Deep Learning-Based Ensembling Technique to Classify Alzheimer's Disease Stages Using Functional MRI.

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Cognitive decline, including memory loss and impaired judgment, affects the elderly, potentially leading to Alzheimer's disease (AD).
• Early detection of Mild Cognitive Impairment (MCI) is vital as it is potentially reversible, unlike advanced AD.
• Current diagnostic methods often identify AD only after it has progressed significantly.

Purpose of the Study:

• To develop an accurate method for diagnosing Alzheimer's disease (AD) and its various stages for timely treatment.
• To explore the efficacy of deep learning models in classifying functional MRI scans for AD detection.
• To advance the multiclass classification of AD into six distinct stages.

Main Methods:

• Utilized functional MRI images from a publicly available dataset.
• Implemented a two-step approach: binary classification (MCI vs. AD) using Custom CNN, followed by multiclass classification (six AD stages) with multiple deep learning models.
• Employed a max-voting ensembling technique to enhance model performance.

Main Results:

• Achieved high accuracy in binary and multiclass classification of Alzheimer's disease stages.
• The max-voting ensembling technique significantly improved the overall performance of the deep learning models.
• An average accuracy of 98.8% was obtained for the classification of Alzheimer's disease stages.

Conclusions:

• Deep learning models, particularly when combined with ensembling techniques, demonstrate significant potential for accurate, early-stage Alzheimer's disease diagnosis.
• The proposed method offers a promising approach for classifying individuals into one of six stages of Alzheimer's disease using fMRI data.
• Accurate staging of Alzheimer's disease is feasible and can aid in timely and effective patient management.