Alzheimers Disease Classification with a Hybrid CNN-SVM Approach on Enhanced MRI Data
Mostafijur Rahman, Md. Sabbir Hossain, Arifa Akter Eva, Md. Mohsin Kabir, M. F. Mridha, Jungpil Shin
Abstract
This research introduces a hybrid model combining Convolutional Neural Networks (CNN) with Support Vector Machines (SVM) for classifying Alzheimer's disease using Magnetic Resonance Imaging (MRI) scans. The dataset used consists of 6400 MRI images, resized to 128x128 pixels and divided into four categories: Mild-Demented, Moderate-Demented, Non-Demented, and Very Mild-Demented. The model is designed to enhance early detection and management of Alzheimer's disease by accurately distinguishing between these categories. Achieving an overall accuracy of 98.59 %, the model's detailed metrics include precision, recall, and F1-score across all classes, highlighting its high reliability. Specifically, precision scores for the classes Mild-Demented, Moderate-Demented, Non-Demented, and Very Mild-Demented were 0.98, 1.00, 0.99, and 0.98, respectively. Recall values were 0.98, 1.00, 0.98, and 0.99, with Fl-scores of 0.98, 1.00, 0.99, and 0.98, respectively. These results suggest the model's effectiveness for potential clinical use, supporting timely and accurate diagnosis of Alzheimer's disease. Future work will focus on further validating the model with larger and more diverse datasets, as well as exploring practical implementation scenarios.