Jain Nishant, Kumar Vinod
Department of Electrical Engineering, Indian Institute of Technology Roorkee, Roorkee, 247667, India.
J Med Syst. 2016 Nov;40(11):249. doi: 10.1007/s10916-016-0623-1. Epub 2016 Oct 4.
In this paper we have proposed an iterative Fuzzy C-Mean (IFCM) method which divides the pixels present in the image into a set of clusters. This set of clusters is then used to segment a focal liver lesion from a liver ultrasound image. Advantage of IFCM methods is that n-clusters FCM method may lead to non-uniform distribution of centroids, whereas in IFCM method centroids will always be uniformly distributed. Proposed method is compared with the edge based Active contour Chan-Vese (CV) method, and MAP-MRF method by implementing the methods on MATLAB. Proposed method is also compared with region based active contour region-scalable fitting energy (RSFE) method whose MATLAB code is available in author's website. Since no comparison is available on a common database, the performance of three methods and the proposed method have been compared on liver ultrasound (US) images available with us. Proposed method gives the best accuracy of 99.8 % as compared to accuracy of 99.46 %, 95.81 % and 90.08 % given by CV, MAP-MRF and RSFE methods respectively. Computation time taken by the proposed segmentation method for segmentation is 14.25 s as compared to 44.71, 41.27 and 49.02 s taken by CV, MAP-MRF and RSFE methods respectively.
在本文中，我们提出了一种迭代模糊C均值（IFCM）方法，该方法将图像中的像素划分为一组聚类。然后利用这组聚类从肝脏超声图像中分割出局灶性肝病变。IFCM方法的优点是n聚类FCM方法可能导致质心分布不均匀，而在IFCM方法中质心将始终均匀分布。通过在MATLAB上实现这些方法，将所提出的方法与基于边缘的主动轮廓Chan-Vese（CV）方法和MAP-MRF方法进行比较。所提出的方法还与基于区域的主动轮廓区域可缩放拟合能量（RSFE）方法进行比较，该方法的MATLAB代码可在作者网站上获得。由于没有在通用数据库上进行比较，因此在我们现有的肝脏超声（US）图像上比较了这三种方法和所提出方法的性能。与CV、MAP-MRF和RSFE方法分别给出的99.46%、95.81%和90.08%的准确率相比，所提出的方法给出了99.8%的最佳准确率。所提出的分割方法进行分割所需的计算时间为14.25秒，而CV、MAP-MRF和RSFE方法分别需要44.71秒、41.27秒和49.02秒。
