- 學位論文
以最大交互訊息進行醫學影像對位
Medical Image Registration using Maximum Mutual Information
摘要
隨著電腦科學的發展，使的高速計算機的功能在影像處理上，各種複雜的運算能有效的發展實行。因此，醫療影像的技術，如電腦斷層掃瞄、核磁共振影像及正子斷層掃瞄，已成熟的發展於臨床診斷與應用。擬以影像處理的技術發展一套影像對位系統，結合結構性與功能性影像的特性，提供醫師對其主治部位之組織，進行準確的評估與診斷。 本論文主要針對結構性核磁共振影像與功能性正子斷層影像進行影像對位。首先利用影像相互位置關係之影像灰階值計算影像交互訊息(mutual information)，再以彼此之交互訊息疊代幾何轉換之參數，經過一連串的疊代法計算，影像之交互訊息與幾何轉換之參數將收斂至一穩定狀態，利用影像最大交互訊息獲得最佳之幾何轉換參數進行影像對位。 經由實驗驗證，本文所提出之對位系統，可利用於不同醫療影像進行對位二維及三維之影像對位，也可利用最大交互訊息之特性進行影像識別。希望藉由此對位系統提供醫學臨床上應用與發展。
並列摘要
Since the computer science attain to maturity, the image processing using high speed compute is exercised. The technique of medical image radiography such as computed tomography (CT), magnetic resonance images (MRI) and positron emission tomography (PET) images is developed for clinical diagnosis and application. Multi-modality image registration is designed to combine the information of medical image is from different modalities and record at different times. The purpose of the study is to register medical images such as MRI and PET images using maximum mutual information (MI). The image registration system is designed two phases: estimation of MI and image registration. One of the phases, the MI is estimated the statistical dependence or information redundancy between the image intensity of corresponding voxels in both images. The other phase is the parameters of the geometric transformation are obtained iterance by estimation of MI. In this study, the 2-D and 3-D registration techniques were developed, and applied on medical image studies. The other application is used the maximum MI after the registration processing for image recognition. The results show the ability to register and recognize the image efficiently. The method is developed and suited for clinical application.