Research Project
Multivariable Artificial Pancreas to Detect and Mitigate the Effects of Unannounced Physical Activities and Acute Psychological Stress
Our research team pioneered the multivariable artificial pancreas (mvAP) that uses physiological and continuous glucose monitoring (CGM) data to mitigate the effects of meals and exercise without any manual inputs. Our mvAP has performed successfully in clinical experiments. We have also collected data in experiments with various types of physical activities and acute stress inducements and developed classification and intensity estimation algorithms for these activities using machine learning. Our multidisciplinary team includes endocrinologists, nurse researchers, biomedical engineers, exercise physiologists, behavioral scientists and statisticians. The overall objective of this proposal is to develop mvAP technology that addresses physical activity and stress-related challenges and complex situations such as their concurrent occurrence.
- Principal Investigator
- Quinn, Lauretta
- Start Date
- 2021-08-01
- End Date
- 2023-07-31
- Funding Source
- Illinois Institute of Technology
Abstract
Our research team pioneered the multivariable artificial pancreas (mvAP) that uses physiological and continuous glucose monitoring (CGM) data to mitigate the effects of meals and exercise without any manual inputs. Our mvAP has performed successfully in clinical experiments. We have also collected data in experiments with various types of physical activities and acute stress inducements and developed classification and intensity estimation algorithms for these activities using machine learning. Our multidisciplinary team includes endocrinologists, nurse researchers, biomedical engineers, exercise physiologists, behavioral scientists and statisticians. The overall objective of this proposal is to develop mvAP technology that addresses physical activity and stress-related challenges and complex situations such as their concurrent occurrence. Our specific aims are: Aim 1. To develop the mvAP algorithms that identify various types of physical activities, acute psychological stress episodes, their concurrent presence and their characteristics in real time. We have already developed machine learning and model identification algorithms that work with historical data. The proposed work will focus on extending these algorithms and developing software to conduct detection, classification, estimation of physical activities and acute psychological stress from real-time data streaming from CGM and wearable devices, and updating the glucose concentration estimation models and control system parameters accordingly. Aim 2: To conduct open-loop studies in clinic and at home with free-living activities to expand the types and intensities of physical activities and acute stress inducements, to enrich our database with data collected during activities of daily living (alone or coupled with stress), training for races and competitions in races, and capture their impact on glucose level variations under real-life conditions. Aim 3: To conduct clinical experiments with the mvAP with an acute stress detection, classification and intensity estimation module, an updated physical activity detection, classification and intensity estimation module, an updated recursive model identification module and control algorithms that take into account the characteristics of physical activities, acute stress and the concurrent presence. We will conduct the experiments at clinical settings and in free living to assess the performance of our mvAP system. At the end of the funding period, we propose to demonstrate a viable, functionally integrated multivariable AP that will mitigate meal, stress and exercise challenges without any manual inputs to better regulate glucose levels. Large-scale clinical studies to assess the mvAP will be proposed in a future proposal to NIDDK.