Performance Evaluation of Deep, Shallow and Ensemble Machine Learning Methods for the Automated Classification of...

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Medicine
• Neuroscience

Background:

• Artificial intelligence (AI) is vital for computer-aided diagnosis (CAD), with deep learning (DL) showing promise in classifying Alzheimer's disease (AD) stages.
• Traditional machine learning (ML) models can match or exceed DL performance, particularly when training data is limited.

Purpose of the Study:

• To propose and evaluate an ensemble ML classifier for AD diagnosis using magnetic resonance imaging (MRI) data.
• To compare the performance of ML classifiers against DL algorithms in data-scarce and data-rich scenarios.

Main Methods:

• An ensemble classifier was developed using popular ML models applied to MRI data.
• Evaluated ML classifiers on the Alzheimer's Disease Neuroimaging Initiative and Open Access Series of Imaging Studies datasets.
• Compared ML ensemble performance against state-of-the-art DL algorithms.

Main Results:

• The proposed ML ensemble classifier achieved an accuracy of 96.52% for AD classification.
• Demonstrated a 3-5% performance improvement over the best individual ML classifier.
• ML classifiers showed effectiveness in both data-scarce and data-rich conditions.

Conclusions:

• Ensemble ML classifiers provide a highly accurate and data-efficient approach for AD diagnosis via MRI.
• This study offers guidance on selecting appropriate AI algorithms for AD classification based on data availability.