You are here

Collin Stultz, ML models of adverse cardiovascular outcomes; Primer by Daphne Schlesinger

Daphne Schlesinger
​Research Laboratory of Electronics Computational Cardiovascular Research Group, Massachusetts Institute of Technology
Primer: Modeling cardiovascular physiology

This talk briefly introduces cardiovascular physiology and pathophysiology, from an engineering point of view. Cardiac structure and function can be modeled with reasonable fidelity using familiar concepts from physics, such as fluid dynamics and electromagnetism. For example, flow through the large vessels of the heart can be expressed using a compact circuit, the Windkessel model. Many such relatively simple models form the basis for the “mental models” that cardiologists utilize regularly in patient care. Therefore, an understanding of these models can be leveraged in the development of translatable computational clinical tools. More specifically, physiological models form the basis for physiologically-inspired machine learning models, which leverage prior scientific and medical knowledge to both improve model performance and provide reliable explanations for model inferences.

 

Collin Stultz
Dept. of Electrical Engineering and Computer Science, 
Institute for Medical Engineering and Science,
Computer Science and Artificial Intelligence Laboratory, MIT; 
Massachusetts General Hospital

Physiology-inspired machine learning models for predicting adverse cardiovascular outcomes
A necessary condition for the success of any machine learning model is that it achieves an accuracy that is superior to pre-existing methods. In the healthcare sphere, however, accuracy alone does not, nor should it, ensure that a model will gain clinical acceptance. So, what constitutes a good machine learning model for clinical applications? Unlike problems outside of the medical domain, poor performance for clinical models can have deleterious consequences for patients. In view of the fact that no model, in practice, has 100% accuracy, attempts to understand when a given model is likely to fail should form an important part of the evaluation of any machine learning model that will be used clinically. Models that provide physiologically motivated explanations for a given prediction are useful because they enable clinicians to leverage their understanding of the underlying physiology to gauge whether a given prediction is likely to be correct. In this talk I will describe novel approaches for building models that are motivated by our understanding of cardiovascular pathophysiology to produce predictions and associated explanations that are easily understood by the practicing clinician. It is our view that such models have a higher chance of being embraced, and used, by the clinical community.