擴散光學斷層掃描系統晶片設計與實現

Abstract

光學醫療照影技術在生醫界中扮演重要的角色，得以在不破壞組織的情況下進行組織特定物質的探測以及影像重建，像是大腦皮質的血氧分佈、淺層組織之癌細胞分佈或其他載色體物質分佈等等，並做為疾病發現之重要診斷來源。本論文提出一個可攜式之擴散光學斷層掃描(diffuse optical tomography)及二維血氧濃度百分比 (2D-oxygen concentration distribution) 處理器，另外也發展了不同的光學前向模型 (optical forward-model)做為重建演算法的驗證以及開發，並透過此模型設計出下一代新穎性的擴散光學斷層掃描重建演算法。 本論文所提出的整個系統是基於指令集架構 (instruction set architecture) 做特殊應用之處理器，每個指令皆包含一個控制訊號和觸發信號處理 (signal processing) 。另外也制訂一套外部命令 (external command) ，用於與外部設備溝通 (external communication) ，可以藉由外部命令進行最高優先權的系統修改。整個系統控制部分做各模組的調配處理 (deployment processing) ，並配合前端電路控制 (front-end control circuit) 以及資料輸出輸入 (input/output interface) ，與外部控制訊號與前端做完整的配合，並包含多項操作模式 (operation mode) 。此整合型腫瘤檢測系統 (integrated tumor detection system) 設計使用台灣積體電路製造股份有限公司90奈米 CMOS技術下線晶片 (TSMC) 90-nm CMOS technology。 另外本論文提出使用蒙地卡羅法 (Monte Carlo method) 的光子行為描述，基於較偏數學型式上做光子行為深入的模擬及探討，並藉此提供重建演算法開發上較準確且公平的驗證方式，更由此模型探討到的結果研發出一套非線性斷層重建演算法 (nonlinear iterative tomography reconstruction algorithms) ，利用光線在不同擴散距離的行為透過修改過的比爾-朗伯定律(modified Beer–Lambert law)簡化出數學前向公式 (forward equation) ，並利用數學疊代重建技術 (iterative method) 找出公式的解而求得組織的異物分佈。最後也使用蒙地卡羅法前向模型做精確的驗證與比較並證明此演算法的實現可能性。

The optical medical imaging technologies play an important role in medical profession, it can detected specific substances and rebuild image in tissue under the situation without body damaging, such as oxygen distribution in cerebral cortex, cancer location of superficial tissue or other distribution of chromophores, which can be an important diagnosis source of disease findings. In this research, we design a portable processor integrating the diffuse optical tomography (DOT) and 2D-oxygen concentration distribution. Beside, also develop different optical forward-model as a verification method of reconstruction algorithms, and design a new generation novelty reconstruction algorithms through this forward-model. The system based on the instruction set architecture, processing control signals and trigger signals. To facilitate external communication with peripheral equipment, we make a set of the external commands to implement the system modification with highest priority. To make deployment processing, we integrate front-end control circuit and input/output interface operation mode to coordinate with the external control signal. Moreover, various operation modes are also included, as well. The chip in our integrated tumor detection system is tapped out by Taiwan Semiconductor Manufacturing Company (TSMC) 90-nm CMOS technology. In other hand, this article proposed an optical forward-model of DOT by Monte Carlo method which is adopted to help simulate the photon propagating behavior with closer inspection. On the basis of mathematical analyses, we hope that the accurate and impartial verification as a pioneer way for new-type reconstruction algorithm. The new generation reconstruction method is nonlinear iterative tomography reconstruction algorithms. Use of light behavior under different diffusion distance to simplify the mathematical formula of Forward equation and using an iterative method to resolve the inverse problem and the result obtain the distribution of distribution of foreign matter. Finally the Monte Carlo optical forward-model also uses to verify this algorithm and proof the implemented possibility. Also, the comparison with conventional works is given and shows the strengths and weaknesses of this innovative reconstruction method.