PPG Neural Network Glucose Prediction for Population-Specific Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing non-invasive blood glucose measurement technologies using photoplethysmography (PPG) lack accuracy and universality due to variations among individuals, particularly those undergoing medical treatments affecting the cardiovascular system, limiting their applicability in a wide population.
Innovation Solution
A PPG-based non-invasive blood glucose prediction system utilizing a neural network trained on data from individuals not undergoing medical treatments that affect the cardiovascular system, incorporating PPG signals and HbA1c measurements to predict blood glucose levels accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a single universal neural network model is used for NIBG prediction across all populations, then the device complexity is reduced and ease of operation is improved, but the measurement precision and reliability deteriorate due to undocumented variations among different population groups
Solution Approach 1:
The patent divides the population into distinct cohorts based on cardiovascular health status and medical treatment history. Separate neural network models are trained for each cohort (e.g., healthy individuals vs. those with cardiovascular conditions), allowing each model to specialize in the specific characteristics of its target population, thereby improving measurement precision without significantly increasing operational complexity.
Solution Approach 2:
The patent applies different modeling approaches and training data characteristics to different population segments. For example, the model for individuals with cardiovascular conditions uses training data specifically from similar populations, while the healthy population model uses data from normocardiac individuals. This localized optimization ensures each model has the quality and characteristics needed for its specific application context.
2Adaptability or versatility
If training data includes individuals undergoing medical treatments affecting the cardiovascular system, then the adaptability of the model to a wider population is improved, but the measurement precision deteriorates due to treatment-induced variations in PPG signals
Solution Approach 1:
The patent segments the training population into distinct groups based on medical treatment status and cardiovascular health. Separate neural network models are developed for each segment (e.g., one model for individuals on cardiovascular medications, another for those without such treatments). This segmentation allows each model to learn the specific PPG signal characteristics associated with its target population, maintaining high measurement precision while achieving broad adaptability through multiple specialized models.
Solution Approach 2:
The patent changes the training parameters and data characteristics based on the target population. For individuals undergoing medical treatments, the training data includes subjects with similar treatment profiles, and the model is trained with adjusted parameters to account for treatment-induced physiological variations. This parameter adaptation allows the model to maintain precision across different population groups.
3Device complexity
If PPG signals from all subjects are processed uniformly, then the device complexity is minimized, but the reliability of blood glucose prediction deteriorates due to individual variations in cardiovascular physiology
Solution Approach 1:
The patent implements a dynamic model selection mechanism that adapts the processing approach based on the subject's characteristics. The system first assesses the subject's cardiovascular health status and treatment history, then dynamically selects or configures the appropriate neural network model. This dynamic adaptation allows the system to maintain high reliability by matching the processing method to the subject's physiological characteristics, while keeping the overall device complexity manageable through automated model selection.
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 provides accurate and reliable non-invasive blood glucose predictions by accounting for individual variations, excluding subjects and training data from those with cardiovascular-altering treatments, enhancing prediction accuracy.
Implementation Method 1
PPG device measures the changes in transmittance or reflectance of near-infrared when blood passes through peripheral capillaries
Implementation Method 2
the light absorption and reflectance of specific wavelengths are sensitive to the body's hemodynamic properties
Data Source
AI summary
The PPG based NIBG neural network prediction system of the present invention comprises a neural network configured to predict BG level of a subject based on PPG signal obtained from the subject wherein the subject is not undergoing medical treatment and the neural network is trained using training data from subjects not undergoing medical treatment. In another embodiment, the PPG based NIBG neural network prediction system of the present invention predicts BG level of a subject based on HbA1c of the subject measured using conventional finger prick method as well as PPG signal obtained from the subject wherein the subject is not undergoing medical treatment and the neural network is trained using training data from subjects not undergoing medical treatment.