Segmentation of Breast Ultrasound image using Regularized K-Means (ReKM) clustering
E. S. Samundeeswari, P.K. Saranya, Radhakrishnan Manavalan · 2016
Breast cancer is a highly heterogeneous disease and very common among women worldwide. Inter-observer and intra-observer errors occur frequently in analyzing the lesion portion of medical images, giving high variability in results interpretations. Computer Aided Diagnosis system (CAD) plays a vital role to overcome this variability. Segmentation is the second critical stage in CAD system to extract the desired portion exactly for distinguishing benign tumor from malignant one. In this work, the traditional K-Means algorithm is incorporated with Ant Colony Optimization and Regularization parameter to segment the lesion portion with maximum boundary preservation. The PRI, VoI, GCE and BDE cluster validation measures are used to compare the segmented result with ground truth image delineated by the radiologist. The proposed work outperforms the traditional K-Means clustering method with 96% similarity (PRI) between segmented tumor image with referred tumor image.