Time Series Machine Learning for Deployment in Healthcare

Explore a 59-minute conference talk on time series machine learning for healthcare deployment, presented by Anna Goldenberg at the EWSC-MIT EECS Joint Colloquium Series. Delve into the application of ML in healthcare settings, including ICU environments and pediatric hospitals. Learn about explainability tools, feature importance, evaluation methods, and representation challenges in medical AI. Discover innovative approaches like HDP Flow States and the Lifeline pipeline for real-time evaluation. Gain insights into model transferability, feature selection, and the use of intervention and continuous data in healthcare ML. Engage with the pressing biomedical questions and foundational machine learning advances discussed in this collaborative series between the Eric and Wendy Schmidt Center at the Broad Institute and AI+D within MIT's EECS Department.

Introduction
SeaKids Hospital
ICU Lawson Labs
Survey
Explainability tool
Feature importance
Evaluation
Representation
Bias
Representations
Trajectory
HDP Flow
States
Model
Underlying States
Blackbox inference
Lifeline pipeline
Realtime evaluation
Challenges
Conclusion
Thank you
Questions
Intervention data
Continuous data
Two short questions
How transferable is this model
Feature selection