Performance Comparison of Several Published Tissue Near-Infrared Spectroscopy Algorithms

We have collected multiwavelength near-infrared (NIR) attenuation spectra on human forearm muscle, the adult rat head, and newborn piglet head to compare the changes in chromophore concentration derived from these data using published algorithms from four groups. We find differences between the results from the algorithms on each data set, particularly in their estimation of cytochrome oxidase (cyt- aa 3) redox changes. We also find some differences when applying the same algorithm to the three data sets, suggesting possible difficulties in transferring algorithms between different physiological systems (e.g., Kurth, C. D., Steven, J. M., Benaron, D., and Chance, B. (1993) J. Clin. Monit. 9, 163-170). We have also compared the algorithms using simulated data generated using measured hemoglobin absorption spectra and a diffusion model for light transport in tissue. We find that while the algorithms from three groups are in broad agreement, that published by Piantadosi (Piantadosi, C. A. (1993) Methods Toxicol. 2, 107-126) produces significantly different results for cyt- aa 3 and HbO 2. Either the hemoglobin spectra used to produce the simulated data are inaccurate or the modeling is incorrect, or this algorithm is erroneous.