Correlation between near-infrared tissue spectra and pH, temperature, and blood flow using partial least squares

It has been shown that near-infrared spectroscopy is a feasible technique to non-invasively measure tissue pH in vivo. Since this technique relies on pH-induced changes in heme protein spectra, other factors that affect those spectra were investigated. In this study, the correlation between spectra collected from the bowel (575-1100 nm) with local tissue temperature and blood flow were investigated simultaneously with pH changes during eight independent swine hemorrhagic shock experiments. Using Partial Least-squares (PLS) and cross-validation techniques to correlate changes in spectra to measured changes in the above parameters, we observed good pH correlation results. The same time points were then used to build PLS calibration models for temperature, and now. Separate models for ischemia and reperfusion periods were calculated for each experiment. Average correlation coefficients (R-2) and standard error of prediction (SEP) are reported for models that had predictive ability. Blood flow was poorly correlated with pH, less than half of the data could be successfully modeled. Temperature showed some correlation with pH and PLS models could be built for most of the reperfusion data. This analysis affirms the ability to develop NIR pH calibration models based solely oh pH reference measurements, however temperature correction may provide improved accuracy when large temperature changes occur.