Hongyu Yang

Achieve state-of-the-art sparse-view CT reconstruction with a diffusion model that generates visually consistent textures using fewer sampling steps, mitigating inherent randomness.

Ditch the RNNs and attention: PFGNet's frequency-guided gating achieves SOTA spatiotemporal prediction with a fully convolutional architecture, slashing parameters and FLOPs.

Diffusion models can learn surprisingly generalizable anatomical representations from unlabeled MRI data, enabling accurate multi-task diagnosis and segmentation across different joints and imaging conditions, even with limited labeled data.