Research

The Machine & Medical Vision Lab (MMVL) focuses on building reliable and interpretable AI systems for real-world computer vision applications. However our primary research focus is on medical image analysis, including disease classification, lesion detection, organ/tissue segmentation, and clinical decision support. We work on vision pipelines that can handle practical challenges such as limited annotations, variations in scanners and imaging protocols, noise, motion artifacts, and class imbalance. A key goal of our lab is to develop methods that are not only accurate on benchmarks, but also generalize well across diverse imaging environments.

We develop various machine learning and deep learning approaches focusing on multimodal learning, federated learning, continual learning, vision and language models, uncertainty estimation, and out-of-distribution (OOD) detection. Further, we explore Generative AI approaches for healthcare, including representation learning, synthetic data generation, and foundation model adaptation.