Segmentation and Classification of Breast Cancer Fine Needle Aspiration Biopsy Cytology Images

Abstract

The proposed system efficiently predicts breast tumour
from microscopic cytology images through image processing techniques coupled with support vector machine (SVM) classifier as
either benign or malignant. The breast cytology image is denoised using non-linear Anisotropic diffusion method to remove random noise prevalent in cytology images. In the cytology images only nucleus is the region of interest for the detection of breast cancer.
Hence, all the nuclei in the cytology image are segmented using seeded region growing (SRG) method. Geometric features of every cell nuclei are extracted. These features are then used in conjunction
with SVMs that classifies breast tumour as cancerous or
non-cancerous. The proposed system implemented on MATLAB takes less than 1 minutes of processing time and has yielded promising results that would supplement in the diagnosis of breast cancer.