淬取大腦影像正中矢狀切面進行三維電腦斷層與磁振造影的對準與融合

IMAGE REGISTRATION OF 3D CT AND MR IMAGES BY USING MID-SAGITTAL PLANE OF THE BRAIN

周銘鐘(Ming-Chong Chou) ； 蕭正英(Chen-Ying Shiau) ； 陳志成(Jyh-Cheng Chen)

《放射治療與腫瘤學》 10卷2期 (2003/06) Pp. 89-96

https://doi.org/10.6316/TRO/200310(2)89

摘要

目的：本篇主要是著重於利用大腦的正中矢狀切面進行三維電腦斷層與磁振造影的影像對準與融合。
材料與方法：由於電腦斷層以及磁振造影的影像均會呈現出左右對稱的大腦半球，利用人為點選的方式找出大腦軸切影像的對稱中線，透過每一張軸切影像上的中線前後兩點，回歸出兩組影像正中矢狀切面所在的平面方程式，再藉由空間幾何轉換使兩組影像的正中矢狀切面對應到同一平面(Z= 0)上。由於每一張影像的中線前的兩點是位於正中矢狀切面上，因此這些點也會被轉換到相同平面(Z=0)上，利用三次曲線內插(cubic spline interpolation)法將這些點內插出正中矢狀切面的邊界，並利用POWELL 最佳化法進行兩組影像的正中矢狀切面圖形的對準，最後再將影像做融合。
結果：本研究評估了兩組影像，結果顯示在x-y方向的誤差為1~3mm，而在z方向的誤差大約為4mm。
結論：本方法利用相同的正中矢狀切面以進行影像對準，降低了對準的自由度，並且也降低了對準的時間。

關鍵字

正中矢狀切面,迴歸,空間幾何轉換,三次曲線內插,最佳化法,影像對準,影像融合

並列摘要

Objective: The Purpose of this study is to register 3D MR images with corresponding CT images by using mid-sagittal plane of the brain.
Materials and Methods: We manually select the end points of each mid-line from each slice from MRI and CT, respectively. From those coordinates we obtain estimated parameters for an equation of a plane using statistical regression method for MRI and CT, respectively. With the usage of geometric transformation, we transform the two sets of image points to fit the same plane with z = 0. Then, we exert cubic spline interpolation to find the edge of the sagittal slices of the two (MRI and CT) images. Finally, we put forth POWELL optimization method to register the two sagittal slices.
Results: In our study, we test two sets of MRI and CT images with some part of the brain to evaluate the accuracy of out method. The registration error is 1~3 mm in the x and y direction, and about 4 mm in the z direction.
Conclusions: We find that the degree of freedom in registration has been decreased in out method, and it is more time saving by using the same mid-sagittal plane for registration has been decreased in out method, and it is more time saving by using the same mid-sagittal plane for registration.

並列關鍵字

Mid-sagittal plane,Regression,Geometric transformation,Cubic spline interpolation,Optimization,Image registration,Image fusion