Cameo improves skin lesion diagnosis fairness and robustness using xai
CAMEO: A Class-Activation-Mapped Equitable Overlay Framework for Fair and Robust Deep Learning-based Skin Condition Diagnosis
Computer Vision and Pattern RecognitionMachine Learning
Summary
Skin condition classifiers can be tricked by backgrounds like skin tone or photo features, making them unfair and less reliable. The authors developed CAMEO, a method that uses AI explanations during training to focus on the actual skin lesion and replace the background with realistic synthetic skin. This reduces errors caused by background cues and helps the model pay attention to the right parts without losing accuracy. Their tests showed that focusing on stable lesion areas makes diagnosis fairer across skin tones and more robust to different images.
What this means in practice
- •For medical imaging software developers: Improve skin lesion diagnostic tools by integrating XAI-driven lesion localization with synthetic background replacement to reduce bias from skin tone and artifacts.$Commercial implications: Enables sale of fairer and more reliable skin disease diagnostic software for dermatology clinics and telemedicine providers.
- •For medical device manufacturers: Incorporate CAMEO's training framework in device-specific model development to ensure robustness across diverse patient skin tones and reduce background confounders.
Authors
Youssef Attia, Debasmita Mukherjee
Abstract
Deep learning classifiers for dermoscopic skin lesions often reach high in-distribution accuracy while quietly relying on spurious background cues such as skin tone, device vignetting, and embedded rulers, rather than on lesion morphology. This undermines robustness and fairness across skin tones. This work asks whether Explainable AI (XAI), typically used only to audit a finished model, can instead be repurposed as an active training signal that corrects this shortcut without sacrificing diagnostic accuracy. We introduce CAMEO (Class Activation Mapped Equitable Overlay), a framework that improves skin-lesion classification by selecting stable model explanations and using them to separate lesions from their backgrounds. It then replaces the background with realistic synthetic skin while keeping the lesion unchanged. On HAM10000 and dark-skin ISIC images, CAMEO maintained accuracy while reducing background-driven errors by nearly four times. It also made the model's attention more consistent when backgrounds changed. Results across multiple tests show that reducing reliance on background information improves robustness, with Fitzpatrick-based backgrounds providing a realistic and interpretable approach. Results show that XAI-guided augmentation can make dermoscopic classifiers measurably more robust and fair at no cost to accuracy. They also clarify that it is the mechanism and not the specific tone palette that matters, and that the lasting contribution of XAI here lies in stability-screened, annotation-free lesion localisation rather than in the robustness number itself.