Related Experiment Video
Updated: Oct 25, 2025
An Integrated Raman Spectroscopy and Mass Spectrometry Platform to Study Single-Cell Drug Uptake, Metabolism, and Effects
Published on: January 9, 2020
A shallow convolutional neural network with elastic nets for blood glucose quantitative analysis using Raman
Feifei Pian1, Qiaoyun Wang1, Mingxuan Wang1
1College of Information Science and Engineering, Northeastern University, Shenyang, Liaoning Province 110819, China; Hebei Key Laboratory of Micro-Nano Precision Optical Sensing and Measurement Technology, Qinhuangdao 066004, China.
This study introduces a novel one-dimensional shallow convolutional neural network with elastic nets (1D-SCNN-EN) for predicting blood glucose concentration using Raman spectroscopy. The 1D-SCNN-EN model demonstrates high accuracy and robustness, especially for small sample sizes.
Area of Science:
- Spectroscopy
- Biomedical Engineering
- Machine Learning
Background:
- Accurate blood glucose monitoring is crucial for diabetes management.
- Traditional methods for glucose measurement can be invasive or time-consuming.
- Raman spectroscopy offers a non-invasive approach for analyzing biological samples.
Purpose of the Study:
- To develop and validate a novel deep learning model for predicting blood glucose concentration.
- To assess the performance of the proposed model against conventional methods.
- To investigate the model's efficacy in scenarios with limited sample data.
Main Methods:
- Acquisition of 106 blood glucose spectra using Fourier transform (FT) Raman spectroscopy.
- Development of a one-dimensional shallow convolutional neural network integrated with elastic nets (1D-SCNN-EN).
- Comparison of the 1D-SCNN-EN model with partial least squares (PLS) and support vector machine (SVM) regression models.
Main Results:
- The 1D-SCNN-EN model achieved a root mean squared error of prediction (RMSEP) of 0.11210 and a determination coefficient of prediction (RP2) of 0.99403.
- The model demonstrated superior performance compared to PLS and SVM, with a residual predictive deviation of prediction (RPD) of 12.94601.
- The 1D-SCNN-EN model exhibited higher prediction accuracy and robustness, particularly with a small sample size.
Conclusions:
- The proposed 1D-SCNN-EN model is a promising tool for non-invasive blood glucose concentration prediction using Raman spectroscopy.
- The model effectively captures deep features and reduces complexity, outperforming conventional regression techniques.
- This approach shows significant potential for applications in diabetes management, especially when dealing with limited spectral data.
