Automated Patient Clinical Trial Identification (2015)
This project automatically assigns diagnoses to discharged patients for potential followup with clinical trial coordinators.
Predictive Models for Medicine (2015)
Predictive models allow clinicians and health professionals to make data driven clinical decisions. We have several projects to predict surgical site infection, readmission, and emergency room triage. In collaboration with IT and various clinical departments, my lab builds, models, and deploys machine learning based clinical models.
Text Classification (2015)
We study the theory and application of machine learning algorithms for text classification. Text classification is an ideal platform to understand how theoretical aspects of machine learning work as the words that make classifications are human understandable.
We implemented a pilot study tracking alcohol use through the Twitter stream. From a public health perspective, active surveillance of alcohol use allows anti-alcohol campaign monitoring and data driven public policy decision making. Our novel work shows that it is indeed possible to track alcohol use through the Twitter stream.
Curriculum Classification (2014)
Assigning topic categories to medical school curricular requires manual effort and domain expertise by coders or lecturers. We apply machine learning based text categorization models from MEDLINE articles to assign MESH topic categories to medical curricular documents. This work serves as an effective example of the use of domain adaptation in applying a machine learning model built in one application to another applications.
Text Messaging for Health Based Interventions (2014)
We built and deployed a targeted text messaging reminder system. In this work, we built a system to support randomization to treatment and control arms to remind patients to return to clinic for addiction and ectopic pregnancy followup.