Abstract

The integration of artificial intelligence (AI) into medical imaging is progressing rapidly. It is essential for these AI tools to be transparent, interpretable, and explainable to gain the trust of clinicians and regulators. Current state-of-the-art explainable AI (XAI) techniques in imaging includes Local Interpretable Model-Agnostic Explanations (LIME), Shapley Additive Explanations (SHAP), and Gradient-weighted Class Activation Mapping (Grad-CAM). Recent studies have shown that LIME often suffers from inconsistency and unreliability which limits their utility in sensitive fields like medical imaging. This paper proposes Influence Map based Explanation (IME), an enhanced variant of the original LIME framework, that aggregates multiple runs to reduce variability and improve explanation reliability. To validate the effectiveness of the proposed approach, we apply it to image-based classification of Alzheimer's Disease (AD) using deep neural networks trained to predict stage of AD from magnetic resonance images. We quantitatively evaluated IME using two complementary metrics (Confidence Impact Ratio (CIR) and Decision Impact Ratio (DIR)) and compared its performance to existing state-of-the-art XAI methods. Results show that the proposed approach performed better than state-of-the-art XAI methods. Furthermore, the results show statistically significant improvements, demonstrating the effectiveness of the proposed IME in producing meaningful explanations.

Department(s)

Electrical and Computer Engineering

Keywords and Phrases

Alzheimer's disease; Grad-CAM; LIME; MRI; SHAP; XAI

Document Type

Article - Conference proceedings

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2026, All rights reserved.

Publication Date

01 Jan 2026

Download

Full Text Link

Share

 
COinS