Voting Ensemble Approach for Enhancing Alzheimer's Disease Classification.

Area of Science:

• Neurology
• Medical Imaging
• Machine Learning

Background:

• Alzheimer's disease (AD) is the most common form of dementia, characterized by cognitive and memory impairments.
• Accurate AD diagnosis is crucial for effective patient treatment and management.
• Current MRI-based classification methods face challenges due to limited data and noise, impacting reliability.

Purpose of the Study:

• To develop a more accurate and reliable classification method for Alzheimer's disease using longitudinal MRI data.
• To overcome the limitations of single classifier models in AD diagnosis.
• To enhance diagnostic performance using an ensemble voting approach.

Main Methods:

• Proposed an ensemble voting classifier model to combine predictions from multiple individual classifiers.
• Utilized longitudinal MRI data for feature extraction and classification.
• Evaluated the model on the Open Access Series of Imaging Studies (OASIS) dataset.

Main Results:

• The ensemble voting classifier demonstrated superior performance over existing methods on the OASIS dataset.
• Achieved high accuracy (96.4%) and Area Under the Curve (AUC) of 97.2% for binary classification (dementia vs. no dementia).
• Showcased improvements in key assessment criteria including accuracy, sensitivity, specificity, and AUC.

Conclusions:

• The proposed ensemble voting method offers a more robust and accurate approach for Alzheimer's disease diagnosis from MRI data.
• This technique addresses the challenges associated with limited sample sizes and data noise in longitudinal MRI analysis.
• The findings suggest a promising direction for improving early and precise detection of Alzheimer's disease.