基于FCM聚类算法与自适应阈值的冠脉钙化分割与量化

Segmentation and Scoring of Coronary Calcium in 3D CTA Data Based on FCM Algorithm and Self-Adapting Threshold Determination

针对基于CTA图像进行冠脉钙化量化时存在的无法克服噪声以及阈值选择不稳定问题,提出一种基于聚类算法与自适应阈值的冠脉钙化分割与量化方法。首先根据CT值和空间位置对冠脉血管内的像素点构建特征向量,继而根据血管骨架点数目构建自适应聚类数,使用模糊C均值(FCM)聚类算法将冠脉区域划分为CT值分布相似的区域;然后使用高斯函数拟合冠脉灰度直方图,根据高斯拟合参数构造自适应阈值,对上述区域进行钙化分割;最后根据分割结果,参考Agatston钙化分量化标准进行钙化分计算。在30组人体冠脉CTA数据的测试结果中,对冠脉钙化量化的灵敏度和特异性分别达到89.5%与98.6%,计算得到的钙化体积和Agatston钙化分与标准结果的皮尔逊系数分别为0.974与0.975,远高于同类型基于一阶微分进行阈值选择方法(DBTD)对应的0.523与0.501。实验结果表明,该方法可用于冠脉钙化分割与量化,且具有全自动、鲁棒性好、能有效抗噪等特点。

Aiming at the problems of image noise and impressionable threshold in coronary calcium quantification with 3D CTA data, a new method based on fuzzy C-means （FCM） clustering algorithm and self- adapting threshold determination was proposed for automatically segmenting and quantifying the coronary calcium. Firstly, feature vectors are constructed for every voxel in the coronary artery, which contains spatial coordinates and CT value information of the voxel, and a clustering algorithm combined with a self-adapted group number referring to vessel skeleton number is used to divide the coronary artery into different candidate volumes; secondly, a robust threshold determination algorithm based on the histogram is used to extract calcium plaques among those candidate volumes acquired; finally, calcium volume and Agatston score are calculated. Result shows that the method proposed in this study has relatively high sensitivity of 89.5% and specificity of 98.6% in calcium detection in 30 coronary CTA data. The calcium volume and Agatston score calculated automatically show a high correlation with the standard result. The corresponding Pearson correlation coefficient being up to 0. 974 and 0. 975, respectively, much higher than the 0. 523 and 0. 501 that calculated by the derivative-based threshold determination （DBTD） method. Experimental results show that this method can be used for coronary calcium segmentation and quantification, and has the characteristics of full automation,robustness and noise immunity.