A fusion of fuzzy clustering and neural networks-based diagnosis of breast cancer by mammography micro-calcification

Abstract

Detecting micro-calcifications in mammography images is a crucial step in the early diagnosis of breast cancer, and enhancing image quality through pre-processing plays a vital role. In this research, we propose a fuzzy clustering algorithm-based approach to accurately identify micro-calcifications in mammography images. To address the low quality of mammograms, pre-processing techniques were applied to improve image clarity. A suitable membership function was defined to identify calcified regions using fuzzy clustering. The detected regions were then compared with those identified by specialists, achieving an improved identification accuracy of 96.7% and a sensitivity of 97.2%, compared to previous methods which had 95% accuracy and 90.5% sensitivity. In the next phase, neural networks were employed to classify the extracted regions into benign and malignant categories. Diagnostic performance was evaluated using identification accuracy, sensitivity, and positive and negative predictive values. The proposed method demonstrated positive results with an identification accuracy of 97.5%, sensitivity of 98.1%, positive predictive value of 98.3%, and negative predictive value of 96.3%. These outcomes indicate that the proposed fusion of fuzzy clustering and neural networks enhances diagnostic precision, owing to its high accuracy in region extraction and the distinctive features it identifies.