基于三维斜截面图像重建技术进行头颅CT同一性认定的研究

发布时间：2018-05-24 17:31

本文选题：法医放射学 + 断层解剖学　；参考：《复旦大学》2010年硕士论文

【摘要】：研究目的探索运用三维斜截面图像重建技术和断层解剖学知识相结合的方法研究MSCT数据集,再现任意CT截面的方位和层面,为不同时间、不同设备头颅CT图像的形态学比对提供一种有效的技术手段。研究方法 26例临床并人不同批次的CT扫描数据,以层厚10mm者为“原始组”,以层厚0.625mm者为“事后组”。对“事后组”及其他层厚为0.625mm的100例临床病人CT数据进行颅骨的三维重建以测量颅长,分别选择与“事后组”具有相同性别、颅长最接近者26例,设为“对照组”。将三组数据导入图像处理软件,分别观察断层影像解剖结构,寻找骨性标志点。参照“原始组”层面,使用软件中的ObliqueSlice模块对“事后组”、“对照组”数据进行方位和层面的重建,使用"fit to points"模块和‘'setplane"命令进行平面校正,使再现截面与“原始组”截面基本吻合。最后分别使用二维切面图像对齐和变换编辑器图像对齐的方法对比对图像进行叠加,进行同一性认定。导入“事后组”或“对照组”数据,参照“原始组”层面,定位相应骨性标志,并创建界标,使用刚性变换模块将“事后组”或“对照组”MSCT数据配准到“原始组”数据中,进行三维体绘制,并与“原始组”二维截面进行叠加比对。研究结果在三维重建下进行最大颅长(g-op)测量,“对照组”和“事后组”之间颅长差值为0～0.11cm,平均值为0.0381±0.03359cm。“事后组”MSCT数据集在ObliqueSlice模块下,观察颅内结构参照物随层面改变而变化,其解剖结构可以与“原始组”分别对应标识。判断中心层面和旋转轴方向,可基本模拟“原始组”方位和层面。“对照组”数据集在ObliqueSlice模块下,其解剖结构亦可以与“原始组”对应标识,但经过中心层面和旋转轴方向多次调整后,与“原始组”方位和层面的拟合度仍比较低。在“事后组”MSCT数据集基本模拟“原始组”方位和层面后,使用"fit to points"模块和"setplane"命令进行平面校正,利用“原始组”截面中三处以上解剖细节确定新截面,使再现截面与“原始组”截面相吻合。而“对照组”数据集重建截面无校正意义,难以同时再现“原始组”截面多数颅内结构参照物。定位“原始组”截面和“事后组”MSCT数据集重建截面为堆叠图像切片文件格式,使用二维切面图像对齐可基本叠加,但比对效果不佳,难以区分特征的吻合情况。而在变换编辑器图像对齐下,使用不同的绘图编辑器模式,图像差别区分明显,能实现有效比对。经过三维图像数据集的配准后,“原始组”二维截面可以与“事后组”MSCT三维体绘制在同一空间下进行立体比对,“原始组”截面的外形轮廓、颧弓的边缘、下颌骨、鼻中隔、以及枕骨均能与“事后组”MSCT三维体绘制吻合。当“原始组”截面用线型编辑器绘制,经Alpha模式实现透明化时,则比对效果明显。结论基于三维斜截面图像重建技术,个体头颅MSCT数据集可以再现不同批次扫描的CT截面的方位和层面,重建的截面与原CT图像可以有效地叠加比对,其中在变换编辑器图像对齐下，，使用不同的绘图编辑器模式,或使用三维图像数据集配准,图像比对效果较好,可以作为基于CT图像同一性认定的一种方法。研究目的使用计算机图像后处理技术重建头颅CT的共同平面,探索运用头颅CT图像的多项解剖特征指标,结合相同个体两次各项指标测量差值与不同个体各项指标测量差值,进行回归分析,探寻建立一种多元的基于头颅CT进行同一性认定的识别指标与方法。研究方法首先对“事后组”和“对照组”MSCT数据集进行重建以模拟“原始组”的截而方位和层面。在蝶窦平面等层面设置骨性标志点，确定参数并进行测量。计算各项指标“事后组”和“对照组”’与“原始组”的差值,进行统计描述。设“事后组”和“原始组”的差值为协变量X,相应因变量Y设为O。设“对照组”和“原始组”的差值为斜变量X,相应因变量Y设为1。运用二元Logistic回归的强迫引入法,根据P值和得分值筛选指标。将选得指标同时输入,运用二元Logistic回归的向后步进法进行回归分析。研究结果根据各层面骨性结构,并参考以往CT测量研究,在蝶窦等五个层面设置34个骨性标志点,本次研究确定头颅CT16项参数进行测量。部分结果如下三组间平均值和标准差无明显差别。事后组、对照组数据分别和原始组数据相减后,事后组、原始组数值差明显小于对照组和原始组数值差运用强迫引入法二元Logistic回归,建立头颅CT同一性认定一元方程16个,各指标进行个体识别的准确率在65.4%至90.4%,其中运用上颌窦宽、颧弓间距进行个体识别正确率较高,达90.4%,其余各项指标中,颞骨岩部角、额窦宽径、额窦平面颅宽等准确率也在80%以上,有检验意义。运用向后步进法进行回归分析,获得以颧弓间距、乙状窦间距、蝶骨大翼角、额窦平面颅宽的四元方程,Y=X3×89.716+X4×21.186+X9×3.185+X16×26.174-53.098,可以使错误率趋于0,效果最好,其OR值为1.1489732385186304E23,有实践价值,可用于头颅CT图像的同一性认定。结论 MSCT数据集重建“原始组”CT体位后,可以提供精确测量的客观数据,相同个体两次各项指标测量差值较小,而不同个体各项指标测量差值较大。经二元Logistic回归分析,获得以颧弓间距、乙状窦间距、蝶骨大翼角、额窦平面颅宽的四元方程,准确率高,并可以提供比较客观的出错率,符合证据规则。
[Abstract]:research objective
The method of combining 3D oblique cross section image reconstruction with sectional anatomy knowledge is used to study the MSCT data set, to reproduce the azimuth and level of arbitrary CT cross section, and to provide an effective technical means for the morphological comparison of CT images of different equipments in different time.
research method
The CT scan data of 26 patients with different batches were "primitive group" with the thickness of 10mm as the "original group", and the layer thickness 0.625mm was the "post group". The three dimensional reconstruction of the skull was carried out to measure the skull length for the "post group" and the other 100 clinical patients with the thickness of the layer of 0.625mm to measure the cranial length, and the same sex was the same as the "post group", and the cranial length was the closest. 26 cases were set up as "control group". Three groups of data were introduced into the image processing software to observe the anatomical structure of the fault images and find the bone mark points respectively. Referring to the "original group" level, the ObliqueSlice module in the software was used to reconstruct the "post group", "control group" data and to use the "fit to points" module and ''. "Setplane" commands plane correction to make the reappearance section basically coincide with the "original group" section. Finally, the image alignment is compared with the method of alignment of the two-dimensional cut surface image alignment and the transform editor. The identity of the image is compared to the same character. The "after group" or "group" data are introduced to the "original group" level. The corresponding bone marks are set up and the boundary marks are created. The "post group" or "control group" MSCT data are registered to the "original group" data using the rigid transformation module, and the 3D body is plotted and compared with the two dimensional cross section of the "original group".
Research results
The maximum cranial length (g-op) was measured by three-dimensional reconstruction. The difference of cranial length between "control group" and "ex post group" was 0 ~ 0.11cm, and the average value was 0.0381 + 0.03359cm. "ex post group" MSCT data set under the ObliqueSlice module. The changes of the intracranial structure reference were observed with the change of the level, and the anatomical structure could be compared with the "original group", respectively. It should be identified. Judging the center level and the axis of rotation, the "original group" orientation and level can be basically simulated. Under the ObliqueSlice module, the data set of the "control group" can also be identified with the "original group", but after many adjustments in the direction of the center and the axis of rotation, the fitting degree to the azimuth and level of the "original group" It's still low.
After the "post group" MSCT data set basically simulates the azimuth and level of the "original group", the "fit to points" module and the "setplane" command are used for plane correction, and the new section is determined by more than three anatomical details in the "original group" section, and the reappearance section is in accordance with the "original group" section. It is difficult to reproduce the majority of intracranial structural reference points of the original group at the same time.
The reconstruction cross section of the "original group" section and the "post group" MSCT data set is the stacked image slice file format, and the alignment can be basically superimposed by the two-dimensional slice image alignment, but the comparison effect is not good, and it is difficult to distinguish the characteristics of the anastomosis. And the image difference is distinguished by the different drawing Editor Mode under the image alignment of the transform editor. It is obvious that effective alignment can be achieved.
After the registration of 3D image data sets, the two-dimensional cross section of the "primitive group" can be compared with the "post group" MSCT three-dimensional body in the same space. The outline of the "primitive" section, the edge of the zygomatic arch, the mandible, the nasal septum, and the occipital bone are all consistent with the "post group" MSCT three-dimensional body. The cross section is drawn by linear editor and transparent by Alpha mode.
conclusion
Based on the 3D oblique cross section image reconstruction technique, the individual head MSCT data set can reproduce the azimuth and layer of the CT cross section of different batch scans. The reconstructed cross section can be effectively overlapped with the original CT image, in which the image alignment of the transform editor, using the different drawing editor mode, or the registration of the 3D image data set, is used. Image matching effect is better, which can be used as a method of image identity recognition based on CT.
research objective
Using the computer image post-processing technique to reconstruct the common plane of the head CT, explore the multiple anatomical features of the head CT image, combine the two indexes of the same individual to measure the difference between the differences of the individual and the various individual indexes, and carry on the regression analysis, and seek to establish a multiple identification recognition based on the identity of the head CT. Indicators and methods.
research method
First, the MSCT data set of "post group" and "control group" was reconstructed to simulate the position and level of the "original group". The bone markers were set on the plane of the sphenoid sinus, and the parameters were determined and measured. The difference between "after group" and "control group" and "original group" was calculated, and the statistical description was carried out.
The difference between "post group" and "original group" is a covariate X, and the corresponding dependent variable Y is set to O. to set "control group" and "original group" to be a skew variable X. The corresponding dependent variable Y is set up to 1. using the forced introduction of two element Logistic regression, selecting the index according to the P value and the score value. The selected index is input at the same time and uses two yuan Logistic back. Regression analysis is carried out by backward step method.
Research results
According to the bone structure in each level, and referring to previous CT measurements, 34 bone markers were set up at five levels, such as the sphenoid sinus. This study determined the CT16 parameters of the head.
There was no significant difference between the average and the standard deviation between the three groups. After the latter group, the data of the control group were subtracted from the original group, and the difference between the original group and the original group was significantly lower than that of the control group and the original group.
By using the two element Logistic regression of the forced introduction, the CT identity of the head was established, and the accuracy rate of individual identification was 65.4% to 90.4%. Among them, the accuracy rate of the individual identification was higher in the maxillary sinus and the zygomatic arch, and the other indexes, the angle of the temporal bone, the width of the frontal sinus and the plane cranial width of the frontal sinus, etc. The accuracy rate is more than 80%, and it has the test significance. Back step method is used to carry out regression analysis to obtain the four element equation with the distance of the zygomatic arch, the spacing of the sigmoid sinus, the great wing angle of the sphenoid bone and the plane cranial width of the frontal sinus. Y=X3 * 89.716+X4 x 21.186+X9 x 3.185+X16 x 26.174-53.098 can make the error rate be 0, the effect is best, and the OR value is 1.1489732385186304E23. The practical value can be used to identify the identity of head CT images.
conclusion
After the MSCT data set reconstructs the "original group" CT body position, the objective data of accurate measurement can be provided. The difference between the two indexes of the same individual is small and the difference between the various individual indexes is relatively large. The four element equation of the zygomatic arch spacing, the inter sigmoid distance, the great wing angle of the sphenoid bone and the plane cranial width of the frontal sinus is obtained by the two element Logistic regression analysis. High accuracy, and can provide a more objective error rate, in line with the rules of evidence.
【学位授予单位】：复旦大学
【学位级别】：硕士
【学位授予年份】：2010
【分类号】：D919

【参考文献】