Breast Tumors Classification Using Adaptive Neuro-Fuzzy Inference System

Abstract

Breast cancer is currently the foremost cause in women’s mortality worldwide. In Sudan, the increasing incidence, detection at late stages and the early onset of the disease makes early detection and diagnosis of breast cancer an overbearing task. Currently, X-ray mammography is the single most effective, low-cost, and highly sensitive technique for detecting small lesions. However, the sensitivity of mammography is highly challenged by the presence of dense breast parenchyma, which deteriorates both detection and characterization tasks. Thus, there is a significant necessity for developing methods for automatic detection and classification of suspicious areas in mammograms, as a means of aiding radiologists to improve the efficacy of screening programs and avoid unnecessary biopsies. The objective of this study is to create a computer interfacing system for the localization, detection and classification of breast masses using ANFIS. The ANFIS classifier was used to detect the breast cancer when five features defining breast cancer indications were used as inputs. The proposed ANFIS model combined the neural network adaptive capabilities and the fuzzy logic qualitative approach. Results demonstrated the proposed methodologies have high potential in enhancing breast images, localizing, detecting and classifying the breast tumor. The system was able to achieve an accuracy of 94.4% sensitivity, 100% specificity, 97.1% positive predictive value, 100% negative predictive value, an Az value of 0.972 and an overall classification accuracy of 98%. The created systems have therefore, proved to effectively enhance the quality of the breast images and discriminate between malignant and benign tumors with an effective level of precision.