With the increasing prevalence of healthcare sensors, and their ability to communicate wirelessly, it is becoming possible to continuously and remotely monitor the health characteristics of individual users. Furthermore, as these sensors communicate with smart phones, it is also possible to capture other contextual user behaviors (e.g., location, nutrition, activity) that deeply influence health.
This presentation describes a solution for monitoring diabetes patients remotely that provides them with predictive alerts to avoid hypoglycemia and improve health outcomes. As part of this solution, sensor date must be integrated with context to build machine learning-based predictive models that provide continuous insights and alerts to users. The models then require quick adaption to account for modified behaviors as users receive feedback. The presentation explores how the models are personalized to the individual.
Learn about core characteristics of predictive models and applications, and the challenges for deploying adaptive and personalized machine learning in healthcare settings.
Use real-time distributed systems to ingest healthcare and other data to build a live view of the context of an end-user
Create personalized machine learning models that predict hypoglycemia
Continuously deploy and adapt these models as user context changes