一种基于知识的CT影像脊髓自动提取方法(修改稿2)教程.doc

CT影像中一种基于知识的脊髓自动提取方法 王 玉1, 王 宏2, 康雁1 （1.东北大学 中荷生物医学与信息工程学院110819；2.东北大学 机械工程与自动化学院110004） 摘 要：本文基于CT影像人体结构的特征知识，提出了用三个主要步骤实现脊髓的自动提取功能。并在检测脊髓概率区关键步骤中，基于脊髓及其周围结构的特征知识，建立了一个全新的特征模型进行脊髓内一点的检测，并基于该点进行区域增长得到脊髓概率区，进而在脊髓概率区内实现脊髓的检测。在临床60例患者CT图像序列的试验中，躯干轮廓检测率全部达到100％，脊髓概率区的检测率有两例99％，其余均100％，基于脊髓概率区的脊髓检测率全部可以达到100％。引入特征模型自适应修正，可以实现全部60例患者CT图像序列脊髓的自动提取。运行于笔记本平台上，患者CT图像序列脊髓检测时间可以达到3秒左右，完全满足临床要求。脊髓自动提取功能的实现，避免了临床放疗医师进行手动勾画的繁琐工作，同时也降低了勾画结果对放疗医师技术水平的依赖，在临床放射治疗领域具有重要意义。 关键词：CT影像；自动提取；特征模型；区域增长；手动勾画 中图分类号：Q-332 文献标识码：A 国家标准学科分类代码510-40 A Knowledge-Based Method of Automatic Detection of the Spinal Cord in CT Images WANG Yu1，WANG Hong2，1 (1. Sino-Dutch?Biomedical and Information Engineering School, Northeastern University, Shenyang 110819, China; 2. Northeastern University POB 319, Shenyang, 110004, China) Abstract：In this article, based on body diagnostic structural knowledge, three key subtasks were performed to contour spinal cord automatically. A new diagnostic model was created to detected spinal cord point according to the structural knowledge of spinal cord and its surroundings for detecting spinal cord possible region. From this point the spinal cord possible region was obtained with region growing method. Then the spinal cord was detected in the spinal cord possible region. The testing results of sixty clinical patients is that body contours were detected successfully for all patient CT series images, spinal cords were detected successfully after spinal cord possible region was obtained for all patient CT series images, ninety-nine percent of spinal cord possible region was detected successfully for two patients while hundred percent of spinal cord possible region was detected successfully for the rest. With the adaptive diagnostic model, the automatic detection of spinal cord for all patients was successful. The detection time was about three seconds on notepad PC which meets the clinical requirements. It can reduce the dependence degree of the operator te