基于FGCM的冠脉OCT图像纤维斑块分割算法

冠脉OCT图像中的纤维斑块精确分割对于冠心病的诊断具有重大意义。针对模糊C均值算法邻域信息和空间信息利用率低,在分割具有弱边界的纤维化斑块过程中容易产生过分割的情况,提出了基于自定义窗口的邻域信息项和具有全局约束项G的FCM算法,即FGCM算法。首先,对OCT图像进行预处理,去除噪声。然后,采用自定义窗口获取像素强度的邻域信息,并将邻域信息和创造性提出的全局约束项G添加到FCM算法,使用FGCM算法分割纤维化斑块。最后,采用数学形态学方法作为后处理操作平滑斑块边界。通过对具有典型纤维化斑块特征的8个病例进行测试,与医生手动标记纤维化斑块(金标准)和其他文献的算法进行对比,测试结果显示所提算法分割精确度进一步提高,达到90%,能够进一步对血管狭窄程度进行定量分析,辅助医生制定诊断方案。

Fiber plaque segmentation algorithm based on FGCM for coronary OCT image

Accurate segmentation of fiber plaques in coronary OCT images is great significance for the diagnosis of coronary heart disease.For the fuzzy C-means algorithm,the neighborhood information and spatial information utilization rate are low,and it is easy to generate over-segmentation in the process of segmenting fiber plaques with weak boundaries.The neighborhood information item of custom window and FCM algorithm with global constraint item G are proposed,namely FGCM algorithm.Firstly,the OCT image is preprocessed to remove noise.Then,the neighborhood window information of the pixel intensity is obtained by custom window,the neighborhood information and the creatively proposed global constraint term G are added to FCM algorithm,the fiber plaque is segmented using the FGCM algorithm.