Blood pH is an important indicator of anaerobic metabolism in exercising muscle. This paper demonstrates multivariate calibration techniques that can be used to produce a general pH model that can be applied to spectra from any new subject without significant prediction error. Tissue spectra (725 ∼ 880 nm) were acquired through the skin overlying the <i>flexor digitorum profundus</i> muscle on the forearms of eight healthy subjects during repetitive hand-grip exercise and referenced to the pH of venous blood drawn from a catheter placed in a vein close to the muscle. Calibration models were developed using multi-subject partial least squares (PLS) and validated using subject-out cross-validation after the subject-to-subject spectral variations were corrected by mathematical preprocessing methods. A combination of standard normal variate (SNV) scaling and principal component analysis loading correction (PCALC) successfully removed most of the subject-to-subject variations and provided the most accurate prediction results.
Vol. 2, Iss. 3 Virtual Journal for Biomedical Optics
Ye Yang, Olusola O. Soyemi, Michelle R. Landry, and Babs R. Soller, "Noninvasive in Vivo Measurement of Venous Blood pH During Exercise Using Near-Infrared Reflectance Spectroscopy," Appl. Spectrosc. 61, 223-229 (2007)