Temporal Predictive Model for Glycemia Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current closed-loop systems for managing blood glucose levels in diabetes patients struggle with predicting future glycemia due to their inability to account for a patient's general glycemic behavior influenced by habits, leading to inaccurate insulin dosing and potential hypoglycemia or hyperglycemia.
Innovation Solution
A device and method utilizing a temporal predictive model that processes time-based data and encoded temporal data to learn glycemic cyclic behavior over a given cyclic period, incorporating patient habits and influencing factors like insulin on board and carbohydrate intake, to accurately predict future glycemia levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If time-based data is used to predict future glycemia in closed-loop systems, then the system can calculate insulin quantity, but the prediction accuracy is insufficient because patient habits and cyclic glycemic behavior are not considered
Solution Approach 1:
The system performs preliminary encoding of temporal information representing patient habits and cyclic glycemic behavior patterns before using them in prediction. This preprocessing step allows the model to incorporate contextual information about typical glycemic variations at different times of day, days of the week, and seasonal patterns, thereby improving prediction accuracy without excessively increasing model complexity during the actual prediction process
Solution Approach 2:
The system transforms temporal data into encoded representations that capture cyclic patterns. By changing the parameter representation from raw time values to encoded temporal features that reflect habitual glycemic behavior, the model can better predict future glycemia levels while maintaining a manageable complexity level through efficient feature encoding
2Measurement precision
If encoded temporal data representing patient habits is incorporated into the predictive model, then prediction accuracy improves, but the data processing complexity increases
Solution Approach 1:
Temporal data encoding is performed in advance to capture patient-specific habitual patterns. This preliminary processing transforms raw temporal information into meaningful features that represent typical glycemic behavior at different times and conditions, improving prediction accuracy while reducing the computational burden during real-time prediction by preparing features beforehand
Solution Approach 2:
The system creates encoded representations (copies) of temporal data that capture essential patterns without requiring the full complexity of original time-series data during prediction. These encoded copies efficiently represent habitual glycemic behavior, allowing accurate predictions with reduced processing complexity
Data Source
AI summary
A device for determining a predicted value of glycemia, the device including a data processing unit (3) adapted to process time-based data. The data processing unit (3) is configured to implement a temporal predictive model trained to learn a glycemia cyclic temporal behaviour, where said glycemia cyclic temporal behaviour is defined by variations of glycemia values in a cyclic way over a given cyclic period of time. The temporal predictive model is configured to: receive, as inputs:said time-based data; andencoded temporal data, said encoded temporal data being encoded as to represent the given cyclic period of time. The temporal predictive model is also configured to deliver, as output, a predicted value of glycemia.