Severity-Level Classification of Alzheimer’s Disease from MRI Scans using Convolutional Neural Networks

Effective diagnosis and monitoring of Alzheimer’s relies increasingly on advanced imaging techniques such as Magnetic Resonance Imaging (MRI). The work in this paper focuses on the application of Convolutional Neural Networks in classifying the stages of this Alzheimer’s disease with the MRI data. In this paper, the used dataset consists of 6400 MRI images, which are basically segmented classification of four categories: Mild Demented, Moderate Demented, Non-Demented, and Very Mild Demented. The data was divided into training, testing, and validation sets to rigorously evaluate the CNN model’s performance. The accuracy of the developed CNN model was 99% overall. Other performance metrics it had were precision, recall, and F1-scores, all of which turned out to be at an average of 99% for the different classes. For perfect classification in the Moderate Demented category, these results describe the high efficacy of CNNs in accurately differentiating between stages of various dementias. The dataset distribution contains 3,200 non-demented images, 896 Mild Demented, 64 Moderate Demented, and 2,240 Very Mild Demented. This success of the classification model in predicting Alzheimer’s disease stages underlines its potential to increase automated diagnostic tools. The work contributes with value to the field, providing a framework that can be reliably used for MRI-based Alzheimer’s classification and lays a foundation for further work in automated disease detection and monitoring.