidance (27 features). Compared to the ConvVAE alone, the proposed Diffusion-VAE achieves a 55% reduction in MSE (0.0022 to 0.0010), a 20% improvement in FID (10.2 to 8.7), and gains in SSIM (0.96 to 0.98), PSNR (32.0 dB to 34.5 dB), and LPIPS (0.083 to 0.058), validating its superior fidelity and diversity. Bootstrap confidence intervals (1,000 resamples) confirmed statistical robustness, while a permutation test (10,000 shuffles) demonstrated the significance of AUC and PR-AUC improvements (p < 0.01). Grad-CAM and SHAP visualizations highlight that the model attends to physiologically meaningful Doppler envelope regions and influential clinical variables such as heart rate and ejection fraction. Classifiers retrained solely on synthetic data achieved 97.9% accuracy, 0.991 ROC-AUC, and 0.982 PR-AUC, underscoring the utility of the proposed model for overcoming data scarcity. This hybrid architecture, combining VAE’s structured representation with diffusion’s refinement capabilities, offers a scalable multimodal augmentation engine to advance reliable AI in cardiovascular oncology.
- 19 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 204337-204354
Open AccessA Hybrid Diffusion-VAE for High-Fidelity Tissue Doppler Imaging Augmentation in Cardiotoxicity Detection
Early detection of chemotherapy-induced cardiotoxicity remains a critical challenge, hindered by the scarcity of large and balanced datasets combining Tissue Doppler Imaging (TDI) and clinical information. This study introduces a conditional Diffusion-VAE hybrid that generates high-fidelity synthetic TDI images conditioned on clinical vectors, enabling robust data augmentation for cardiotoxicity prediction. A skip-connected Convolutional VAE (ConvVAE) first encodes each 2D TDI scan into a structured 256 × 10 × 24 latent tensor. After freezing the encoder, a cross-attention–conditioned U-Net iteratively denoises these latents under clinical guidance (27 features). Compared to the ConvVAE alone, the proposed Diffusion-VAE achieves a 55% reduction in MSE (0.0022 to 0.0010), a 20% improvement in FID (10.2 to 8.7), and gains in SSIM (0.96 to 0.98), PSNR (32.0 dB to 34.5 dB), and LPIPS (0.083 to 0.058), validating its superior fidelity and diversity. Bootstrap confidence intervals (1,000 resamples) confirmed statistical robustness, while a permutation test (10,000 shuffles) demonstrated the significance of AUC and PR-AUC improvements (p < 0.01). Grad-CAM and SHAP visualizations highlight that the model attends to physiologically meaningful Doppler envelope regions and influential clinical variables such as heart rate and ejection fraction. Classifiers retrained solely on synthetic data achieved 97.9% accuracy, 0.991 ROC-AUC, and 0.982 PR-AUC, underscoring the utility of the proposed model for overcoming data scarcity. This hybrid architecture, combining VAE’s structured representation with diffusion’s refinement capabilities, offers a scalable multimodal augmentation engine to advance reliable AI in cardiovascular oncology.
- Open Access
- 1 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 57568-57588
A Multimodal Deep Learning Model Integrating CNN and Transformer for Predicting Chemotherapy-Induced Cardiotoxicity
Chemotherapy-induced cardiotoxicity presents a major risk to cancer patients, often leading to severe cardiac complications such as heart failure, myocardial infarction, and arrhythmias. Early detection is crucial for preventing long-term damage and improving patient outcomes, yet existing diagnostic methods struggle to effectively capture the complexity of multimodal medical data and often lack interpretability. In this study, we propose an innovative approach that integrates multimodal deep learning with Explainable AI (XAI) techniques to enhance early cardiotoxicity detection. Our model combines clinical data (e.g., age and cardiovascular metrics) with Tissue Doppler Imaging (TDI), a functional imaging technique that captures myocardial velocity during the cardiac cycle. To overcome data limitations, we employed Conditional Generative Adversarial Networks (cGANs) and Conditional Tabular Generative Adversarial Networks (CTGANs) to augment the dataset, improving its diversity and balance for better model training. We developed three architectures that integrate Convolutional Neural Networks (CNNs) for feature extraction from TDI images with Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Transformer models to capture temporal dependencies and enhance prediction accuracy. Additionally, we incorporated SHapley Additive Explanations (SHAP) to interpret the contribution of input features, increasing model transparency and clinical applicability. Our Transformer-based model achieved the highest accuracy of 96%, outperforming the GRU (94%) and LSTM (89%) models, significantly surpassing traditional approaches. These findings highlight the potential of transformer-based architectures in multimodal deep learning for precise cardiotoxicity prediction, supporting early intervention and personalized treatment strategies while improving interpretability through XAI techniques such as SHAP.
- 1 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 57568-57588
Open AccessA Multimodal Deep Learning Model Integrating CNN and Transformer for Predicting Chemotherapy-Induced Cardiotoxicity
Chemotherapy-induced cardiotoxicity presents a major risk to cancer patients, often leading to severe cardiac complications such as heart failure, myocardial infarction, and arrhythmias. Early detection is crucial for preventing long-term damage and improving patient outcomes, yet existing diagnostic methods struggle to effectively capture the complexity of multimodal medical data and often lack interpretability. In this study, we propose an innovative approach that integrates multimodal deep learning with Explainable AI (XAI) techniques to enhance early cardiotoxicity detection. Our model combines clinical data (e.g., age and cardiovascular metrics) with Tissue Doppler Imaging (TDI), a functional imaging technique that captures myocardial velocity during the cardiac cycle. To overcome data limitations, we employed Conditional Generative Adversarial Networks (cGANs) and Conditional Tabular Generative Adversarial Networks (CTGANs) to augment the dataset, improving its diversity and balance for better model training. We developed three architectures that integrate Convolutional Neural Networks (CNNs) for feature extraction from TDI images with Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Transformer models to capture temporal dependencies and enhance prediction accuracy. Additionally, we incorporated SHapley Additive Explanations (SHAP) to interpret the contribution of input features, increasing model transparency and clinical applicability. Our Transformer-based model achieved the highest accuracy of 96%, outperforming the GRU (94%) and LSTM (89%) models, significantly surpassing traditional approaches. These findings highlight the potential of transformer-based architectures in multimodal deep learning for precise cardiotoxicity prediction, supporting early intervention and personalized treatment strategies while improving interpretability through XAI techniques such as SHAP.