Nicholas Houstis, MD
Innovation Fellow Project | 2020
As a practicing cardiologist and intensive care physician, Dr. Houstis observed that physicians in the CCU are often presented with an overwhelming volume of data. Moreover, physicians are asked to reason about complex cardiovascular physiology from nonintuitive data like streaming pressure waveforms from multiple sites in the body. Dr. Houstis wondered whether such analyses could in principle be carried out more systematically and quantitatively using computational tools.
During his HTL fellowship, Dr. Houstis set out to develop an algorithm that would analyze a patient’s hemodynamic data and advise the clinician on treatment selection for patients with cardiogenic shock. He used a rapidly evolving technique known as reinforcement learning to successfully train an algorithm to treat cardiogenic shock in a simulation setting. He also co-led a team that assembled and cleaned a large database of CCU data, and he is now using this database to assess how the computational treatment algorithm might perform in a real-world setting. His work as an Applied Innovation Fellow set him up for receipt of the NIBIB, Trailblazer R21 award and he continues his work today.