基于流形学习的血管内超声图像序列关键帧的提取及应用

发布时间：2019-02-27 17:54

【摘要】：目的以流形学习为基础,提出一种基于血管内超声(IVUS)图像序列的关键帧门控方法,抑制IVUS序列纵切方向上的运动伪影。方法应用流形学习方法中的拉普拉斯特征映射算法,将高维IVUS图像序列降到低维流形中,利用低维特征向量,构建一个距离函数来反映心脏运动规律,将IVUS图像分为心脏舒张末期和非心脏舒张末期两类,从而提取关键帧,组成门控序列。结果临床采集13组IVUS序列(图像915±142帧,血管长度15.24±2.37 mm),计算门控前后图像序列的血管容积、管腔容积和平均斑块负荷。统计实验结果,表明门控序列血管容积、管腔容积显著小于原始序列,门控前后序列的平均斑块负荷差异性不显著,满足临床诊断要求。血管面积方差和管腔面积的方差显著小于原始序列,表明门控序列较原始序列稳定。在IVUS图像序列的纵切图像上,门控序列减少了锯齿形状的运动伪影,与原始序列形状一致,且具有良好的连续性。并将本文方法与已有的提取门控序列方法进行对比。结论本文方法算法简单稳定,抑制了IVUS图像序列的纵向运动伪影。
[Abstract]:Aim based on manifold learning, a key frame gating method based on intravascular ultrasound (IVUS) image sequences is proposed to suppress motion artifacts in the longitudinal direction of IVUS sequences. Methods the Laplacian feature mapping algorithm of manifold learning method was used to reduce the high dimensional IVUS image sequence to the low dimensional manifold. A distance function was constructed to reflect the law of cardiac motion by using the low dimensional feature vector. IVUS images were divided into two types: end-diastolic and non-cardiac end-diastolic, then the key frames were extracted and the gating sequence was formed. Results the blood vessel volume, lumen volume and mean plaque load were calculated in 13 groups of IVUS sequences (915 卤142frames, 15.24 卤2.37mm),) before and after gating. The statistical results showed that the vessel volume and lumen volume of the gated sequence were significantly lower than those of the original sequence. The average plaque load of the gated sequence was not significantly different before and after gating, which met the requirements of clinical diagnosis. The variance of vascular area and lumen area was significantly smaller than that of the original sequence, indicating that the gated sequence was more stable than the original sequence. On the longitudinal image of the IVUS image sequence, the gate sequence reduces the motion artifact of the serrated shape, which is consistent with the original sequence shape and has good continuity. The method of this paper is compared with the existing method of extracting gate-controlled sequence. Conclusion the algorithm presented in this paper is simple and stable, and the longitudinal motion artifact of IVUS image sequence is suppressed.
【作者单位】：南方医科大学生物医学工程学院;南方医科大学南方医院心内科;
【基金】：国家自然科学基金(61271155)~~
【分类号】：R445.1;R54

【参考文献】