As a research scientist in Department of Clinical Informatics, I am engaged at large scale Healthcare Data Analytics. By collaborating with other Machine Learning scientists, Medical Imaging experts, Physicians and Radiologists, we use Machine Learning, Feature Engineering, Deep Learning based image recognition and NLP techniques, to facilitate representation of medical concepts, provide clinical decision support in Radiology and derive data-driven strategy for optimizing healthcare services and resources. Some of our research topics include: - Automatic Annotation of Chest X-ray Images Using Convolutional Neural Networks - Representation Learning of Finding Codes in Structured Echocardiogram Reports

- We investigate a general framework of multiplicative multitask feature learning, with applications in analyzing predictors of Alzheimer’s Disease, drug discovery and robotics.- We construct learning models to learn from multiple biomedical annotations based on bi-convex optimization. The models learn from inconsistent labels and evaluate annotators' labeling expertise.- We construct learning models to deal with two sources of annotation ambiguity: (1). multiple labelers provided inconsistent versions of labels; (2). a class label is associated with a bag of instances rather than an individual instance.- Working with UCONN Health Center (UCHC), we construct computer vision based system to predict if a tympanostomy tube is mounted and if the ear has been infected. - Working with UCONN health center, we develop the web-based application for enhancing asthma guideline compliance. http://www.labhealthinfo.uconn.edu/EasyBreathing/

Working with Big Data Analytics group and Design team at Philips, we develop a data-driven system to investigate and visualize CMS referral patterns for deriving market strategies. Prototype is designed, implemented and successfully transferred to Healthcare Transformation Services at Philips.