Lead: Dr. Christopher Re
A framework that integrates diverse modeling paradigms and data types will maximize predictive power, while an emphasis on presenting transparent models will accelerate adoption by researchers and translation to the clinic.
Different types of data (e.g., high-fidelity experimental data from labs and low-fidelity data from wearable sensors) and models (e.g., biomechanical models and statistical models) are currently isolated and hard to integrate. We will dissolve these barriers and develop trained systems that integrate diverse models and data types using scalable statistical inference. In addition to making accurate predictions, the systems will clearly communicate with clinicians and integrate with their decision-making process, a current barrier to the translation of predictive models to medical practice. We will apply machine learning methods to develop novel automatic tools for models to explain themselves and help doctors improve their decision-making processes.