Machine Learning

The study will develop and implement a validated machine learning model to optimize voiding cystourethrogram timing and use for diagnosing vesicoureteral reflux (VUR) in children, aiming to reduce the significant health and economic impacts of VUR and recurrent febrile urinary tract infections (fUTIs) by standardizing practices, minimizing unnecessary procedures, and ensuring timely diagnosis for those at highest risk, ultimately seeking to prevent renal injury from fUTIs.

The study will develop, validate, implement, and assess the usability of a machine learning based clinical decision support tool designed to reduce laboratory testing overutilization in pediatric intensive care unit patients.

This study aims to optimize the use of donor hearts for infants and children awaiting heart transplantation by developing predictive models to assess in real-time the potential for transplant success and to evaluate risk. Researchers plan to display these data through intuitive visualizations on a custom-built interface to reduce clinicians’ cognitive burden, enhance decision making confidence, and help ensure the best donor match for pediatric patients.

This research uses large language models and machine learning to retrospectively analyze electronic health records of patients with suspected sepsis and identify patterns in treatment outcomes, with the goal of shaping future clinical guidelines that help doctors select the most effective antibiotics for each patient, reduce unnecessary use of broad-spectrum antibiotics, lower the risks of drug resistance, and ultimately improve patient outcomes.