Glucose Event Detection Using Recurrent Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current glucose level monitoring technologies lack effective solutions for detecting glucose events and trends using noninvasive sensors and wearables, which are essential for managing diabetes, a growing global health crisis.
Innovation Solution
A glucose event detection system employing an artificial recurrent neural network (RNN) architecture with long short-term memory (LSTM) cells that predicts blood glucose levels based on heart beats per minute input streams, utilizing a supervised training mechanism and photoplethysmogram data to generate accurate predictions and notifications for glucose events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If noninvasive sensors and wearables are used for glucose monitoring, then patient comfort and ease of use are improved, but measurement precision and reliability of glucose event detection deteriorate
Solution Approach 1:
The patent uses heart rate data from noninvasive wearables as an intermediary signal to indirectly infer glucose events. Instead of directly measuring glucose, the system detects correlations between heart rate patterns and glucose events, using the heart rate as a mediator that bridges the gap between noninvasive sensing and glucose event detection.
Solution Approach 2:
The patent replaces direct mechanical/chemical glucose sensing with an optical/electrical heart rate monitoring system. By substituting the direct glucose measurement mechanism with a cardiovascular response measurement, the system achieves noninvasive operation while maintaining detection capability through physiological correlations.
2Measurement precision
If iterative supervised training with loss function updates is implemented, then prediction accuracy is improved, but computational complexity and training time increase
Solution Approach 1:
The patent applies partial action by implementing iterative training with a fixed number of epochs rather than exhaustive training. The system performs a sufficient number of training iterations to achieve acceptable accuracy thresholds while avoiding excessive computational expenditure, balancing precision gains against complexity costs.
Solution Approach 2:
The patent implements feedback mechanisms through loss function calculations that guide weight updates in the neural network. The loss function provides continuous feedback on prediction accuracy, enabling the system to iteratively improve performance by adjusting parameters based on measured errors between predicted and actual glucose events.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system effectively detects glucose events by analyzing RNN model output streams, reducing error through iterative training and loss function updates, and mitigates false positives with accelerometer data, providing timely and accurate notifications for managing diabetes.
Implementation Method 1
The glucose event detection system, in accordance with embodiments, may include a photoplethysmogram (PPG) based device for monitoring the heart rate of the subject
Data Source
AI summary
Disclosed are methods, circuits, devices, systems and functionally associated machine executable code for glucose event detection. A system for glucose event detection includes a recursive neural network (RNN) model for generating, for a monitored subject, blood glucose level (BGL) output streams for respective, system fed, heart beats per minute (BPM) input streams of a monitored subject. A supervised training mechanism, for training the artificial recurrent neural network (RNN) model, compares model generated blood glucose level (BGL) output streams to time-aligned blood glucose level (BGL) output streams from a continuous glucose monitoring (CGM) device concurrently monitoring the same subject.