利用斜角光纖定量雙層組織的光學參數

Abstract

在如口腔組織的鱗狀細胞瘤癌變的過程中，主要是從上皮結構開始產生退行性發育，因此在用於早期癌症診斷的光纖式漫反射光譜的研究中，準確的定量跟上皮層組織相關的光學參數是非常重要的，尤其是跟上皮細胞結構相關的散射係數。然而在雙層組織中，由於上皮層比起下層組織(基質)而言非常的薄，漫反射光譜大部分容易受到下層的影響，使得上皮層的光學參數相當難以準確地定量；而在前人提出的不同光纖設計的研究中，以模擬而言，對於上皮層散射係數，定量的誤差範圍約在4~42%，上下層多個光學參數的整體誤差約為1~49%。為了改善傳統光纖設計對上皮層光學參數不靈敏的缺點，我們利用沿著組織表面傾斜且平行排列的斜角光纖，及自行開發的光譜擬合工具來達到準確定量上皮層光學參數的目的。 本論文主要為數值模擬的研究，首先利用雙層縮放(scaling)蒙地卡羅(Monte Carlo)演算法來加速傳統蒙地卡羅演算法計算時間較長的缺點，此光譜模擬工具由C++所編寫，其計算時間顯著降低，且與傳統蒙地卡羅演算法相比，誤差小於1%。本研究利用此程式模擬調整上下層光學參數後，觀察空間解析光譜的靈敏度變化趨勢，由模擬結果可看出斜角光纖確實對上皮層光學參數的變化有較高的靈敏度，但對下層的血紅素吸收較不靈敏，此缺點可由圈選適當的光纖收光區域來改善。 接下來，本研究利用MATLAB編寫反向擬合工具來模擬定量雙層組織中的多個光學參數之準確度，利用最佳化的斜角光纖設計，分別擬合調整血紅素濃度或上皮層厚度的光譜，其定量上層散射係數的平均誤差約為1.5%，且上下層光學參數的整體誤差皆小於7%。因此本研究所提出的光纖設計及擬合方法不僅改善了定量上皮層散射係數的準確度，同時也準確地定量雙層組織中的多個光學參數，有利於光纖式漫反射光譜對於早期癌症的診斷。

Fiber-based diffuse reflectance spectroscopy has been applied to detect absorption and scattering properties associated with dysplasia in the tissue, which is a potential precursor of epithelial cancers. The more accurate measurement and quantification of optical properties in the thin epithelial layer is important to early dysplasia detection. The aim of this simulation study is to improve the accuracy of quantifying the scattering coefficients of the thin epithelial layer in a two-layered mucosal tissue model. The proposed method is to use obliquely oriented fibers which consist of a source fiber and multiple detection fibers parallel to each other and oriented obliquely to the tissue surface, instead of perpendicular fibers which have the drawback of weak performance of quantifying the optical properties of the thin epithelium. The major algorithm was Monte Carlo method. Since the conventional Monte Carlo simulation has high time complexity, a novel scaling method was used to speed up a forward model for simulating the spatially-resolved reflectance spectra. We implemented this simulation tool by C++. There was a significant improvement of time consumption, and the average deviation in the reflectance between our tool and the MCML open source code was 0.60±0.44%. We utilized our code to investigate the sensitivity of spatially-resolved spectra to the optical coefficients of the two layers. The results show that the oblique fibers had much higher sensitivity than perpendicular fibers to changes of the optical coefficients of the thin epithelial layer but lower sensitivity to stromal absorption and scattering coefficients. An inverse model was implemented with MATLAB to fit simulated spectra in order to quantify the unknown optical properties of the two-layered model. The fitting results show that the oblique fibers significantly reduced the root mean square percentage error of extracted epithelial scattering coefficient in comparison to the perpendicular fibers (from 32% to 1.5%). The average error in other optical properties was below 7% using the oblique fibers.