Chest X-ray reasoning paper accepted at MICCAI 2026

Yeji X's research team had a paper accepted at MICCAI 2026 (Medical Image Computing and Computer Assisted Intervention) — widely regarded as the top international conference for medical imaging AI. The work will be presented this September in Strasbourg, France.

The accepted paper proposes a diagnostic reasoning model that mirrors how radiologists actually read a chest X-ray: starting from broad anatomical regions and progressively narrowing to specific findings, rather than treating the image as a single black-box classification. The model draws its conclusion along one of 197 anatomically valid reading paths, giving clinicians a traceable rationale behind each output instead of an opaque score.