A Novel CNN Architecture for Alzheimer's Disease Classification Using MRI Images: Comparison with Traditional Machine Learning Models

Automated medical image processing has significantly improved with recent advances in deep learning and imaging technologies, particularly in the area of neuroimaging-based Alzheimer's disease (AD) diagnosis. Although various imaging modalities are available for individual patients, there is still a lack of research on the creation of multi-modal models that incorporate different modalities. AD is a neurological condition that worsens with time and affects daily activities. It affects memory, cognition, and behavior. AD cannot be cured, but it can be slowed down in its progression and quality of life enhanced with early care. Medical research is using machine learning (ML) more and more to diagnose and track diseases. Magnetic Resonance Imaging (MRI) can be used to detect AD, although it is still challenging to spot subtle changes in the brain in the early stages of the illness. In order to categorize four classes associated with AD symptoms Mild Demented (L-1), Very Mild Demented (L-2), Non-Demented (L-3), and Moderate Demented (L-4) we present a novel CNN architecture in this study. In order to identify the best accurate model, we also evaluate the performance of six ML algorithms. We trained and tested our proposed technique on the DATASET, which consists of 6400 MRI images and is publicly available on Kaggle database. Our results demonstrate that the proposed CNN model outperforms existing state-of-the-art methods with an accuracy of 98.7% and an F1-score of 98.9%. The most effective machine learning model that was looked at was Linear Regression (LR), which had an accuracy of 84.1% and an F1-score of 83.7%.