Control Subject to Human Behavioral Disturbances: Anticipating Behavioral Influences in the Control of Diabetes

Abstract:

This project addresses the design of cyber-physical systems that respond to behavioral disturbances introduced by human users. The primary motivating example of this research is the design of “artificial pancreas” algorithms for the control of blood glucose in patients with Type 1 diabetes who require external insulin throughout the day to maintain glucose homeostasis. Whereas individual patients have their own preferences with regard to the degree of automation needed (with some desiring fully automatic delivery of insulin, others wanting only advice), the control problem itself is largely defined in terms of the uncertainties associated with daily activities: meals and exercise.
From recent clinical studies of fully automated control of diabetes, it is clear that deterministic model predictive methods can effectively deal with the measurement and actuation delays associated with current-generation continuous glucose monitors and insulin pumps. However, it is also clear that outside of the controlled environment of a clinical research center it is necessary to make special accommodations for spurious human behavioral disturbances. Specifically, (i) system disturbances, such as meals and exercise, may have to be announced by the patient (requiring an intrusion upon the patient’s lifestyle), (ii) it may be necessary to integrate specialized algorithms for detecting meals and exercise (which are prone to errors and introduce delay), and/or (iii) the system may need to be de tuned so that unannounced meals and exercise can be safely absorbed by the system. In this work, we take the view that stochastic methods can play a significant role in bridging the gap between model predictive control and uncertain system disturbances caused by human behavior. In particular, we assert that many system disturbances can be modeled (i) as random processes, but not as zero-mean white noise processes, and (ii) as occurring with statistical regularity, but not as periodic. Accordingly, we are developing new mathematical models of human behavioral disturbances. Our premise is that appropriate statistical characterizations of routine behavior allow us to derive new control algorithms that anticipate behavioral disturbances and improve upon the existing (deterministic) model predictive methods currently being evaluated for human-subject clinical experiments without compromising safety. Estimating Behavioral Disturbances via Deconvolution and Patient-Individualized Models – “Blind” System Identification: In prior years of this project we had developed a framework for “blind” estimation of human behavioral disturbances by reconciling continuous sensor data (e.g. a diabetic patient’s continuous BG monitoring data) with known actuation signals (e.g. a patient’s insulin pump log file). These methods are important since humans are prone to reporting relevant behaviors inaccurately (e.g. patient meal diaries). To the extent that behavioral influences can be inferred from objective measurements, we have an alternative means of characterizing patient behavior. In the context of type 1 diabetes, we use deconvolution to infer carbohydrate “net effect” signals from continuous glucose monitoring data and insulin pump records, where the output of the process is an estimated meal-arrival signal. A database of net effect curves for a patient is representative of the patient’s eating and exercise behavior and can be used as a means of “replaying” each day in the database and evaluating the effect of alternative insulin strategies. To achieve accurate predicted responses it is important to ensure that the dynamic system model is actually representative of the patient’s physiology. To address this issue, we have developed an approach for estimating model parameters, so that patients can benefit from net effect- driven insulin advice in the short term (requiring say a week of data between updates) and from improved model accuracy in the long term as more data becomes available. Behavioral Initialization of Closed-Loop Control of Type 1 Diabetes: Over the last year, we have used our “net effect” framework (above) for improving the care of patients with type 1diabetes. Specifically, we address the opportunity to optimize a patient’s basal rate insulin profile, targeting either (i) improved conventional insulin pump therapy or (ii) improved closed-loop control of diabetes. (Regarding the latter, in the closed-loop systems that we are developing, control actions are computed relative to the open-loop profile of basal insulin delivery. To the extent that the open loop basal rate profile is well adapted to the patient’s routine, the closed-loop system is only needed to address transient metabolic disturbances.) The system that we have developed takes seven days of historical data, including continuous glucose monitoring data and insulin pump records, and iteratively applies clinical recommendations for basal profile adjustment, ultimately producing a recommendation for the patient that is adapted to his/her eating and exercise behaviors. Here, we use the seven-day database of net effect curves to validate that the proposed adjustment to the profile profile actually leads to improved outcomes for the patient. This system is being tested now in ongoing clinical trials.

  • CPS Domains
  • Medical Devices
  • Control
  • Modeling
  • Health Care
  • CPS Technologies
  • Foundations
  • National CPS PI Meeting 2013
  • 2013
  • Poster
  • Academia
  • CPS PI Poster Session
Submitted by Stephen Patek on Mon, 10/28/2013 - 10:45