变形模型在冠状动脉二维运动跟踪中的应用

Applications of Deformable Model in Tracking Coronary Arterial Dynamics from Xray Angiogram Sequence

本文提出一种采用变形模型对X射线造影图像序列中的冠状动脉血管进行运动跟踪的方法。首先采用多尺度高斯滤波对造影图像序列进行预处理,将血管区域从背景中提取出来,灰度图像被转化成二值图像。然后,采用改进的离散活动轮廓模型对图像序列进行感兴趣血管段的运动跟踪,其中模型的初始位置是自动获取的。通过在外部能量中引入对端点的灰度相似度和对中间点的运动相似度的测量,有效地避免了模型沿血管收缩或漂移的现象。同时为了保证模型的连续性和光滑性,降低变形结果对权重参数的敏感性,在每次变形结束后对模型进行重新采样。对临床得到的造影序列进行了实验,结果证明了方法的可行性。

An approach to track dynamic coronary arteries in X-ray angiogram sequence based on deformable model is developed. Firstly, multi-seale Gauss filtering is applied to segment vessel regions from projections, Then, a technology based on improved active contour model is used to automatically track the interesting vessel segment, labeled by start point and end point selected by operator in the beginning of the process, through model deforming from its initial pasition which is automatically located in our method. Measurements of intensity similarity on the end points and displacement similarity on other points are introduced into the external energy to avoid model shrinking and shift along the object during deforming, which are common in open snake. A resampling process is implemented after each deformation to obtain continuous and smooth models with an effect of decreasing the sensitivity to weighting parameters, Results of applying the method to clinical data are presented.