基于单图谱分割的脊髓自动分割算法研究  

RESEARCH ON AUTOMATIC ATLAS-BASED SEGMENTATION OF THE SPINAL CANAL TOMOGRAPHIC IMAGES

作　　者：刘露露 马钟 贺占庄[1] 毛远宏 Liu Lulu;Ma Zhong;He Zhanzhuang;Mao Yuanhong(Xi’an Microelectronics Technology Institute,Xi’an 710065,Shaanxi,China)

机构地区：[1]西安微电子技术研究所,陕西西安710065

出　　处：《计算机应用与软件》2022年第8期253-257,297,共6页Computer Applications and Software

摘　　要：针对当前脊髓CT图像分割过程中存在的分割精度不高、自动化程度不足的问题,提出一种改进的图谱自动分割算法。该算法利用图谱分割中先验解剖知识的优势,结合模糊连接算法实现脊髓的自动分割。该算法基于待分割图像与图谱图像间的联合概率分布,计算归一化互信息,构建最佳图谱并进行图谱分割,基于图谱分割结果构建生长种子区域,利用脊髓目标与周围区域的模糊连接特性构造合适的模糊隶属度函数,以此完成对脊髓目标的自动分割。实验表明,该方法可以提高配准精度,具有较好的鲁棒性和自动化程度。Aimed at the problems of low segmentation accuracy and insufficient automation in the current spine canal CT image segmentation process,an improved automatic image segmentation algorithm is proposed.This algorithm took advantage of the prior anatomical knowledge in image segmentation,and combined the fuzzy connection algorithm to realize the automatic segmentation of the spine cord.Based on the joint probability distribution between the image to be segmented and the atlas image,it calculated the normalized mutual information,constructed the optimal atlas and performs atlas segmentation,constructed the growth seed area based on the atlas segmentation result.The fuzzy connection characteristics between the spinal cord target and the surrounding area were used to construct a suitable fuzzy membership function to complete the automatic segmentation of spine cord targets.Experimental results show that this method can improve the registration accuracy,and has better robustness and automation.

关 键 词：脊髓CT图像 图像自动分割 单图谱分割 图像配准 模糊连接 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]