Edge Analytics for Personalized Virtual Care

For chronic diseases such as Diabetes, continuous medical care is required to minimize the risk of acute and long-term complication. The therapy is complex and often comprises various medical and life-style related measures. Smart health technologies like continuous glucose monitoring (CGM) and healthcare wearables offer large data sets used for early prevention or improvement of the therapy in the hospital as well as at home. Together with data-mining and machine learning (ML) technology, connected health brings full potentials and open problems for academia and industry.

In UCSD, we work with Kaiser Permanente to enhance Diabetes therapy and focus on three targets: 1) define the valuable problems in Diabetes therapy and providing personalized and actionable knowledge to users with ML techniques. 2) propose a hierarchical edge-cloud data analytic architecture to tackle the privacy and resource (computing / communication / data storage) management problems regarding personal sensor data. 3) build an edge analytic testbed continuously collecting, tracking and analyzing sensor signals.



Healthcare Informatic: Data Analytic and Personalized Recommendation using Wearable

Nowadays emerging consumer health wearables enable people to collect personal health data, including activity tracking (e.g. exercise and sleep) and vital signal (e.g. blood pressure and heart rate), in good granularity. Personalized healthcare including monitoring and recommendation based on such data has great potential to fight chronic diseases (e.g. diabetes and hypertension). However, it is difficult to gain insight from such data without proper analysis. The objective of this study is using machine learning techniques to 1. Find associational and causal inference between target variables in chronic disease (e.g. blood pressure and glucose level) and health behaviors. 2. Provide personalized recommendation to users to improve their health targets.  


Hybrid Edge Analytics

Many current Internet services rely on models trained on user data. Commonly, both the training and inference tasks are carried out using cloud resources processing personal data collected at edge devices from users. There are two major disadvantages: 1. Time and cost exchanging raw data and analyzed result between cloud and edge. 2. Risk of transmitting private and sensitive data of users.

We plan to setup a prototype environment with connected health technology in the new inter-disciplinary engineering building. Unlike traditional point-of-care, this connected health environment will integrate various kinds of data (vital/ behavioral/ environmental) and provide interactive experience for users. With state-of-the-art sensors and data gateway, the collected data will be streamed and aggregated in order to provide comprehensive evaluation on the targeted problem. Furthermore, the new space can lead to more possible inter-disciplinary cooperation between medical science and engineering toward next-generation health applications.