Closed-loop

The "closed-loop" concept for an Artificial Pancreas




The artificial pancreas must function to deliver insulin as needed in a highly dynamic environment. The insulin pump can deliver the insulin, and the continuous glucose measuring system (CGMS) can help determine the insulin actually needed. The goal of the artificial pancreas is twofold:
 * 1) to improve insulin replacement therapy until glycemic control is practically normal as evident by the avoidance of the complications of hyperglycemia, and
 * 2) to ease the burden of continual therapy for the insulin-dependent.

The basic concept considered here involves "closing the loop" between the insulin pump and the CGMS. As in a modern control system, the blood glucose values from the CGMS is a feedback signal that will be compared to a desired blood glucose value, and the error between them will be used to control the insulin pump for automatic control of the future blood glucose levels. Some goals are:
 * to control the basal insulin when it is possible to do so by increasing insulin when the blood sugar tends high and decreasing insulin when the blood sugar tends low, and denying insulin when the predicted blood sugars could be very low;
 * to adjust the bolus insulin given to cover meals to achieve a balance of insulin with carbohydrates based on the actual blood glucose levels;
 * to actually "learn" the needs of the pump user by keeping track of past success and failure of the insulin program, and using that information to automatically "adapt" the control variables to continuously monitor and improve the insulin program.
 * and someday combining multiple hormones (insulin, amylin, and glucagon) to copycat the glucose regulation of a real pancreas.

When controlling the basal rate alone, the closed loop can still correct a meal bolus error that was too large or small for the food consumed by balancing the insulin on board with the CGMS feedback. The artificial pancreas algorithm must:
 * track insulin on board after a meal bolus,
 * know how much insulin is needed to bring BG values back in range after the meal,
 * recognize an imbalance between the insulin bolused and the insulin actually needed,
 * automatically bolus or increase the basal rate to correct a shortage of insulin,
 * automatically reducing or interrupting the basal rate to correct an abundance of insulin,
 * and using adaptive filtering techniques to "learn" the carbohydrate to insulin ratios for each meal bolus.

The focus of this artificial pancreas project differs however from others: lets avoid the use of the general purpose PID controller, focusing instead on a custom approach combining modern control systems theory and real-time embedded controls. Our approach will:
 * mimic the expertise taught by a certified diabetes educator
 * use the natural states and modes of the insulin pump
 * use CGMS feedback to control the basal rate when possible
 * use CGMS feedback to learn the adaptive variables