At the Lucile Packard Children’s Hospital in California, Suchi Saria and colleagues analyzed mountains of continuously streaming data, monitoring the vital signs of vulnerable, premature infants. Using this information they were able to predict, with over 90% accuracy, which of the tiny babies were headed for life-threatening complications.
The team’s revolutionary real-time electronic assessment and scoring tool, PhysiScore, provides healthcare workers the information they need for targeted medical interventions. The tool, touted on the cover of Science Translational Medicine, underscores the value of applying machine learning to the often messy, noisy world of healthcare systems data. “We have access to high granularity patient data from very large health systems. This data is collected as a by-product of patient care but previously, most of it was ignored. For complex diseases like autism that we barely understand, we could significantly advance clinical practice. The trick is to build techniques that can draw insights from this data and present them within the right clinical context,” she says.
Saria joins the Johns Hopkins’ Department of Computer Science faculty after she completes a year-long NSF Computing Innovation Fellowship at Harvard Medical School. She received her PhD from Stanford, where her thesis explored computational techniques for decoding longitudinal electronic health record data.
Saria’s grand hope is to dramatically reduce the feedback loop in medicine by building rapidly learning health systems. Highly qualified students interested in machine learning and health informatics, with a background in computer science, statistics, may apply to the graduate programs in the Computer Science, Biomedical Engineering or Biostatistics.