COMPUTERIZED DIAGNOSIS OF BREAST LESIONS ON ULTRASONOGRAPHY

In this article, a computer-aided diagnosis (CAD) system for breast ultrasound is proposed. The system consists of three main stages: lesion segmentation, feature selection, and classification. The central characteristic of the CAD system is the classifier learning algorithm. Furthermore, a differential evolution technique was used to optimize the structure of a radial basis function (RBF) neural network in terms of number of hidden units and position of the RBFs centers. The dataset consisted of 641 breast ultrasound images: 228 carcinomas and 413 benign masses. Based on the ROC analysis the performance of the proposed CAD was assessed, resulting in a sensibility of 90.2%, specificity of 90.9% and accuracy of 90.4%. Moreover, it was achieved a 0.97 for the area under the ROC curve.