An Explainable AI Paradigm for Alzheimer's Diagnosis Using Deep Transfer Learning

Area of Science:

• Artificial Intelligence
• Neuroscience
• Medical Imaging

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder impacting millions globally, necessitating early and accurate diagnosis for effective management.
• Deep learning shows promise in analyzing neuroimaging for AD diagnosis, but lack of interpretability limits clinical trust and adoption.
• Explainable AI (XAI) is crucial for understanding AI decision-making processes in healthcare.

Purpose of the Study:

• To develop an explainable AI (XAI)-based framework for Alzheimer's disease diagnosis.
• To enhance the interpretability and transparency of deep learning models in medical image analysis for AD.
• To improve clinical trust and acceptance of AI tools in AD diagnosis.

Main Methods:

• Leveraged deep transfer learning with pre-trained Convolutional Neural Networks (CNNs) including VGG16, VGG19, DenseNet169, and DenseNet201.
• Developed ensemble models (Ensemble-1: VGG16/VGG19, Ensemble-2: DenseNet169/DenseNet201) to improve diagnostic performance.
• Integrated XAI techniques, specifically saliency maps and gradient-weighted class activation mapping (grad-CAM), into a novel diagnostic model.

Main Results:

• Individual CNN models were evaluated, with ensembles showing superior performance.
• Ensemble models achieved up to 95% accuracy, precision, recall, and F1 scores.
• The novel XAI-integrated model demonstrated a remarkable accuracy of 96%, providing visual insights into diagnostic reasoning.

Conclusions:

• Combining deep transfer learning with XAI techniques offers a powerful approach for Alzheimer's disease diagnosis.
• The developed XAI framework enhances model interpretability, providing visual evidence for clinical decision support.
• This research paves the way for more transparent, trustworthy, and clinically applicable AI models in neurodegenerative disease diagnosis.