<![CDATA[Segmentation and overlapped cells separation are important phases in microscopic image processing of breast cancer, because the accuracy of overlapped cells separation result determines the accuracy of breast cancer cell calculation. The amount of breast cancer cells is considered by doctor in determining the action towards patients. Two of the most common topics discussed in previous studies are the problem of increasing the accuracy of overlapped cancer cell separation result by calculating the number of cancer cell and over-segmentation problem. Compared to watershed method, clustering method produces higher accuracy in separating overlapped cancer cells. In this paper, a combination of Spatial Fuzzy C-Means (SFCM) and Rapid Region Merging (RRM) method is proposed to separate the overlapped cells and handling the over-segmentation problem. The input image of overlapped cells separation phase is the result of breast cancer cell identification by Gram-Schmidt (GS) method, while the clustered cancer cells are overlapped cancer cells which are detected based on the area of geometric feature. 40 microscopic breast cancer cells image of benign and malignant type is used as the datasets. The average value of Mean Square Error (MSE) for cell identification is 0.07 and the average accuracy of overlapped cells separation using SFCM and RRM is 78.41%.]]>
Copyrights © 2017
