My research focuses on computer vision and medical AI, with particular interest in endoscopic image&video analysis and brain imaging diagnosis. I am also interested in data-limited learning in medical scenarios, including learning with noisy labels, semi-supervised learning, and weakly supervised learning, to improve model robustness under realistic clinical annotation settings. In parallel, I explore multimodal representation learning through vision-language alignment and pretraining, and I further work on multimodal large language models, especially efficient MLLMs for clinically grounded reasoning and deployment.