| Alzheimer Detection Using CNN Models Based on Anatomical Feature Extraction from MRI Images |
| Paper ID : 1033-ICEEM2025 (R1) |
| Authors |
|
Alaa Emad Moataz1, Mariam Ibrahim Ibrahim1, Mohamed N. Saad2, Tamer M. Nassef *1 1Faculty of Computer Science, October Modern University for Sciences and Arts (MSA) 2Biomedical Engineering Department, Faculty of Engineering, Minia University, Egypt |
| Abstract |
| Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that requires early and accurate diagnosis to enable timely intervention. Magnetic resonance imaging (MRI) provides structural biomarkers for AD, yet many automated detection approaches rely solely on end-to-end deep learning models, often overlooking domain-specific anatomical features. This paper presents a hybrid framework that integrates anatomical feature extraction with convolutional neural networks (CNNs) for AD classification. MRI images are preprocessed and segmented using K-Means clustering to estimate gray matter (GM) and white matter (WM) percentages, which are then concatenated with image features in a custom-designed CNN optimized for grayscale medical imaging. The proposed model is evaluated against three pretrained architectures—ResNet50, VGG16, and DenseNet121—using the publicly available Alzheimer Dataset by Dr. Saeed Mohsen. Experimental results show that the custom CNN achieves 97.54% accuracy, substantially outperforming pretrained models (49.89–61.30%). The findings demonstrate that integrating clinically relevant handcrafted features can enhance model performance, interpretability, and efficiency, addressing critical limitations in conventional deep learning pipelines for medical image analysis. |
| Keywords |
| Alzheimer’s disease, Magnetic resonance imaging (MRI), Convolutional neural networks (CNNs), Deep learning, Feature extraction, Gray matter, White matter, Transfer learning, Medical image analysis, Pretrained models. |
| Status: Accepted |