Events

NYU Langone Medical AI Seminar: Unleashing the Power of Real-World Data with Machine Learning

Speaker: David Sontag

Location: On-Line : (passcode 447268)
Videoconference link: https://nyulangone.zoom.us/j/95994962949?pwd=MW9kN3BHSFJUUS9OUGphOTJqWG1jUT09

Date: Friday, November 4, 2022

Real-world data from electronic medical records has the potential to transform the pharmaceutical industry and clinical care. In this talk, I first discuss recent work on learning patient simulators of disease progression, which could be used for predicting survival time, personalized side-effect profiles, and treatment effects. Using data from over a thousand patients with multiple myeloma, I show how our recently developed neural PK-PD state space model (Hussain et al., ICML 2021) can more accurately predict a patient's future biomarkers compared to previous time-series modeling approaches. I next turn to the question of discovering disease subtypes from censored data, and introduce a deep generative modeling approach that can disentangle disease subtypes from disease stage (Chen et al., AAAI 2022). I close by discussing what I view as the most important open problem for the field -- how to accelerate the generation of high-quality real-world data. I discuss work on our new MedKnowts system for AI-driven clinical documentation and the results of piloting it at Beth Israel Deaconess Medical Center (Murray et al., UIST 2021).