Determination of optical properties in double integrating sphere measurement by artificial neural network based method

Abstract

An accurate inversion technique in double integrating sphere (DIS) measurement is essential for determining the optical properties of biological tissue. Although there are several established techniques, the computational time and complexity for spectral analysis require some approximations of the anisotropy factor g and refractive index n. We aim to demonstrate an artificial neural network (ANN) based method to determine the absorption $\mu _\mathrm{a}$ and scattering $\mu _\mathrm{s}$ coefficients of biological tissue from the diffuse reflectance R, total transmittance T, g, and n. ANNs were trained using dataset generated by calculating light transport in the DIS setup with a Monte Carlo method. The measured R and T spectra and the wavelength-dependent g and n were inputted to calculate $\mu _\mathrm{a}$ and $\mu _\mathrm{s}$. Due to the simple and fast calculation, the ANN-based method can calculate the $\mu _\mathrm{a}$ and $\mu _\mathrm{s}$ spectra assuming the wavelength dependence of g and n. The relative errors of reconstruction by the trained networks were 1.1% and 0.95% for $\mu _\mathrm{a}$ and $\mu _\mathrm{s}$, respectively. Each optical property spectra (total 662 points) was obtained in 1.1 ms. The proposed method can determine $\mu _\mathrm{a}$ and $\mu _\mathrm{s}$ in the DIS measurement assuming wavelength dependent g and n.