基于形变模型和微分同胚配准的超声应变分析

Strain estimation of ultrasound images based on modified hybrid deformable model and diffeomorphic non-rigid registration

目的通过改进的三维形变混合模型分割算法及微分同胚非刚性配准算法对三维超声图像进行应变分析,以得到稳定可靠的超声应变图像。方法首先,使用各向异性扩散滤波器进行降噪处理,再使用二维投影的区域生长算法对左心室进行粗分割,然后通过形态学操作得到较为精确的分割。在此基础上使用三维参数形变模型,实现精确的心脏图像分割。最后以心动周期内心室第一帧图像的三维分割为参考模型,利用微分同胚非刚性配准算法对后续图像配准,获取心室位移场,并进行应变估计。结果实验结果显示,该方法具有很高的效率和准确率。结论本方法得到的应变分析,符合心脏解剖学特征,对辅助诊断心室疾病具有一定意义。

Objective In order to obtain robust and reliable ultrasound strain images,this paper presents an improved method of strain estimation from 3 D ultrasound images by combining 3 D hybrid deformable model and diffeomorphic non-rigid registration.Methods First,we use an anisotropic filtering method to filter out noises,and region growing algorithm based on 2D projection to achieve an initial rough segmentation.Then we apply morphologic closing computation for the more precise results of region growing.In the next step,we achieve accurate cardiac segmentation by deforming the 3 D deformable model.We choose the result segmented by the first frame image from 3 D echocardiography in a cycle to register other sequence images based on diffeomorphic non-rigid registration method,and subsequently calculate the displacement field and estimate the strain state.Results The experiments show that the proposed method is efficient and accurate.Conclusions The strain analysis of the proposed method is consistent with the property of cardiac anatomy,and is helpful in the diagnosis of cardiac diseases.