Advancing Neurodegenerative Disorder Diagnosis: A Multiclass Approach Utilizing Machine Learning on fMRI Data

Abstract

This study pioneers advancements in mental health diagnostics by integrating advanced machine learning techniques with neuroimaging to enhance the accuracy of classifying depression, Alzheimer’s disease, and normal controls using fMRI data processed with SPM12 software. The methodology involves extracting deep features from fMRI data through the VGG19 model, capturing intricate patterns indicative of mental health conditions. Subsequently, three machine learning algorithms are employed for classification, with k-nearest neighbor (KNN) emerging as the top performer, achieving an impressive accuracy of 98.73%. This highlights the robustness of the chosen classifiers in managing multiclass neuroimaging data.The findings underscore the efficacy of the proposed methodology, particularly the superior accuracy of KNN in distinguishing between individuals with depression, Alzheimer’s disease, and normal controls. A thorough evaluation of the classifiers sheds light on their respective strengths and limitations, offering a deeper understanding of their clinical applicability.This research introduces a novel approach that not only advances the field but also sets new benchmarks in neuroimaging-based diagnostics, promising a transformative shift in how machine learning is applied to mental health diagnosis.