Abstract
正子造影術是一種醫學影像上所使用的成像技術,可以將人體內部生理功 能變化狀態,轉化成為一係列斷層影像的輸出,以提供做為醫療上病因的 診斷依據。然而由於在執行正子造影時,其所需的成像時間甚常,通常 為 10 到 30 分鐘之久。病人在這相當長的掃描時間內,會有位移或轉動 ,也就在所難免。但是如果一但病人在此成像期間有了移動,往往就會導 致所取得的影像發生模糊現像,其嚴重程度輕重不一,有的甚致於無法從 中做任何的醫學分析與醫療診斷。本研究提出了一個估計方式,來估計病 人在掃描過程中,其姿式變動的描述參數,然後再利用此參數模型,來修 正因病人移動所造成的模糊現像。此一估測方式包括,利用成像終止後所 計錄到的光子計量,根據其當病人有所移動時,光子計量所會呈現的特殊 現像,設計一個演算法則,從中抽取出重要的資訊。在利用這些資訊來轉 換重組出物體移動前後的輪廓,接著再使用重組而得的輪廓資訊來進一步 獲得平移和旋轉的描述參數。再來從有雜訊的影像中,利用其平均數效果 ,先降低雜訊對其該影像的影響度,然後求得相對應的相對流滯時間參數 。最後轉換所有的前置限定,以及移動模型,分別求得相對的凸集正交投 射運算子,再用凸集式投影方法,對得到的受移動現像所影響的影像做修 正,以期去除影像中的模糊現像。最後我們並以就模擬數值和真實正子照 影數值作分析與修正的實驗結果,來說明這處理方法對掃描中的病人位移 估計和影像修正的能力。 Positron Emission Tomography (PET) is a medical imaging technique which can provide internal functioning information of the human body for diagnosis. The reconstructed PET image, emission image, represents the radionuclide distribution of the human body. During the lengthy scanning process, patients sometimes move during the emission scan. This will cause severe artifacts which will deteriorate the image quality substantially and will make the information of interest in those images hard to obtained. We propose techniques which will estimate the motion parameters of the patient during emission scan. The technique involves reconstructing the shape of the object from the photon count data recorded during the emission scan, obtaining the shifting and rotation parameters from the reconstructed shape, estimating the relative duration parameter base on the reconstructed emission image considering the image is quite noisy. Afterward, we use the method of convex projection (POCS) to correct motion artifacts in the reconstructed emission images. The complete algorithm is implemented. Experimental results on both the simulated data and the real PET data will be presented using the algorithm proposed in this report to show the efficacy of the methods in analyzing patient's motion and correcting artifacts of the emission image in PET scan. The same principle can also be applied in estimating motion parameters in transmission scan.