Research Projects | Overview
Personalized medicine using novel machine learning algorithms
With the in-depth understanding of new analytical approaches such as machine learning, we have developed high-performing and relevant predictive models in patient care and clinical operations. If we can accurately predict individual patient outcome, we can then be truly selective in diagnostics, imaging, and interventions for those who truly require them. Based on RIVUR trial data, we led a collaborative team from HMS and MIT (Professor Dimitris Bertsimas, Operation Research Center) to develop high-performance models predicting significant clinical outcomes that necessitate invasive diagnostics and treatments in children with initial urinary tract infection (UTI) and vesicoureteral reflux (VUR). Those at high risk of both VUR and recurrent UTI would be managed aggressively with VCUG, while low risk patients would be observed. By assigning management according to true clinical risk, we could achieve optimal outcome on a population level while avoiding adverse effects and costs from unnecessary interventions.
Next-generation urodynamic studies: a deep dive in signal processing and modeling to improve bladder function
Another focus of CHAP uses new analytical approaches for urodynamics (UDS), the gold-standard modality for evaluating bladder and urinary sphincter function prior to urologic and neurologic interventions in patients with complex conditions such as spina bifida. Unfortunately, consistent interpretation is elusive and little consensus exists among experts. Using signal processing and mathematical modeling, computerized and standard interpretations of UDS results is possible. To that end, we have led a collaborative effort between applied mathematicians and clinicians in completing a UDS data pipeline, and preliminary analysis shows great promise in using advanced algorithms to provide consistent and objective UDS pattern recognitions for future interpretation.
Business analytics and operations in healthcare
Maximizing efficiency for clinicians' workflow is paramount to physician satisfaction and prosperity of hospitals and practices. By utilizing the data accumulated in the day-to-day operation, strategies can be formulated into the workflow to increase efficiency. We have focused our attention on operating room scheduling optimization and clinic time predictions with significant preliminary results. There is great promise in revolutionize the way in the operations for the hospitals.