Physiology-independent control algorithm for an artificial pancreas

An insulin pump control system that works automatically without user input

Problem:

Diabetic patients can use insulin pumps to deliver insulin after a meal. Continuous glucose monitoring (CGM) systems are also used to record glucose levels throughout the day. These systems could be unified via a closed-loop control algorithm to create an artificial pancreas that requires no input, manual blood testing, or manual dosing by the user. However, this is a difficult problem to solve because of the physical variation among individuals, the difficulty of quantifying nutrition, and the risks of serious outcomes if the incorrect dosage of insulin is given.

Solution:

Researchers at the University of Pennsylvania have developed a closed-loop control algorithm to enable an artificial pancreas system for Type 1 Diabetes patients. This algorithm is independent of the user’s physiology and requires no user input about nutrition. The algorithm uses past information about insulin doses provided by the pump and glucose data provided by the CGM. From only this data, the system is trained to accurately guess when a meal has taken place and what the appropriate magnitude insulin dose should be, without any user input about the timing or content of the meal itself. The new algorithm was tested using an FDA-accepted diabetes physiological model with varying patient parameter settings resulting in a 99.6% correct detection rate, outperforming other models.

Advantages:

  • Accurate meal prediction without user input
  • Improved patient quality of life

Stage of Development:

Prototype software tool

Intellectual Property:

US 10,792,423

Reference Media:

Towards a Model-Based Meal Detector for Type I Diabetics

Desired Partnerships:

  • License
  • Co-development

Docket # 16-7626

Patent Information: