Kalman Filter On-Off Switch for Insulin Pump
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
People with type 1 diabetes face challenges in managing blood glucose levels, particularly at night due to ineffective alarms during hypoglycemia, and there is a need for a method to prevent or mitigate low blood glucose levels caused by excessive insulin delivery.
Innovation Solution
A Kalman filter-based automatic on-off switch system for insulin pumps that uses continuous glucose monitor readings to predict low blood glucose levels, suspending or resuming insulin delivery based on set safety rules, including time windows and activity thresholds, to prevent hypoglycemia and hyperglycemia.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If alarms are used to alert patients about low blood glucose levels, then patients can be warned of hypoglycemia, but alarms are ineffective during nighttime due to torpor caused by hypoglycemia
Solution Approach 1:
The system performs preliminary action by predicting low blood glucose levels before they occur using the Kalman filter to analyze glucose trends. This predictive capability allows the system to suspend insulin delivery in advance, preventing hypoglycemia before it happens, rather than merely alarming after it occurs. The safety rules evaluate multiple conditions including predicted glucose levels, rate of change, and time windows to ensure timely preventive action.
Solution Approach 2:
The system implements self-service by automatically suspending and resuming insulin delivery based on predicted glucose levels without requiring patient intervention. The automated on-off switch monitors glucose trends continuously and makes independent decisions to adjust insulin delivery, eliminating the need for patient response to alarms and providing reliable protection even during nighttime torpor.
2Reliability
If continuous monitoring and automatic pump control are implemented, then hypoglycemia can be prevented, but system complexity increases
Solution Approach 1:
The system uses feedback by continuously monitoring glucose levels through CGM readings and using the Kalman filter to process this data. The filter incorporates process noise and measurement noise models to accurately estimate current glucose levels and trends. This feedback loop enables reliable prediction of future glucose levels and automatic adjustment of insulin delivery based on actual physiological responses.
Solution Approach 2:
The system applies parameter changes by dynamically adjusting insulin delivery rates based on predicted glucose levels. The Kalman filter estimates glucose state parameters including current level, rate of change, and curvature. Safety rules evaluate multiple parameters such as predicted glucose thresholds, time window durations, and maximum shut-off times to make informed decisions about pump control, optimizing the balance between preventing hypoglycemia and avoiding unnecessary interruptions.
3Reliability
If insulin delivery is suspended to prevent hypoglycemia, then patient safety is improved, but risk of hyperglycemia increases if suspended too long
Solution Approach 1:
The system implements dynamics by making insulin delivery adjustable and responsive to real-time glucose conditions. The automated on-off switch dynamically suspends or resumes insulin delivery based on continuously evaluated safety rules and predicted glucose levels. This dynamic control allows the system to adapt to changing physiological states, preventing both hypoglycemia and prolonged hyperglycemia by responding flexibly to actual glucose trends rather than using fixed delivery schedules.
Data Source
AI summary
Techniques for controlling an insulin pump include determining values for parameters selected from a group including a first prediction time horizon, a predicted glucose threshold (Goff) for turning the insulin pump off, a maximum shut off time within a time window, and duration of the time window. A safety rule is determined based on the maximum shut off time within the duration. Glucose readings are collected up to a current time. An expected current glucose value G and glucose temporal rate of change are determined based only on the glucose readings and a Kalman filter configured for noisy glucose readings. A glucose level (Gh1) is predicted for a future time that is the prediction time horizon after the current time. A command is issued to shut off the insulin pump if it is determined both that Gh1 is less than Goff and that the safety rule is satisfied.