基于模板匹配的心脏三维超声最优切面搜寻方法

目前实时三维超声是诊断复杂型先天性心脏病的主要手段.然而,利用实时三维超声系统,在心脏三维超声图像中,靠医生人工寻找病变的最佳观察面比较费时、可重复性较差,且容易造成漏诊;为便于复杂先心病的三维超声诊断,在孙锟等提出的复杂型先心病最优观察切面基础上,本文提出了基于模板匹配的自动搜寻心脏三维超声最优观察切面的方法,其中基于互信息方法的模板匹配能实现较高正确率的寻找最优切面.该方法易于实现,实时性好,避免了人工寻找切面的弊端,对测量及配准等后续的计算机辅助诊断意义重大.

Automatic Search for the Best Cross-sections of the Three-dimensional Ultrasound Image of Heart by Template Matching

Now real-time three-dimensional echocardiography (RT3DE) is an important tool to diagnose the complex congenital heart malformation. However, searching the cross-sections to view the cardiac anomalies in a three-dimensional ultrasound image of the heart by hand on the RT3DE system is time consuming, not repeatable, and easy to miss some places. In order to improve the existing diagnostic tool, we adopted the methods based on template matching for automatically finding the best cross-sections which are defined by Sun Kun et al. to view the cardiac anomalies in RT3DE's full-volume data. Among the methods is the entropy correlation coefficient which measured the similarity of the best cross-sections with the highest accuracy. The automatic search for these cross-sections is much quicker than the manual search. And with good repeatability it will make good preparation for the subsequent computer-aided diagnosis such as measure and registration, which is significant for the diagnosis of the complex congenital heart malformation.