Noninvasive determination of hemoglobin and hematocrit using a temperature-controlled localized reflectance tissue photometer

We performed visible/near-infrared optical measurements on the forearm of human subjects. We conducted four studies: one study using a commercial diffuse reflectance spectrometer, and three studies using a breadboard temperature-controlled localized reflectance tissue photometer. Calibration relationships were established between skin reflectance signal and either the reference blood hemoglobin (Hb) concentration or the hematocrit values (Hct). Prediction results were expressed as the prediction correlation coefficient (r(p)) and the standard error for cross-validation prediction (CV-SEP). Using diffuse reflectance measurement, r(p) = 0. 8, CV-SEP = 0.9 g/dL for Hb and r(p) = 0.7, CV-SEP = 3.3% for Hct (n = 40). In a localized reflectance study involving calculating the absorption and scattering coefficients and including effect of change of skin temperature in the calibration model, the best prediction results were r(p) = 0.9, CV-SEP = 0.8 g/dL for Hb and r(p) = 0.8, CV-SEP = 3% for Hct (n = 26). In a second localized reflectance study on a population having diverse skin colors at 34 degrees C, the optimal model led to r(p) = 0.8, CV-SEP = 0.9 g/dL for Hb and r(p) = 0.9, CV-SEP = 2.1% for Hct (n = 28) using the localized reflectance values without deducing the absorption and scattering coefficients. Improvement in the correlation was more noticeable for Hct than for the case of Hb. The photometer was used to screen prospective blood donors with low Hb concentration. It was possible to predict anemic subjects in the limited prospective blood donor population.