Determination of oxygen saturation and hematocrit of flowing human blood using two different spectrally resolving sensors.

The blood parameters oxygen saturation and hematocrit were determined by two different spectral sensors using reflectance spectra from 550 to 900 nm and partial transmission spectra centered at 660 nm. The spectra were analyzed by the method of partial least squares. One sensor consists of a miniature integrating sphere, while the other was fiber-guided. The results show that the geometry of the sensors and different blood flows do not influence the spectral analysis significantly. Independent of the sensor geometry, both hematocrit and oxygen saturation could be determined with an absolute predicted root mean square error of less than 3%. Furthermore, the analysis showed that hematocrit prediction requires eight wavelength regions and oxygen saturation prediction requires four wavelength regions using reflectance spectroscopy. This implies that if the measurement is restricted to reflectance, a spectrometer is indispensable for determining both blood parameters. Hematocrit determination could be improved using reflectance measurements in combination with transmission.