AI for Healthcare Imaging
Machine learning for radiology, pathology, and cardiology with emphasis on clinically meaningful prediction, localization, and explainability.
Healthcare imaging presents a rich but challenging setting for machine learning: clinically relevant findings may be subtle, labels can be incomplete, and deployment requires trust and interpretability. This theme develops AI methods for radiology, pathology, and cardiology that emphasize clinically aligned prediction, localization, and transparent model behavior.
We design models that integrate limited supervision, human workflow signals, and robust representation learning to improve both performance and reliability. A central goal is to produce systems that are not only accurate on benchmark tasks but also inspectable and useful in clinical decision support, where understanding model evidence and failure modes is essential.
Methods
Application Areas
Selected Projects
Contrastive Learning for Histopathological Image Analysis
Representation learning and weak / limited supervision for robust pathology classification and grading.
View project →
Interpretable AI for Radiology
Interpretable deep learning for radiology—combining clinical workflow signals, localization supervision, and expert-aligned explanations so imaging models are accurate and inspectable.
View project →
ECG Machine Learning
Machine learning for electrocardiograms—supervised baselines, self-supervised contrastive pretraining, and transformer models for cardiac phenotypes, diagnostic support, and clinically reliable prediction with limited labels.
View project →Selected Publications
CLASS-M: Adaptive stain separation-based contrastive learning with pseudo-labeling for histopathological image classification
F2FLDM: Latent Diffusion Models with Histopathology Pre-Trained Embeddings for Unpaired Frozen Section to FFPE Translation
HistoEM: A Pathologist-Guided and Explainable Workflow for Histopathological Image Analysis Using Expectation Maximization
Performance of off-the-shelf machine learning architectures and biases in low left ventricular ejection fraction detection
Prediction of Obstructive Lung Disease from Chest Radiographs via Deep Learning Trained on Pulmonary Function Data
