AIMA’s Website
Our lab focuses on advancing artificial intelligence methods for scientific discovery and real-world impact. We develop principled machine learning and deep learning approaches, with an emphasis on robust representation learning, optimization, and generalization. Our research bridges theory and practice, aiming to design algorithms that are both mathematically grounded and empirically effective across diverse applications..
Research Interests
Robust & Data-Efficient Learning in Medical AI
We develop robust and data-efficient artificial intelligence methods for healthcare, focusing on learning under imperfect and limited supervision. Our research addresses real-world clinical challenges such as annotation scarcity, domain shift, imbalanced subpopulations, occlusion, and noisy acquisition conditions in medical imaging.
Key themes include:
• Learning with limited and noisy annotations
• Semi-supervised, self-supervised, and few-shot learning
• Domain adaptation and robustness analysis
• Uncertainty quantification in clinical AI
Medical Image Computing & Foundation Models
We investigate fundamental problems in medical image computing, including segmentation, reconstruction, registration, and classification across multiple imaging modalities such as MRI, CT, ultrasound, and microscopy.
Our work spans:
• 2D and 3D medical image segmentation
• Foundation models in medical imaging
• Multi-modality learning and imaging–clinical data integration
• Benchmarking and systematic robustness evaluation
Translating AI into Clinical Impact
Through rigorous benchmarking and principled model design, we aim to bridge methodological advances in machine learning and real-world clinical deployment. Our goal is to build reliable, generalizable, and interpretable AI systems that can operate under realistic clinical constraints.
We welcome collaborations and motivated students interested in AI for healthcare.