Multi-Head Glucose Prediction Network for CGM Gaps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Continuous glucose monitoring devices may malfunction or be removed, leaving users without a consistent way to monitor blood glucose levels, necessitating a reliable method for estimating glucose values.
Innovation Solution
A multi-head temporal convolutional neural network (MH-TCNN) analyzes past glucose levels and biometric data to predict future glucose values, using separate convolution layers for different data inputs and combining them through a concatenated layer for accurate prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a continuous glucose monitoring device is used to monitor blood glucose levels, then measurement precision is improved, but device reliability deteriorates when the device malfunctions or is removed
Solution Approach 1:
The patent creates a virtual copy of the CGM device functionality through a machine learning model that replicates glucose level prediction capabilities. The MH-TCNN model processes alternative biometric data (heart rate, temperature, activity) to generate predicted glucose values, effectively copying the monitoring function without requiring the physical CGM device to be present or functioning.
Solution Approach 2:
The patent introduces an intermediary system that translates alternative biometric measurements into glucose level predictions. When CGM data is unavailable, the MH-TCNN model acts as a mediator, converting data from other sensors (heart rate monitors, temperature sensors, activity trackers) into meaningful glucose estimates, thereby maintaining monitoring reliability through indirect measurement.
2Measurement precision
If a multi-headed temporal convolutional neural network is used to predict glucose values, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent segments the glucose prediction task into multiple specialized sub-tasks, each handled by a dedicated convolutional head in the MH-TCNN model. Different heads process different types of biometric data (heart rate, temperature, activity) independently, allowing each to be optimized for its specific data type while contributing to the overall prediction, thereby managing complexity through functional decomposition.
Solution Approach 2:
The patent creates a universal model architecture that handles multiple data types and prediction tasks through a single multi-headed network. The shared base layers of the MH-TCNN process common temporal patterns across all biometric inputs, while the specialized heads handle type-specific transformations, achieving multi-functionality that reduces overall system complexity compared to separate models for each data type.
Data Source
AI summary
Systems and methods for determining a glucose value for a user are disclosed herein. The method includes receiving a plurality of data inputs associated with biometric data of the user, the plurality of data inputs including at least one data input representative of a past estimated glucose value of the user and processing the plurality of data inputs with a multi-headed temporal convolutional neural network to generate a blood glucose value for the user. The method also includes providing a notification to the user based at least in part on the blood glucose value.