Investigation on accuracy and sensitibity of three-dimensional computed tomography, and its application for simulation surgery.

本研究的目的如下。1.研究3D-CT图像和颅骨复制品的准确性。2.研究CT图像和颅骨复制品上骨缺陷的原因。3.寻找减少假阴性数据的方法。4.评估无框架三维数据配准系统的准确性。以此体积数据为基础，制作0.2 mm层面的CT阈值为200HU的颅骨复制品，在额鼻缝水平切开颅骨，从上方拍摄眶底位的照片，然后用0.2 mm的层面数据重建眶底骨表面图像，阈值CT值分别为200HU和100HU，然后在工作站上以2.0 mm的间隔和0.2 mm的插值法进行重建。在此基础上，分别用阈值为200HU和100HU的CT阈值重建眼眶底三维骨面图像，并在个人计算机上对图像和三维骨面图像的骨丢失面积与整个眼眶底面积的比值进行了比较。研究结果表明：1.阈值和内插过程中产生的误差是造成假性颅底骨缺损的主要原因。2.通过制作CAD数据积累数据，造成假性负性骨缺损。取阈值CT值为100HU的数据可以减小误差。3.空格模型具有较高的精度，最大误差为1.96 mm，最小误差为0.25 mm，平均误差为0.73 mm。4.平均配准误差小于1个像素(=0.39 mm)

The purpose of this research were as follows.1.Investigation of accuracy of 3D-CT image and skull replicas.2.Investigation of the cause of bone deficits on CT image and skull replicas.3.Search the way to reduce the false negative data.4.Evaluation of the accuracy of the frameless three-dimensional data registration system.Helical volume CT data of the whole head were accumulated. Using this volume data ; 0.2 mm slice data were prepared and a skull replica fabricated with a threshold CT value of 200 HU.The skull was cut at the level of the frontonaseal suture and a photograph of the orbital floors was taken from above.3D osseous surface images of the orbital floors then were reconstructed with the 0.2 mm slice data with threshold CT values of 200 HU and 100HU.Slice data then were prepared with an interval of 2.0 mm and images interpolated at 0.2 mm on the workstation. From these interpolated slice data, 3D osseous surface images of the orbital floors were reconstructed with threshold CT values of 200 HU and 100HU.The ratios of the area of bone loss to the whole orbital floors in the photograph and 3D osseous surface images were evaluated with a personal computer and then compared with each other.We could conclude by the studies as follows.1.The errors produced during the precess of thresholding and interpolation had play the main role in causing false bone deficits in the orbital floors of skull replicas.2.False negative bone deficits were created on the way from accumlating data through making CAD data. The error could be reduced by taking data with a threshold CT value of 100HU.3.Skull model have relatively high accuracy which errors are 1.96 mm maximum, 0.25 mm minimum, 0.73 mm average.4.The average registration error was liss than one pixel (=0.39 mm)