- 學位論文
SPECT 基於互資訊和內插CT下應用影像特徵做CT 與腦部影像對位
CT and SPECT Brain Image Co-Registration Using Image Features Based on Mutual Information and CT Interpolation
摘要
隨著電腦科學的發展，使得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，已於臨床診斷與應用上成熟發展，而目前臨床上用於診斷的影像根據其成像的方式可以區分為解剖性(Anatomical)及功能性(Functional)這兩大類。解剖性影像是指取得的影像能夠清楚顯示及辨別器官或組織的輪廓及外形，磁振造影(MRI)及電腦斷層造影(CT)即屬於此類；而功能性影像是指能夠獲得組織或器官的新陳代謝情形，正子斷層造影(PET)和單光子電腦斷層造影(SPECT)即屬於此類。 顯而易見地,結合功能性及解剖性影像能夠提供醫生更多的資訊，更利於疾病的診斷及治療。所以我們藉由電腦影像處理的方式來使之結合在一起，稱為影像對位(Image Co-Registration)。並由於電腦技術進步的快速，使的電腦的運算速度越快功能越強大但成本卻越來越便宜。 在此，我們針對電腦斷層造影與單光子電腦斷層造影進行影像對位，並提出利用內插式影像來增加對位準確度。方法首先是物體內部對位(Intra-Subject Registration)，利用理想中心矢狀切面演算法(ideal Mid-Sagittal Plane Algorithm) 以腦部對稱性結構來將歪斜的影像調整回中心線，並在影像Z軸上利用影像內插法結合DICOM資訊來產生多張內插影像。再來是物體之間對位(Inter-Subject Registration)，利用互資訊(Mutual Information)對內部校準完的結果作更精確的調整。經過實例測試之後,採用影像內插法可確實提升影像對位的準確度，同時我們建立了一套具有人機介面的醫學影像處理與校準系統。
並列摘要
According to formation image, the medical image can be divided into two major classes that are anatomical image and functional image. Magnetic resonance image (MRI) and computed tomography (CT) belong to the anatomical image, which can show an outline of an organ and a tissue clearly. The functional image can show the image of the organ and the tissue metabolism situation. Positron emission tomography (PET) and single photon emission computed tomography (SPECT) are also part of the functional image. Obviously, combining the functional and anatomical image can offer more information to doctors who make diagnoses. These images benefit the diagnosis for diseases so the doctors can provide better treatment for their patients. Moreover, the combination of these two images that we deal with by image process of the computer, called image co-registration. Because of the fast computer technology development, people nowadays have high speed computer and powerful computer but cheaper. In the study, we aim at the CT image and SPECT image to registration mainly and adopt interpolationing CT image to improve the system..First, intra-subject registration utilizes ideal Mid-Sagittal Plane Algorithm (iMSP) and image interpolation. IMSP is used to adjust the skew image into symmetry to the straight mid-line in the brain. Utilizing image interpolation based on DICOM information to produce several interpolationing images in the Z-axis can provide more image information to process. Second of all, the inter-subject registration utilizes the mutual information (MI), which regulates the result from the intra-subject registration so that the accurate adjustment can be made. According to an actual example, adopt interpolationing CT image can improve the registration accurate. Finally, a co-registration system with window-based interface built by Borland C++ Builder is introduced.