基于水平集方法的冠状动脉CT图像分割

Coronary CT image segmentation based on level set method

目的采用水平集方法实现对CT图像中冠状动脉的高效自动识别与分割,借此简化三维模型重建工作,从而提高使用冠脉CT造影计算的血流储备分数(computed tomography derived fractional flow reserve,FFRct)整体的计算效率。方法选取20名冠心病患者作为研究对象,提取26支狭窄冠脉血管的CT图像。分别利用水平集方法和手动方法对CT图像进行冠脉的识别与分割处理从而生成三维模型。利用计算流体力学(computational fluid dynamics,CFD)仿真方法分别获取各冠脉段的血流储备分数(fractional flow reserve,FFR)数值,并通过Bland-Altman方法分析两种方法所得FFR的一致性,用以验证水平集方法在冠脉识别精度上与手动方法的差异。结果对比了两种冠脉识别方法获取FFR数值整体所需的时间。与手动方法相比,水平集方法的模型重建时间减少了约10倍。经检验,两种冠脉识别方法的FFR数值计算结果一致性良好。结论基于水平集方法可以实现冠脉CT图像的高效自动识别,具有一定的临床应用意义。

Objective The level set method is used to realize efficient automatic identification and segmentation of coronary arteries in CT images,thereby simplifying the reconstruction of the three-dimensional model and improving the overall calculation efficiency of FFRct.Methods Twenty patients with coronary heart disease were selected as the research objects,and CT images of 26 narrow coronary vessels were extracted.The level set method and manual method were used to identify and segment the coronary artery on CT images to generate a three-dimensional model.Computational fluid dynamics(CFD)simulation method was used to obtain the FFR value of each coronary artery segment,and the consistency of the FFR obtained by the two methods was analyzed by the Bland-Altman method to verify the difference between the level set method and the manual method in the coronary recognition accuracy.Results The overall time required to obtain the FFR value of the two coronary artery identification methods was compared.Compared with the manual method,the model reconstruction time of the level set method was reduced by more than 10 times.After testing,the FFR numerical calculation results of the two coronary artery identification methods were in good agreement.Conclusions The level set method can realize the high-efficiency automatic identification of coronary artery in CT images,which has certain clinical application significance.