Classification of Mild, Very Mild, Moderate, and Non-Demented Alzheimer’s Disease MRI Scans with SVM

Over 55 million people worldwide are affected by Alzheimer's disease (AD), which causes a substantial financial impact on society. The estimated expenses of health care and long-term care for individuals with forms of dementia are expected to approach around $360 billion in 2024. Recent developments in neuroimaging have yielded important insights into AD. Machine learning (ML) methods have become widely available as effective instruments for interpreting complex neuroimaging data. Previous studies on classifying magnetic resonance imaging (MRI) images found significant success in identifying neurological disorders. In this study, supervised learning is employed to investigate if ML models can distinguish between mild dementia, moderate dementia, very mild Dementia, and non-demented MRI images. The dataset was sourced from the Open Access Series of Imaging Studies (OASIS), a publicly available dataset that provides cross-sectional and longitudinal MRI data of the brain. We trained a linear support vector machine (SVM) classifier with stochastic gradient descent (SGD) on 75% of the dataset. To evaluate performance, we calculated the accuracy level from the testing data. This model produced an accuracy of 74.21%, demonstrating that SVM classification is a promising avenue to classify MRI images.