CGM Meal Detection Neural Network for Automated Carbohydrate Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated insulin delivery systems require users to manually count carbohydrates and announce meals, which is challenging, inaccurate, and leads to suboptimal glycemic control due to high prevalence of postprandial hyperglycemia and hypoglycemia.
Innovation Solution
A machine-learning-based multioutput neural network using CGM and insulin data to detect meals and estimate carbohydrate content, implemented in smart devices or distributed systems, providing accurate meal detection and size estimation for improved insulin dosing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual carbohydrate counting is used, then users can control insulin dosing, but accuracy and ease of operation deteriorate leading to suboptimal glycemic control
Solution Approach 1:
The system performs automated meal detection and carbohydrate estimation using CGM glucose data and insulin delivery data without requiring user input. The algorithm independently identifies meals and estimates their carbohydrate content, allowing the system to serve itself rather than requiring manual user operation for each meal event.
Solution Approach 2:
The patent replaces the manual mechanical process of carbohydrate counting with an automated computational system. Machine learning algorithms process CGM and insulin data to detect meals and estimate carbohydrate content, substituting human cognitive effort with automated data processing and pattern recognition.
2Ease of operation
If automated meal detection algorithms are used, then user burden is reduced, but reliability deteriorates due to high false positive rates
Solution Approach 1:
The system uses CGM glucose measurements as feedback to validate meal detection events. By monitoring actual glucose responses and comparing them against predicted patterns, the algorithm can confirm or reject detected meals, reducing false positives while maintaining high automation levels.
Solution Approach 2:
The meal detection algorithm dynamically adjusts its parameters and thresholds based on individual user characteristics, glucose patterns, and meal contexts. This adaptability allows the system to optimize detection sensitivity and specificity for each user, improving reliability while maintaining automation.
3Loss of time
If meal detection occurs within 60 minutes of intake, then insulin dosing timeliness is improved, but detection difficulty increases due to glucose signal overlap
Solution Approach 1:
The system performs preliminary meal detection by identifying glucose rise patterns that precede or coincide with meal intake. By detecting meals at or near the time of intake rather than waiting for full glucose response, the system enables timely insulin dosing while using multiple data features to maintain detection accuracy despite signal overlap.
Solution Approach 2:
The algorithm moves beyond single-dimensional glucose magnitude analysis to multi-dimensional feature space, incorporating rate of glucose change, insulin on board, time of day, and historical patterns. This dimensional expansion allows accurate meal detection within 60 minutes by identifying patterns that would be indistinguishable in simpler analyses.
Data Source
AI summary
Disclosed is a meal detection and meal size estimation machine learning technology. In some embodiments, the techniques entail applying to a trained multioutput neural network model a set of input features, the set of input features representing glucoregulatory management data, insulin on board, and time of day, the trained multioutput neural network model representing multiple fully connected layers and an output layer formed from first and second branches, the first branch providing a meal detection output and the second branch providing a carbohydrate estimation output; receiving from the meal detection output a meal detection indication; and receiving from the carbohydrate estimation output a meal size estimation.