| Summary: | 碩士 === 國立成功大學 === 光電科學與工程學系 === 103 === In this study, we propose a method to measure human skin with SDS = 1mm by FDPM system. In order to improve the shortcomings of traditional photon propagation model, we established a new photon propagation model which combined the GPU-MCML and ANN method to derive the sample absorption coefficient、reduced scattering coefficient corresponding to the amplitude and phase delay with the frequency. The new model can substitute traditional Monte Carlo model and standard diffusion equation which have some using limitation.
In simulation, Monte Carlo is considered as a gold standard model of photon propagation. However it consumes lots of time to do the simulation. Compared with Monte Carlo, the percent deviation of amplitude and phase which simulated by artificial neural networks are about 4% and 6%, respectively. In the other hand, comparing with Monte Carlo, the percent deviation of amplitude and phase which simulated by standard diffusion equation are about 30% and 17%, respectively. Therefore compared with diffusion equation the new model with a 1mm source to detector separation indeed has its absolute advantage. We also use the homogeneous phantom which optical properties is known to confirm the feasibility of the artificial neural networks model.
In experience,we measured three position of three healthy adults, including the finger, the inner forearm and outer forearm with the artificial neural networks model extrapolated absorption coefficient、reduced scattering coefficient with three wavelengths. And then we put the absorption coefficient into the chromophore fitting to quantify the chromophore concentrations of tissue and StO2.
In conclusion, our new artificial neural networks model has better efficiency and wider applied range than other traditional model, which can accurately calculate the optical properties of the superficial skin tissue.
|
|---|
