Broadening the bands for improving the accuracy of noninvasive blood component analysis

Abstract

Signal to noise ratio (SNR) of the spectrum is the overriding factor that limits the accuracy of noninvasive blood component measurement based on spectral analysis. In this paper, we applied the multi-dimensional spectral measurement strategy of the “M+N” theory to the quantitative blood component analysis based on the dynamic spectrum (DS) theory, in the form of increasing the wavelength range of the collected spectrum, and discussed the effects of spectral difference coefficient on measurement accuracy, to investigate the effects of wavelength range on the accuracy. We collected photoplethysmogram (PPG) signals across a broad range (313–1388 nm) using a spectrometer system consisting of multiple light sources and spectrophotometers covering different wavelength range assembled on-site. Using DS data obtained with the optimizing differential extraction method, we established calibration models for hemoglobin concentration. The new calibration model showed that compared with the model used the DS under the generally used range (550–1050 nm), the root mean square error of calibration set (RMSEC) decreased from 10.154 g/L to 5.509 g/L, and the correlation coefficient of calibration set (Rc) increased from 0.655 to 0.912. This study shows that the acquisition of signals from a wider wavelength range can improve the accuracy of noninvasive blood component measurement.