[Noninvasive detection of hematocrit and the mean corpuscular hemoglobin concentration levels by Vis-NIR...

Hematocrit (HCT) and mean hemoglobin concentration(MCHC) play a very important role in preventing cardiovascular disease and anemia. A method was developed on the basis of spectroscopy to detect HCT and MCHC non-invasively and accurately. The anatomical study showed that the blood rheology abnormalities and blood viscosity's changes can cause the changes of tongue, so there is a certain correlation between tongue and blood components. Reflectance spectrums from the tongue tips of 240 volunteers were collected, then the tongue pictures were captured and the biochemical analysis results were recorded at the same time. The 240 samples were separated into two parts: calibration sample and test sample. Spectra were then subjected to a partial least squares regression (PLSR) analysis to develop mathematics models for predicting HCT levels. The correlation between the data and prediction of HCT and MCHC yielded calibration samples value of 0.998 and 0.938. HCT and MCHC levels of test samples predicted by this model from Visible-Near infrared spectra provided a coefficient of determination in prediction of 0.979 and 0.883 with an average relative error of prediction of 1.65% and 1.88%, a root mean square error of prediction of 4.066 and 4.139. From the experiment results we can see that the model which was built before can better predict the HCT and MCHC, and the results also showed that spectrometry method may provide a promising approach to the noninvasive measurement of human HCT and MCHC with a combination of PLSR analysis.