Objective:

Early Prediction of Alzheimer's Disease from Asymptomatic Individuals Using Structural MRI Neuroimaging Analysis.

Background:

Alzheimer's disease (AD) poses a significant global health challenge, with an increasing prevalence and substantial economic burden. This study leverages machine learning techniques to develop a predictive model for early AD detection using structural brain MRI features, addressing the critical need for early intervention strategies.

Design/Methods:

The study utilized the OASIS-3 dataset, comprising 1,378 participants. Structural MRI data were segmented using FreeSurfer, and feature importance analysis identified the top 10 influential features. Seven classification algorithms were evaluated, with performance assessed before and after implementing Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance. Model interpretability was enhanced through Explainable AI techniques.

Results:

Support Vector Machine (SVM) emerged as the best-performing model, achieving 93.3% accuracy and 68.9% sensitivity post-SMOTE implementation. Right inferior lateral ventricle volume, right middle temporal thickness, and right temporal pole thickness were identified as the most important predictive features. SMOTE application significantly improved model performance across all algorithms, particularly in sensitivity and F1-scores.

Conclusions:

This study demonstrates the potential of machine learning, especially SVM, in detecting preclinical AD using structural MRI analysis. While the results are promising, limitations such as the short follow-up period and reliance on a single dataset necessitate cautious interpretation. Future research should focus on external validation, multimodal approaches, and addressing ethical implications for clinical translation. The findings underscore the importance of early detection in AD management and clinical trial design.