以漫反射光譜術與蒙地卡羅模擬於活體定量人體頸部肌肉及淺層組織光學參數

Abstract

定量組織光學參數有利於使用光子進行活體診斷和治療。然而現有文獻缺乏在測量活體肌肉組織的光學參數時，考慮淺層組織的光學參數。本研究目標為穩定地測量頸部肌肉組織光學參數。測量部分為自建漫反射光譜系統，採用寬頻發光二極體及三通道自製光纖（SDS=4.5, 7.5, 10.5 mm），分析波長範圍為711∼880nm。根據受試者的真皮和皮下脂肪厚度建立蒙地卡羅模型，並訓練神經網路代理模型作為順向工具加速模擬，代理模型的平均方均根誤差小於2%。模型分為表皮、真皮、皮下脂肪和肌肉層平行均質四層，其μa 和μs 範圍參考了文獻。在表皮層中，散射相位函數是透過時域有限差分模擬得出的，g=0.94。而其他組織層則使用Henyey-Greenstein 散射相位函數，真皮層中g=0.715、其餘組織層g=0.9，折射率皆設定為1.4。在擬合過程中，μa(λ) 是透過計算每個組織層內各種吸光物質的濃度決定，μs(λ) 是透過逆冪律決定。本研究採用非線性迭代曲線擬合方法萃取組織光學參數。本研究對三名受試者的頸部區域進行了測量。我們提取的每個組織的μa(λ) 與先前文獻提供的範圍一致，而受試者B 的肌肉層μ′s(λ) 略低於文獻，受試者D 的真皮層與肌肉層μ′s(λ) 略高於文獻。三名受試者的實驗光譜皆擬合良好，光譜誤差分別僅有1.52%, 2.39% 及3.72%。為了估計定量光學參數的準確性，將我們系統上測量的雜訊添加到測試光譜中並進行擬合。μa(λ) 和μ′s(λ) 的平均方均根誤差最大為真皮層μa 的23%。本研究也對受試者的前臂進行了動脈和靜脈閉塞實驗，其光強度的變化與預期的生理狀態改變一致。

Quantifying tissue optical parameters facilitates the use of photons for in-vivo diagnosis and treatment. However, the existing literature lacks consideration of optical parameters of superficial tissues when measuring optical parameters of muscle tissue. The goal of this study is to stably measure the optical parameters of neck muscle. The measurement part is a self-built diffuse reflectance spectroscopy system, which uses broadband lightemitting diodes and three-channel self-made optical fibers (SDS=4.5, 7.5, 10.5 mm), and also the analysis wavelength range is 711∼880 nm. A Monte Carlo model was established based on the layer thickness of the subject’s dermis and subcutaneous fat tissue, and a neural network surrogate model was trained as a forward tool to accelerate simulation. The average root mean square error of the surrogate model was less than 2% . The model is divided into four parallel and homogeneous layers: epidermis, dermis, subcutaneous fat and muscle layers. The ranges of μa and μs refer to the literature. In the epidermis layer, the scattering phase function is obtained through finite-difference time domain simulationwith g=0.94. The other tissue layers use the Henyey-Greenstein scattering phase function, with g=0.715 in the dermis layer, g=0.9 in the other tissue layers, and the refractive index is set to 1.4. During the fitting process, μa(λ) is determined by calculating the concentration of various chromophores in each tissue layer, and μs(λ) is determined through the inverse power law. In this study, a nonlinear iterative curve fitting method was used to extract tissue optical parameters. This study measured the neck region of three participants. The μa(λ) we extracted for each tissue was consistent with the range of previous literature. However, The muscle layer μ′s(λ) of subject B is slightly lower than the literature, and the dermis layer and muscle layer μ′s(λ) of subject D is slightly higher than the literature. The experimental spectra of the three subjects all fit well, with spectral errors of only 1.52%, 2.39% and 3.72% respectively. To estimate the accuracy of the quantitative optical parameters, the noise measured on our system was added to the test spectra and fitted. The maximum average root mean square errors of μa(λ) and μ′s(λ) is 23% of the dermis μa. This study also conducted arterial and venous occlusion experiments on the subject’s forearm, and the changes in light intensity were consistent with expected physiological state changes.