Stacked Glucose Prediction Models for Event-Aware CGM Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional glucose prediction systems fail to accurately predict future glucose levels due to their reliance on historical data alone, neglecting the impact of events such as meals, exercise, and insulin administration, leading to inaccurate predictions that can result in dangerous health conditions.
Innovation Solution
A stacked configuration of multiple machine learning models is employed, including neural networks and state machines, to generate glucose and event predictions, utilizing historical glucose measurements and additional data from continuous glucose monitoring systems, with confidence-based input selection to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional glucose prediction systems use only historical glucose measurements as input, then the system complexity is low, but the prediction accuracy deteriorates due to inability to account for events like meals, exercise, and insulin administration
Solution Approach 1:
The prediction system is segmented into multiple independent machine learning models, each specialized for predicting specific glucose-impacting events (meals, exercise, insulin administration). This segmentation allows each model to focus on specific patterns while collectively providing comprehensive event detection, resolving the contradiction between accuracy and complexity by distributing computational tasks across specialized components rather than using a single complex model
Solution Approach 2:
The machine learning models are designed with multi-functionality to detect various types of glucose-impacting events (meals, exercise, insulin administration) using a unified architecture. This universal approach allows the system to handle diverse event types without requiring separate specialized systems for each event type, maintaining manageable complexity while achieving comprehensive prediction accuracy
2Measurement precision
If a CGM platform processes enormous amounts of continuous glucose measurement data to identify patterns, then the prediction accuracy improves, but the data processing complexity becomes impossible for human operators
Solution Approach 1:
The patent replaces human manual data processing with automated machine learning models that can efficiently analyze enormous volumes of continuous glucose measurement data. The machine learning algorithms automatically identify patterns and predict glucose-impacting events without human intervention, resolving the contradiction by substituting mechanical human analysis with computational automation capable of handling large-scale data processing
Solution Approach 2:
The system uses historical glucose measurement patterns as training data to create predictive models that replicate expert pattern recognition capabilities. By copying and learning from historical data patterns, the machine learning models can identify recurring glucose-impacting event patterns without requiring human operators to manually analyze each data point, achieving high accuracy while maintaining manageable operational complexity
Data Source
AI summary
Glucose measurement and glucose-impacting event prediction using a stack of machine learning models is described. A CGM platform includes stacked machine learning models, such that an output generated by one of the machine learning models can be provided as input to another one of the machine learning models. The multiple machine learning models include at least one model trained to generate a glucose measurement prediction and another model trained to generate an event prediction, for an upcoming time interval. Each of the stacked machine learning models is configured to generate its respective output when provided as input at least one of glucose measurements provided by a CGM system worn by the user or additional data describing user behavior or other aspects that impact a person's glucose in the future. Predictions may then be output, such as via communication and/or display of a notification about the corresponding prediction.