AI-Assisted MI diagnosis from echocardiogram videos: does explainability enhance human-AI collaborative accuracy?

Key Points

• Diagnostic performance may be impaired by current explainability techniques, leading to misalignment with clinical reasoning.
• AI’s visual explanation strategies must align better with clinician expertise for effective integration into practice.
• Echocardiogram videos serve as vital tools for detecting myocardial infarction, enhancing diagnostic accuracy when combined effectively with AI.
• Future AI models should prioritize explainability aligned with clinical workflows, potentially improving collaborative accuracy.

Abstract