Performance improved method for subtracted blood volume spectrometry using empirical mode decomposition.

Subtracted blood volume spectrometry (SBVS) can eliminate the background information in near infrared spectroscopy (NIRS) noninvasive biochemical sensing. However, the spectrum obtained by this method is accompanied by serious noises which are to the disadvantage of the calibration models. Empirical mode decomposition (EMD) was applied to restrict the noises in order to improve the performance of subtracted blood volume spectrometry. Certain criteria were used to evaluate the performance of the method, such as the average correlation coefficient, and the average and standard deviation of the Euclidean distance. EMD was applied to three subtracted spectra with different ΔL, and the criteria were calculated accordingly. All of the criteria were improvement. Especially for the subtracted spectra with ΔL=0.5mm, the correlation coefficient increased from 0.9970 to 0.9999, the average Euclidean distance decreased from 0.0265 to 0.0118, and the standard deviation of the Euclidean distance decreased from 0.0148 to 0.0033 after EMD filtering. The PLS models of the processed spectra were promoted as well. These preliminary results suggest that EMD is a promising means of improving the performance of subtracted blood volume spectrometry.