(in English) |
Explainability is one of the key elements required in medical image diagnosis using deep image recognition models. In this talk, the characteristics of explainable AI models, such as pixel attribution, explanation by prototypes, natural language explanation, conceptual explanation, and counterfactual explanation, have been introduced. Then, provide an overview of the necessary technologies for explainable AI from the perspective of which stakeholders will seek explanations and for which purposes. |