| abstract |
A method is provided for predicting a property of a matter of biological origin, such as biological fluid, containing water, in a dynamic condition where the biological fluid may be approximated to contain two compartments where one compartment has a proportionally larger or smaller amount of water than the other compartment having the property of interest. The method involves establishing a training set in the near-infrared (NIR) region with independent quantification of the property of the fluid using known techniques. The training set is mathematically analyzed according to a correlation developed by regression analysis after employment of a pre-processing technique such as a multiple derivative transformation of spectra or a ratioing of two wavelengths in the spectra. The result is a mathematical transformation equation which quantitatively relates spectral intensities at specific wavelengths to the property of interest. This transformation equation may be applied to unknown samples so as to predict their properties, thereby eliminating need for the reference method except for validation or recalibration. The method provides rapid and accurate prediction of the property of the unknown sample, which may be the property of hematocrit or hemoglobin concentration in whole animal blood. Other analyses of properties in the biological fluid such as oxygen saturation in hemoglobin in whole animal blood may be included in the mathematical analysis to further refine the prediction of the property of interest. Also, a loop from the patient is disclosed for the purpose of monitoring the property of interest nearly simultaneously with changes in that property of interest. |