Explainable AI for Dermatological Precision.
Moving beyond black-box models. Odis utilizes ProtoPNet architectures to provide interpretable, case-based reasoning for dermatological analysis.
The ProtoPNet identified high-spatial symmetry in the lesion structure, matching learned prototypes of benign nevi.
Core Methodology
The Scientific Foundation.
Bridging the semantic gap between neural latent spaces and clinical observations via dual-stage prototypical grounding.
Neural Interpretability
Unlike standard black-box CNNs, Odis identifies specific morphological features aligned with dermatological pathology, providing a transparent audit trail for every diagnosis.
Case-Based Reasoning
"This looks like that." The system identifies 105 clinical anchors from the HAM10000 dataset to justify its findings, mimicking the deductive process of human experts.
Empirical Validation
Accuracy on HAM10000
Learned Clinical Prototypes