NIR Glucose Prediction via Binary Classification and Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Glucose monitoring using Near-Infrared (NIR) spectroscopy is challenging due to the low absorption values of glucose compared to other constituents, leading to distorted data and noise, making it difficult to predict blood glucose levels with high accuracy.
Innovation Solution
A method involving the extraction of a feature set from NIR glucose spectra using machine learning regression, which includes identifying glucose-dependent features, removing collinearity, and employing a binary classification and regression model to predict glucose values, utilizing an artificial neural network framework for classification and regression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If NIR spectroscopy is used for glucose monitoring, then non-invasive measurement is achieved, but measurement precision deteriorates due to low glucose absorption values and noise
Solution Approach 1:
The patent divides the continuous glucose monitoring problem into discrete segments by classifying glucose levels into different ranges (e.g., hypoglycemic, normoglycemic, hyperglycemic). This segmentation allows the model to focus on specific glucose level characteristics, improving prediction accuracy within each segment while maintaining non-invasive measurement capabilities.
Solution Approach 2:
The patent transforms the raw NIR spectral data by changing parameters such as wavelength selection, absorption coefficient calculations, and signal processing parameters. By optimizing these parameters and focusing on specific wavelength regions where glucose absorption is more prominent, the method enhances the detectability of glucose signals against the background noise of other tissue constituents.
2Device complexity
If conventional regression models are used on raw NIR data, then prediction process is simple, but measurement precision deteriorates due to collinearity and noise in the data
Solution Approach 1:
The patent extracts relevant features from the raw NIR spectral data by identifying and isolating the most informative wavelength regions and absorption characteristics. This feature extraction process removes collinear and noisy components while retaining the essential glucose-related information, thereby improving prediction accuracy without requiring overly complex models.
Solution Approach 2:
The patent performs preliminary data processing and feature extraction before applying the final regression or classification model. Steps such as baseline correction, noise filtering, and feature selection are executed in advance to prepare the data, which significantly improves the performance of subsequent simpler prediction models.
3Quantity of substance
If all NIR spectral features are used for prediction, then data utilization is maximized, but measurement precision deteriorates due to inclusion of non-glucose related constituents
Solution Approach 1:
The patent applies local quality by focusing analysis on specific wavelength regions and spectral features that are locally more sensitive to glucose concentration changes. Rather than treating all spectral data uniformly, the method identifies and emphasizes local characteristics (specific absorption bands) that provide glucose-specific information while minimizing interference from other tissue constituents.
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
This approach enables accurate non-invasive prediction of blood glucose levels by segregating data into binary classes and generating regression models for each class, improving the accuracy of glucose monitoring and overcoming noise issues in NIR data.
Implementation Method 1
The NIR spectroscopy has been used for continuous glucose monitoring, in which NIR waves are generated to pass through the skin and a spectrum indicating absorption of the NIR waves by the blood underneath the skin is used in determining the glucose level.
Implementation Method 2
The absorption of the NIR waves is defined by BEER-Lambert law: A=εCd Where ε is an absorption co-efficient, C is a concentration of a component in sample, and d is a penetration depth.
Data Source
AI summary
A method for predicting a blood glucose level using a near-Infrared (NIR) spectrometer is provided. The method may include obtaining a feature set from an NIR glucose spectra; and predicting glucose values from the feature set based on a binary classification of the NIR glucose spectra and an in-class prediction of glucose using Machine Learning Regression.