Quantifying Explanation Consistency: The C-Score Metric for CAM-Based Explainability in Medical Image Classification
2026-04-09 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionArtificial Intelligence
AI summaryⓘ
The authors looked at ways to check if computer models used in medical imaging explain their decisions consistently, not just correctly. They created a new score called the C-Score that measures how similarly the model focuses on important areas across different patients with the same disease, without needing expert labels. Testing six explanation methods on chest X-rays, they found that some models' explanations became unreliable even before their overall accuracy dropped. Their C-Score can warn when a model’s explanations are about to fail, helping choose better models for medical use based on explanation quality, not just accuracy.
Class Activation Mappingdeep learningmedical imagingexplanation consistencyC-Scorechest X-raytransfer learninglocalisation fidelitysoft IoUmodel stability
Authors
Kabilan Elangovan, Daniel Ting
Abstract
Class Activation Mapping (CAM) methods are widely used to generate visual explanations for deep learning classifiers in medical imaging. However, existing evaluation frameworks assess whether explanations are correct, measured by localisation fidelity against radiologist annotations, rather than whether they are consistent: whether the model applies the same spatial reasoning strategy across different patients with the same pathology. We propose the C-Score (Consistency Score), a confidence-weighted, annotation-free metric that quantifies intra-class explanation reproducibility via intensity-emphasised pairwise soft IoU across correctly classified instances. We evaluate six CAM techniques: GradCAM, GradCAM++, LayerCAM, EigenCAM, ScoreCAM, and MS GradCAM++ across three CNN architectures (DenseNet201, InceptionV3, ResNet50V2) over thirty training epochs on the Kermany chest X-ray dataset, covering transfer learning and fine-tuning phases. We identify three distinct mechanisms of AUC-consistency dissociation, invisible to standard classification metrics: threshold-mediated gold list collapse, technique-specific attribution collapse at peak AUC, and class-level consistency masking in global aggregation. C-Score provides an early warning signal of impending model instability. ScoreCAM deterioration on ResNet50V2 is detectable one full checkpoint before catastrophic AUC collapse and yields architecture-specific clinical deployment recommendations grounded in explanation quality rather than predictive ranking alone.