基于统计形状模型的三维解剖结构预测  

Three-dimensional anatomical structure prediction based on statistical shape model

作　　者：徐铸业 赵小强[1,2,3] 罗文菲 XU Zhuye;ZHAO Xiaoqiang;LUO Wenfei(College of Electrical and Information Engineering,Lanzhou University of Technology,Lanzhou 730050,China;Key Laboratory of Gansu Advanced Control for Industrial Processes,Lanzhou University of Technology,Lanzhou 730050,China;National Experimental Teaching Center of Electrical and Control Engineering,Lanzhou University of Technology,Lanzhou 730050,China)

机构地区：[1]兰州理工大学电气工程与信息工程学院,甘肃兰州730050 [2]兰州理工大学甘肃省工业过程先进控制重点实验室,甘肃兰州730050 [3]兰州理工大学国家级电气与控制工程实验教学中心,甘肃兰州730050

出　　处：《华中科技大学学报（自然科学版）》2021年第11期58-63,共6页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61763029)。

摘　　要：针对在构建患者病患部位的预测模型过程中存在数据缺损的问题,提出一种根据缺损样本构建三维(3D)解剖结构预测模型(3DASPMB-DS)的方法.首先对输入数据集进行数据处理,将感兴趣区域从图像中分割出来,并将其作为训练样本集;然后手动模拟训练样本集中的缺损样本在所有情况下发生的数据缺损,并根据缺损样本构建出患者病患部位的预测模型;最后将预测模型与测试样本拟合,通过计算预测模型中的点与测试样本中对应点之间的距离来判断二者的拟合程度,并用相似度量函数来进行评估.采用30组骨盆的计算机断层(CT)扫描图像作为输入数据集来进行实验验证,结果表明:3DASPMB-DS具有较小的距离误差和较高的相似性系数,并且能够有效地构建出股骨的预测模型.Aiming at the problem of data defect in the process of constructing anatomical structure prediction model of patient,a three-dimensional(3 D) anatomical structure prediction model based on defect specimens(3 DASPMB-DS) constructing method was proposed.First,the input specimens were processed to segment the regions of interest from an image,and the segmented regions of interest were used as the training specimens.Then,the data defects of the defect specimens in the training specimens were simulated manually in all cases,and the prediction model of patient location was constructed according to the defect specimens.Finally,the prediction model was fitted with the test specimens,and the fitting degree was judged by calculating the distance between the points in the prediction model and the corresponding points in the test specimens,and the similarity measurement function was used to evaluate the fitting degree.30 computed tomography(CT) scan images of pelvis were used as input data set for experimental verification,and results show that the 3 DASPMB-DS has the smallest distance error and the highest similarity coefficient,and can effectively construct the prediction model of femur.

关 键 词：医学图像处理 三维建模 统计形状模型 数据缺损 形状分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]