摘要
在计算机辅助心血管诊断中,其中最重要的步骤就是对超声心动视频的分类。因此提出了基于LSOE和AK-SVM的超声心动视频分类算法。首先为了克服超声心动视频本身存在的低对比度和快速不规则运动问题,提出基于局部时空方向能量(LSOE)对超声心动视频的特征进行提取并将其转换为可描述的中间文件,接着为得到最好的分类效果提出了加性核SVM对其进行学习和分类并得出其分类的准确率。在文献[1]超声心动视频数据集上取得了目前最好的分类效果,超声心动影像的平均分类准确率为75.37%。
In echo cardiac clinical computer-aided diagnosis,the important step is toautomatically classify echocardiography videos. We propose a kind of echocardiography video classification algorithm based on Local Spatio-temporal Oriented Energy( LSOE) and Additive Kernel SVM( AK-SVM). First of all,we propose the Local Spatio-temporal Oriented Energy in echocardiography sequence and get the video features and convert the described file. It can overcome the influence of the rapid and irregular movement of the echocardiography video and get the more robust tracking results. Then we propose the Additive Kernel SVM for classification and get the classified accuracy. This paper get the best classification results in the data of echocardiography video in literature[1]. The average classification accuracy of echocardiography video is 75. 37%.
出处
《南阳理工学院学报》
2016年第2期44-49,共6页
Journal of Nanyang Institute of Technology
基金
国家自然科学基金(61471124)
关键词
计算机辅助诊断
超声心动视频
方向能量
加性核SVM
分类
computer-aided diagnosis
echocardiography video
oriented energy
additive kernel SVM
classification