漫反射光譜學（Diffuse Reflectance Spectroscopy, DRS）可藉由量測不同光源－偵測器間距（Source-Detector Separation, SDS）下的漫反射率，進一步反演組織的吸收係數與縮減散射係數，本次研究比較多層介質蒙地卡羅（Monte Carlo Multi-Layered, MCML）與極限蒙地卡羅（Monte Carlo eXtreme, MCX）在均質介質漫反射模擬、人工類神經網路（Artificial Neural Network, ANN）建模及實際假體量測中的差異。
首先比較 MCML 與其圖形處理器（Graphics Processing Unit, GPU）加速版本 CUDAMCML 的模擬結果，結果顯示在不同吸收係數、散射係數與SDS條件下，兩者的相對誤差皆小於 1%，證實 CUDAMCML 可視為 MCML 的加速版本，因此後續以 CUDAMCML 取代 MCML 模擬資料庫，並接著在相同光學參數條件下，比較 MCML 與 MCX 於不同SDS所計算的漫反射率，結果顯示兩者的差異在短距離SDS下較為明顯，且相對誤差會隨吸收係數與縮減散射係數增加而上升，當SDS增加時，兩者結果則逐漸接近。
為探討造成差異的原因，本次研究進一步分析光子步長（Step Size）與偵測器定義，將吸收係數設為零後，MCML 與 MCX 之間仍存在反射率差異，顯示步長並非主要影響因素，而MCML 採用圓柱對稱與徑向環形區域統計反射光子，而 MCX 則以三維體素（Voxel）描述介質，並以球形範圍判定偵測光子，在偵測器幾何形狀、收光條件等差異，可能使 MCX 高估單位面積反射率，並造成散射增加時兩者誤差上升。
本次研究亦分別使用 MCML 與 MCX 的模擬資料訓練正向人工類神經網路(ANN)及反向人工類神經網路(inANN)，建立吸收係數、縮減散射係數與 1 mm、2 mm 漫反射率之間的非線性映射關係，模型於大部分訓練範圍內均具有良好預測能力，但 MCX 模型在部分參數邊界區域出現較大的反演誤差，並使用實驗室自製矽膠光學假體與漫反射光譜量測系統進行實驗驗證，兩種模擬資料所建立的反演模型在實際量測中的吸收係數與縮減散射係數結果整體相近，尚未呈現顯著差異，但在較高吸收與較高散射的假體條件下，兩者差異有增加的趨勢。
綜合而言，MCML 適合用於均質介質及穩定光學參數資料庫的建立，MCX 則具有三維幾何建模、偵測器配置彈性、白蒙地卡羅（White Monte Carlo, WMC）計算及 GPU 加速等優勢，更適合模擬複雜組織與實際探頭幾何結構。

Diffuse Reflectance Spectroscopy (DRS) estimates tissue absorption and reduced scattering coefficients by measuring diffuse reflectance at different source–detector separations (SDSs). This study investigated and experimentally validated differences between Monte Carlo Multi-Layered (MCML) and Monte Carlo eXtreme (MCX) simulations in homogeneous turbid media. The comparison included theoretical reflectance calculations, artificial neural network (ANN) modeling, inverse optical-property retrieval, and silicone phantom measurements. Conventional MCML was first compared with its graphics processing unit (GPU)-accelerated implementation, CUDAMCML. Their relative differences remained below 1% under all tested conditions, confirming that CUDAMCML can be regarded as an accelerated MCML implementation. MCML and MCX were then compared under matched optical parameters. Their differences were greatest at short SDSs and increased with the absorption coefficient and reduced scattering coefficient, whereas the results became more similar as SDS increased. Further analysis indicated that photon step size was not the principal cause. Differences in detector geometry, photon-collection criteria, and reflectance normalization were more likely explanations. ANN and inverse ANN models trained with both databases showed good overall performance, although MCX-based models produced larger errors near the training boundaries. Phantom measurements generated similar optical-property estimates for the two models, with slightly larger differences under higher-absorption and higher-scattering conditions.

[1] Yu-Wen Chen, Jun-Yen Guo, Shih-Yu Tzeng, Ting-Chun Chou, Ming-Jen Lin, Lynn Ling-Huei Huang, Chao-Chun Yang, Chao-Kai Hsu, and Sheng-Hao Tseng, "Toward reliable retrieval of functional information of papillary dermis using spatially resolved diffuse reflectance spectroscopy," Biomed. Opt. Express 7, 542-558 (2016)
[2] Y. Chen, C. Chen, P. Huang, and S. Tseng, "Development of an advanced model for determining the optical properties below and beyond 600 to 1000 nm wavelength range from frequency domain diffuse reflectance," in Frontiers in Optics 2016, OSA Technical Digest (online) (Optica Publishing Group, 2016), paper FW3C.4.
[3] Wang, L., S.L. Jacques, and L. Zheng, MCML--Monte Carlo modeling of light transport in multi-layered tissues. Comput Methods Programs Biomed, 1995. 47(2): p. 131-46.
[4] https://www.atomic.physics.lu.se/fileadmin/atomfysik/Biophotonics/Software/CUDAMCML.pdf
[5] Erik Alerstam, Tomas Svensson, Stefan Andersson-Engels "Parallel computing with graphics processing units for high-speed Monte Carlo simulation of photon migration," Journal of Biomedical Optics 13(6), 060504 (1 Nov 2008)
[6] Alerstam E, Lo WC, Han TD, Rose J, Andersson-Engels S, Lilge L. Next-generation acceleration and code optimization for light transport in turbid media using GPUs. Biomed Opt Express. 2010 Sep 1;1(2):658-75. doi: 10.1364/BOE.1.000658. Epub 2010 Aug 23. PMID: 21258498; PMCID: PMC3018007.
[7] Erik Alerstam, Stefan Andersson-Engels, Tomas Svensson "White Monte Carlo for time-resolved photon migration," Journal of Biomedical Optics 13(4), 041304 (1 Jul 2008)
[8] Qianqian Fang and David A. Boas, “Monte Carlo Simulation of Photon Migration in 3D Turbid Media Accelerated by Graphics Processing Units.” Opt. Express, vol. 17, issue 22, pp. 20178-20190 (2009)
[9] Q. Fang and D. A. Boas, "GPU accelerated Monte Carlo simulation for 3D photon migration," in Biomedical Optics and 3-D Imaging, OSA Technical Digest (CD) (Optica Publishing Group, 2010), paper BME5.
[10] Rupawala M, Dehghani H, Lucas SJE, Tino P and Cruse D (2018) Shining a Light on Awareness: A Review of Functional Near-Infrared Spectroscopy for Prolonged Disorders of Consciousness. Front. Neurol. 9:350. doi: 10.3389/fneur.2018.00350
[11] Wang L, Jacques S L. Monte Carlo Modeling of Light Transport in Multi-layered Tissues in Standard C[M]. Houston, TX: University of Texas M. D. Anderson Cancer Center, Laser Biology Research Laboratory, 1992.
[12] https://www.edinst.com/resource/the-beer-lambert-law/
[13] Yu-Wen Chen and Sheng-Hao Tseng, "Efficient construction of robust artificial neural networks for accurate determination of superficial sample optical properties," Biomed. Opt. Express 6, 747-760 (2015)