Stephanie Hyland - CHIMERA seminar series
28 April 2021, 3:00 pm–4:00 pm
Predicting near-term circulatory failure in the ICU with machine learning
I will discuss my recent paper, “Early prediction of circulatory failure in the intensive care unit using machine learning” (Nature Medicine, 2020). Using a large, high-resolution dataset from a Swiss ICU we built a classifier of near-term (8-hour) circulatory failure in ICU patients with invasive blood pressure monitoring. The model, using gradient-boosted trees, achieved strong predictive performance (AUC >0.9; precision ~30% at 90% recall), with generally consistent performance across patient cohorts. A key component of this work was a careful data processing and feature extraction pipeline, which I will be happy to discuss in detail.
About the Speaker
at Senior Researcher at Microsoft Research in Cambridge
Stephanie Hyland is a Senior Researcher at Microsoft Research in Cambridge. She received her PhD in Computational Biology and Medicine from the Tri-Institutional Training Program of Cornell, Weill Cornell and Sloan Kettering Cancer Centre. Her research focuses on the development of machine learning methods for problems in healthcare, with a focus on intensive and perioperative care, time-series modelling, and the challenges of real-world ML systems.