Haejun Chung

CSWinUNETR achieves unprecedented segmentation accuracy for thin anatomical structures, outperforming existing methods without complex post-processing.

Adaptive Binning transforms how we leverage unlabeled medical tabular data, achieving significant performance gains without the need for costly expert labels.

OTCHA's optimal transport-driven approach not only refines token representations but also filters out irrelevant information, leading to superior multi-view classification performance.