MRI Breast Image Segmentation Using Artificial Bee Colony Optimization with Fuzzy Clustering and CNN Classifier

Abstract

Breast cancer is the most leading disease for women and many lost their lives without proper treatment and early diagnosis helps to reduce the death rate. In clinical procedure identifying the tumor take much processing time and extracting the abnormal part leads error at some time. To overcome this problem and helps the radiologist an automated framework with the integration of artificial bee colony and fuzzy C means clustering for extracting the cancerous part and classify the tumor as benign and malignant with CNN classification. Various objective measures like entropy, eccentricity, contrast, correlation, homogeneity, mean, variance, kurtosis, dice similarity index, hands-off distance and subjective measures like precision, recall, accuracy, MSE, PSNR were evaluated to ensure the efficiency and accuracy of segmentation and classification of MRI breast images with T1, T2—contrast enhanced images. Our hybrid approach yield 98% segmentation accuracy and 98.6% classification accuracy and it take average of 5 s for processing. We utilized public dataset like DDSM, MIAS and INbreast dataset for validation.