A Feature-Augmented Explainable Artificial Intelligence Model for Diagnosing Alzheimer's Disease from Multimodal...

Area of Science:

• Artificial Intelligence in Medicine
• Neuroscience
• Medical Imaging Analysis

Background:

• Alzheimer's disease (AD) diagnosis requires integrating complex data like clinical information, cognitive scores (MMSE), and MRI scans.
• Current AI diagnostic tools often lack transparency, hindering clinical adoption.
• Explainable AI (XAI) methods are crucial for building trust and facilitating the use of AI in healthcare.

Purpose of the Study:

• To evaluate a Feature-Augmented explainable AI (XAI) approach for supporting Alzheimer's disease (AD) diagnosis.
• To assess the impact of AI explainability on clinician trust and understanding in AD diagnosis.
• To integrate clinical data, MMSE scores, and MRI scans within an XAI framework.

Main Methods:

• A survey-based evaluation involving five experienced physicians specializing in AD.
• Participants assessed AI-generated diagnostic outputs, including clinical feature interpretations, MRI heat maps, and combined explanations.
• The XAI approach combined rule-based reasoning with example-based visualization for enhanced interpretability.

Main Results:

• The explainable AI model achieved a 100% trust score among participating clinicians.
• 80% of clinicians expressed conditional trust, indicating a need for further clarification alongside AI insights.
• The integrated explanation format significantly improved clinician understanding and confidence in AI-assisted AD diagnosis.

Conclusions:

• This study is the first to gather expert clinician feedback on an XAI model for AD diagnosis.
• Explainability is a critical factor in fostering trust and usability of AI tools in clinical practice.
• XAI approaches are particularly valuable for supporting experienced clinicians in complex diagnostic tasks like AD.