Alzheimer's disease classification using a hybrid deep learning approach with multi-layer U-net segmentation and XAI...

Area of Science:

• Artificial Intelligence in Medicine
• Neuroscience
• Medical Imaging Analysis

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder significantly impacting cognitive function.
• Early and accurate diagnosis of AD is crucial for effective therapeutic interventions and patient management.
• Artificial intelligence (AI), particularly deep learning, shows significant promise for improving AD classification accuracy.

Purpose of the Study:

• To propose and evaluate a novel deep learning framework for classifying Alzheimer's disease (AD), Mild Cognitive Impairment (MCI), and Cognitively Normal (CN) individuals.
• To enhance diagnostic accuracy through advanced segmentation and hybrid classification techniques.
• To improve model interpretability and clinical trustworthiness using explainable AI (XAI).

Main Methods:

• A four-phase deep learning methodology was employed, starting with whole brain and gray matter segmentation using Multi-Layer U-Net.
• Feature extraction and classification were performed using a hybrid approach combining Multi-Scale EfficientNet with Support Vector Machines (SVM).
• Explainable AI (XAI) techniques, specifically Saliency Map Quantitative Analysis, were integrated for model interpretability.

Main Results:

• The proposed deep learning model achieved high classification performance across three categories: Alzheimer's Disease (AD), Mild Cognitive Impairment (MCI), and Cognitively Normal (CN).
• Overall accuracy reached 97.78% ± 0.54%, with precise metrics for precision, recall, and F1 scores across all classes.
• The results demonstrate the model's effectiveness in accurately differentiating between AD, MCI, and CN stages.

Conclusions:

• The developed deep learning framework, integrating segmentation and hybrid classification with XAI, shows significant potential for accurate Alzheimer's disease diagnosis.
• The high performance metrics underscore the proposed approach's capability in classifying different stages of Alzheimer's disease.
• Future research will focus on validating the model on public datasets and incorporating advanced XAI for enhanced diagnostic reliability.