Applying ensemble machine learning techniques to MRI scans to predict Alzheimer's disease.

Area of Science:

• Neuroimaging
• Machine Learning
• Alzheimer's Disease Research

Background:

• Alzheimer's disease (AD) is a leading cause of dementia, necessitating early prediction for timely intervention.
• Predicting AD conversion in cognitively normal (CN) individuals before symptom onset is crucial.
• Structural magnetic resonance imaging (MRI) offers potential for early AD detection.

Purpose of the Study:

• To develop and evaluate a machine learning framework for predicting future AD conversion in CN individuals.
• To utilize structural MRI data for identifying preclinical AD signatures.
• To assess the efficacy of an ensemble machine learning approach for early AD risk identification.

Main Methods:

• A machine learning framework employing transfer learning with a pre-trained VGG16 model for feature extraction from structural MRI.
• Processing of five representative 2D MRI slices per scan to generate imaging descriptors.
• Classification using an ensemble of support vector machines (SVM), random forests (RF), and artificial neural networks (ANN) with soft voting.

Main Results:

• The ensemble model achieved a median AUC-ROC of 0.951, accuracy of 0.872, recall of 0.923, and F1 score of 0.811.
• Evaluation on 1,093 subjects from the OASIS-3 dataset using person-wise stratified cross-validation.
• Demonstrated high performance in detecting preclinical AD signatures from structural MRI data.

Conclusions:

• Ensemble machine learning effectively detects preclinical Alzheimer's disease signatures from structural MRI.
• The developed framework provides a practical, accessible, and cost-effective tool for early AD risk identification.
• This approach supports timely interventions by enabling early prediction of AD conversion.