Seeing Isn't Always Believing: Analysis of Grad-CAM Faithfulness and Localization Reliability in Lung Cancer CT Classification

Teerapong Panboonyuen
2026
PDF Cite Code Project Source Document ArXiv Hugging Face

Abstract

This work critically evaluates the faithfulness and localization reliability of Grad-CAM for lung cancer CT classification across CNN and Vision Transformer architectures. Results reveal strong model-dependent variability and reduced reliability for transformer-based models, highlighting key limitations of saliency-based XAI in medical imaging.

Type
Conference paper
Publication
In arXiv:2601.12826 [cs.CV]

explainable-ai grad-cam medical-imaging lung-cancer ct-classification xai-evaluation model-interpretability Vision Transformers Convolutional Neural Networks saliency-maps clinical-ai trustworthy-ai computer-vision
Teerapong Panboonyuen
Teerapong Panboonyuen

My research focuses on leveraging advanced machine intelligence techniques, specifically computer vision, to enhance semantic understanding, learning representations, visual recognition, and geospatial data interpretation.

Related