A Cascaded Mutliresolution Ensemble Deep Learning Framework for Large Scale Alzheimer's Disease Detection Using Brain...

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Imaging

Background:

• Alzheimer's disease (AD) is a progressive dementia impacting cognitive functions.
• Current AD diagnosis lacks a definitive single test, often relying on imaging like MRI.
• Identifying reliable diagnostic tools for AD is crucial for early intervention.

Purpose of the Study:

• To develop an advanced deep learning framework for accurate Alzheimer's disease diagnosis.
• To enhance predictive performance in detecting Alzheimer's disease using brain Magnetic Resonance Imaging (MRI).

Main Methods:

• An integrative, multi-resolutional ensemble deep learning framework utilizing a hierarchical PartialNet design.
• The framework was evaluated on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset comprising 379 subjects.
• Comparison against established deep learning models like ResNet and DenseNet variants.

Main Results:

• The proposed PartialNet framework achieved superior performance in both multi-class and binary-class AD detection.
• Demonstrated enhanced learning capabilities through diversified depth, deep supervision, and feature reuse.
• Outperformed state-of-the-art deep learning approaches on the benchmark ADNI dataset.

Conclusions:

• The developed deep learning framework offers a powerful and accurate tool for Alzheimer's disease diagnosis via MRI.
• PartialNet's architecture provides advantages in gradient flow, parameter efficiency, and training time.
• This approach holds significant potential for improving early and accurate detection of Alzheimer's disease.