生物資源暨農學院: 生物產業機電工程學研究所指導教授: 周呈霙許悅Hsu, YuehYuehHsu2017-03-022018-07-102017-03-022018-07-102015http://ntur.lib.ntu.edu.tw//handle/246246/272485正子斷層掃描目前已經是一個成熟的醫療診斷技術，另一方面由於藥物的開發需要先做小動物實驗，因此本研究使用雙平板小動物系統。此系統有高靈敏的感測以及方便調整平板的特性，但是此系統也有資料遺失的問題。由於飛行時間法的提出，搭配飛行時間法的正子斷層掃描系統已經大幅改善重建影像的品質。另一方面為了解決衰減校正的問題，目前常見的為搭配電腦斷層掃描或核磁共振造影去得到衰減影像。不過這幾年，已經有團隊提出maximum likelihood activity and attenuation estimation (MLAA)，此方法僅使用飛行時間正子斷層掃描系統即可同時估算正子分部影像以及衰減影像，有效降低診斷所需的時間。 而本次研究我們使用了考慮時間飛行法的系統矩陣，並提出使用交替方向乘子法(alternating-direction method of multipliers estimation)，目的是希望藉由交替方向乘子法基於全變差方法的概念改善雙平板系統資料遺失的問題。另一方面我們也改善MLAA的演算法使其搭配全變差的演算法，由於兩方法都有使用全變差，這讓我們的比較更為公平。最後根據重建影像分析，我們可以得知MLAA搭配全變差演算法以及交替方向乘子法都必須迭代15次才會收斂，而交替方向乘子法重建正子分布影像以及衰減影像的百分誤差分別為59%和34%；另外MLAA搭配全變差的正子分布影像以及衰減影像的百分誤差分別為67%和42%。Positron emission tomography (PET) is a mature technique of medical diagnosis. If we want to develop the tracer for PET, we should do animal experiment first. It is the reason why we utilized the dual-head small-animal positron emission tomography (DHAPET) system. DHAPET has characteristics of high detection sensitivity and flexible system configurations. However, the system geometry may result in severe parallax errors and undetected events along the trans-axial direction, thus reducing the image quality. With the advancement of detector hardware, time-of-flight (TOF) of photons can be detected, which has been demonstrated to significantly improve image quality, and can greatly benefit its use in research studies and clinics. On the other hand, the conventional attenuation correction requires performing additional transmission scan, which represents more radiation dose. To reconstruct activity distribution and attenuation coefficient simultaneously, the maximum likelihood activity and attenuation estimation (MLAA) was proposed. In this work, we numerically computed the DHAPET system response matrix that includes TOF information and employed the alternating-direction method of multipliers estimation (ADMM) based on total variation (TV)-minimization. By applying the ADMM, the reconstructed images can be anticipated to have improved quality with better spatial resolution. On the other hand, we combined the MLAA with TV for improving the MLAA. Then we compared the ADMM with MLAA-TV, they are convergence after 15 iteration, and the normalized error of activity and attenuation in ADMM are 59% and 34%, and the normalized error of activity and attenuation in MLAA-TV are about 67% and 42%.論文使用權限: 不同意授權正子斷層掃描衰減影像飛行時間法全變差PETattenuationTOFtotal variation (TV)ADMMthesis