Deep Learning Blood Glucose Prediction for Type 1 Diabetes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for predicting blood-glucose concentrations in individuals with type 1 diabetes lack accuracy, necessitating improved methods for effective glucose intervention regulation.
Innovation Solution
A system utilizing a blood-glucose monitoring device with a sensor, processor, and memory that accesses data on meal intake, insulin doses, physical activity, and blood-glucose levels, employing a deep learning network to predict glucose concentrations and determine an insulin administration schedule, which can be adjusted in real-time based on detected glucose levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current prediction techniques are used, then the system is simple to operate, but the prediction accuracy of blood-glucose levels is insufficient
Solution Approach 1:
The system segments the prediction task into multiple components: data collection module (meal intake, insulin doses, physical activity, blood-glucose levels), deep learning network module (prediction engine), and output module (future blood-glucose concentration predictions at multiple time intervals). This segmentation allows each component to be optimized independently while maintaining overall system manageability.
Solution Approach 2:
The patent introduces a deep learning network as an intermediary between the input data (meal intake, insulin doses, physical activity, blood-glucose levels) and the output predictions. This intermediary processes the complex relationships between multiple variables to generate accurate predictions, resolving the contradiction by adding computational intelligence without requiring direct complex hardware modifications.
2Measurement precision
If deep learning network is implemented, then the prediction accuracy improves, but the computational resources and processing time increase
Solution Approach 1:
The system performs preliminary data collection and processing by gathering meal intake, insulin dose, physical activity, and blood-glucose level data before running the deep learning prediction. This preparation allows the prediction model to process pre-organized data, reducing actual prediction time while maintaining accuracy.
Solution Approach 2:
The system generates predictions at periodic time intervals (e.g., 30 minutes, 60 minutes, 90 minutes ahead), allowing computational resources to be distributed across multiple prediction points rather than requiring continuous high-computation processing. This periodic approach balances accuracy with processing efficiency.
Data Source
AI summary
A system for predicting blood-glucose concentration includes a blood-glucose monitoring device including a sensor configured to capture blood-glucose levels, a processor, and a memory. The memory includes instructions, which, when executed by the processor, cause the system to: access data of meal intake for a person with type 1 diabetes, data of insulin doses administered to the person, data corresponding to a physical activity of the person, and data of blood-glucose levels in the person, captured by the sensor; predict by a deep learning network blood-glucose concentration levels for the person at predetermined time intervals based on the accessed data of meal intake, data of insulin doses, data corresponding to the physical activity, and data of blood-glucose levels in the person; and determine an insulin administration schedule based on the predicted blood-glucose concentration levels for the person, the insulin administration schedule including a dosing schedule.