Early Alzheimer's Detection with MRI-Based Deep Convolutional Neural Networks and Transfer Learning

Alzheimer's disease (AD) is a progressive neurodegenerative disorder that lacks a definitive cure and imposes a growing burden on healthcare systems. Early and accurate detection is essential for managing the disease and improving patient outcomes. In this study, we investigate the effectiveness of two pre-trained convolutional neural networks, Inception V۳ and ResNet۵۰, for classifying AD stages using T۱-weighted MRI scans. Leveraging transfer learning, both models were fine-tuned on an augmented dataset consisting of four classes: no dementia, very mild, mild, and moderate dementia. Comprehensive preprocessing techniques were employed to enhance image quality and reduce noise. The results demonstrate that both Inception V۳ and ResNet۵۰ achieve high accuracy in multi-class classification, highlighting their potential in assisting early AD diagnosis. While Inception V۳ showed strong feature extraction capabilities, ResNet۵۰ offered a balance between performance and computational efficiency. These findings suggest that CNN-based models can provide scalable, reliable tools for clinical decision support in neurodegenerative disease diagnostics.

کلیدواژه ها:

Alzheimer's disease ، transfer learning ، MRI ، convolutional neural networks