BACKGROUND AND OBJECTIVE: This paper compares the ability of three analytic techniques to predict chromophore concentration from fluorescence emission spectra of homogeneous, turbid samples with optical properties similar to human tissue. STUDY DESIGN, MATERIALS AND METHODS: Two models of light propagation were implemented (exponential attenuation, two flux Kubelka-Munk theory); a priori information about sample optical properties was used to analyze data with the two flux Kubelka-Munk model. The third data analysis technique utilizes the method of partial least squares (PLS) to develop an empirical, linear model of sample fluorescence from a training set with optical properties and known concentrations representative of those to be predicted. This model can be applied to predict chromophore concentrations in the unknown samples. RESULTS: Of the three methods, PLS achieved the most accurate results and was able to predict fluorophore concentration to within +/- 6% of true values. CONCLUSION: We investigated conditions under which PLS predictions were most accurate and find that best results are achieved when predictions are based on fluorescence emission spectra at more than one excitation wavelength with inclusion of the tail of Rayleigh scattering at the excitation wavelength.
BACKGROUND AND OBJECTIVE: This paper compares the ability of three analytic techniques to predict chromophore concentration from fluorescence emission spectra of homogeneous, turbid samples with optical properties similar to human tissue. STUDY DESIGN, MATERIALS AND METHODS: Two models of light propagation were implemented (exponential attenuation, two flux Kubelka-Munk theory); a priori information about sample optical properties was used to analyze data with the two flux Kubelka-Munk model. The third data analysis technique utilizes the method of partial least squares (PLS) to develop an empirical, linear model of sample fluorescence from a training set with optical properties and known concentrations representative of those to be predicted. This model can be applied to predict chromophore concentrations in the unknown samples. RESULTS: Of the three methods, PLS achieved the most accurate results and was able to predict fluorophore concentration to within +/- 6% of true values. CONCLUSION: We investigated conditions under which PLS predictions were most accurate and find that best results are achieved when predictions are based on fluorescence emission spectra at more than one excitation wavelength with inclusion of the tail of Rayleigh scattering at the excitation wavelength.
Authors: Brian W Pogue; Scott C Davis; Frederic Leblond; Michael A Mastanduno; Hamid Dehghani; Keith D Paulsen Journal: Philos Trans A Math Phys Eng Sci Date: 2011-11-28 Impact factor: 4.226