Spectral data quality assessment based on variability analysis: application to noninvasive hemoglobin measurement by dynamic spectrum

• Published in: Analytical methods Vol. 7; no. 13; pp. 5565 - 5573
• Main Authors: He, Wenqin, Li, Xiaoxia, Wang, Mengjun, Li, Gang, Lin, Ling
• Format: Journal Article
• Language: English
• Published: 01.01.2015
• Subjects: blood; Coefficients; Computer simulation; data quality; Dynamic tests; Hemoglobin; Mathematical analysis; Quality assessment; quantitative analysis; simulation models; Spectra; spectral analysis; Stability
• ISSN: 1759-9660; 1759-9679; 1759-9679
• DOI: 10.1039/c5ay00669d

The quality of spectral data is crucial to the accuracy of quantitative spectral analysis, especially the noninvasive measurement of blood components. As an important part of research, establishing an effective and reliable quality assessment metric to select data before modelling is indispensable. According to the principle of Dynamic Spectrum (DS) and the characteristics of the photoplethysmogram (PPG), a novel method to assess the spectral quality - stability coefficient (SC) - is proposed in this study and we propose an analytical formula. To verify the feasibility, in simulation analysis, we calculated the stability coefficient of simulated spectra and evaluated the performance of extraction by the Root Mean Square Error (RMSE). The result shows a negative correlation between the stability coefficient and RMSE. After simulation analysis, we conducted a control experiment based on data from 427 subjects by developing calibration models between the DS data and hemoglobin concentration. The average correlation coefficient is 0.875 in the test set of the experimental group, while that of the control group is only 0.715. The actual experimental result is consistent with the simulation analysis, which demonstrates that the assessment method can evaluate the quality of spectral data efficiently and accurately. This new quantitative method provides a reliable way to assess and screen spectral data. It could be applied not only to the noninvasive measurement of blood components but also to other related fields such as spectral analysis, and signal measurement and processing. It could be applied not only to the noninvasive hemoglobin measurement but also to other related fields such as signal processing.