论文部分内容阅读
目的针对心血管内超声(IVUS)图像中钙化斑块、声影等干扰因素影响外弹力膜(EEM)轮廓检测准确性的问题,提出结合先验形状信息和序贯学习分类的心血管内超声外弹力膜检测的改进算法。方法首先用多类多尺度序贯学习(M2SSL)将IVUS图像分割七大不同组织;然后在分类结果的基础上,结合血管先验形状信息筛选出外弹力膜轮廓的关键点;最后,结合IVUS图像的梯度和相位信息,采用Snake模型,获得最终的EEM轮廓。结果临床采集22组IVUS序列,挑选出具有代表性的153帧图像做实验。统计数据显示:本文算法检测结果的平均Jacc指标为88.5%,满足临床诊断要求,性能优于国内近年来较好的算法。结论本文的EEM自动检测算法简单有效,相比国内已有算法,提高了对钙化、纤维斑块以及声影区域的识别能力,对含钙化斑块、纤维斑块或血管中心偏移的高频IVUS图像具有较高的适用性。
Objective To evaluate the accuracy of contour detection of extra-elastic membrane (EEM) by interfering factors such as calcified plaque and acoustic shadow in intracardiac ultrasound (IVUS) images and to propose extracardiac extracorporeal elastography combined with a priori shape information and sequential learning classification Improved Algorithm for Membrane Detection. Methods Firstly, multislice multiscale sequential learning (M2SSL) was used to segment the IVUS images into seven different tissues. Based on the classification results, the key points of the contour of the elastic membrane were screened based on the prior information of vessel shape. Finally, Gradient and phase information, using the Snake model, to obtain the final EEM profile. Results The 22 sets of IVUS sequences were collected clinically and 153 representative images were selected for experiment. Statistics show that the average Jacc index of the algorithm in this paper is 88.5%, which meets the requirements of clinical diagnosis and has better performance than the better algorithm in recent years in China. Conclusion The EEM automatic detection algorithm in this paper is simple and effective. Compared with the existing algorithms in China, the EEM automatic detection algorithm improves the ability of recognizing calcifications, fibrous plaque and acoustic shadow areas, and high frequency of calcified plaque, fibrous plaque or blood vessel center IVUS image has high applicability.