Investigation of noninvasive in vivo blood hematocrit measurement using NIR reflectance spectroscopy and partial least-squares regression

Hematocrit (Hct), the volume percent of red cells in blood, is monitored ro utinely for blood donors, surgical patients, and trauma victims and require s blood to be removed from the patient. An accurate, noninvasive method for directly measuring hematocrit on patients is desired for these application s. The feasibility of noninvasive hematocrit measurement by using near-infr ared (NIR) spectroscopy and partial least-squares (PLS) techniques was inve stigated, and methods of iii vivo calibration were examined. Twenty Caucasi an patients undergoing cardiac surgery on cardiopulmonary bypass were rando mly selected to form two study groups. A fiber-optic probe was attached to the patient's forearm, and NIR spectra were continuously collected during s urgery, Blood samples were simultaneously collected and reference Hct measu rements were made with the spun capillary method, PLS multivariate calibrat ion techniques were applied to investigate the relationship between spectra l and Hct changes. Single patient calibration models were developed with go od cross-validated estimation of accuracy (similar to 1 Hct%) and trending capability for most patients. Time-dependent system drift, patient temperat ure, and venous oxygen saturation were not correlated with the hematocrit m easurements. Multi-subject models were developed for prediction of independ ent subjects. These models demonstrated a significant patient-specific offs et that was shown to be partially related to spectrometer drift. The remain ing offset is attributed to the large spectral variability of patient tissu e, and a significantly Larger set of patients would he required to adequate ly model this variability. After the removal of the offset, the cross-valid ated estimation of accuracy is 2 Hct%.