ADERGHAL Karim

I'm a researcher in computer vision and medical image analysis at the University of Bordeaux currently working on deep learning applications for Alzheimer's disease diagnosis.

I previously did research working with Jenny Benois-Pineau, Gwenaëlle Catheline, and Karim Afdel, collaborating on multimodal neuroimaging analysis. My work focuses on developing CNN for the classification of structural MRI and diffusion tensor imaging data to improve early detection and staging of Alzheimer's disease and mild cognitive impairment. My research specifically explores the fusion of different magnetic resonance imaging modalities using deep learning techniques, with a particular emphasis on transfer learning and cross-modal approaches. I have developed novel 3D CNN architectures, including Inception-based networks, that can effectively process and combine structural MRI (sMRI) and mean diffusivity maps from diffusion tensor imaging (MD-DTI) to enhance diagnostic accuracy. Through my work on the ADNI dataset, I've demonstrated how 2D+ε projection methods and multi-modal fusion can significantly improve classification performance for distinguishing between normal controls, Alzheimer's patients, and subjects with mild cognitive impairment.
A key aspect of my research involves addressing the challenge of limited training data in medical imaging through innovative data augmentation techniques and cross-modal transfer learning strategies. I have shown that models pre-trained on structural MRI data can be effectively adapted to work with diffusion tensor imaging modalities, reducing overfitting and improving generalization.

My publications have contributed to advancing the field of computer-aided diagnosis for neurodegenerative diseases, with particular focus on making deep learning approaches more robust and clinically applicable for early Alzheimer's detection.