Differentiating between Benign and Malignant non-Mass Enhancing Lesions in Breast DCE-MRI by Using Curvelet-based Textural Features

Breast DCE-MR imaging plays an important role in effective detection and diagnosis of breast cancer. Non-mass enhancing breast lesions have been less studied in CADx systems because of their challenging intrinsic. In this study, a CADx system is proposed for differentiating benign and malignant non-mass enhancing lesions in breast DCE-MRI. Proposed system uses dynamic information of the 4D DCE-MRI data to segment the lesions on the basis of a fuzzy clustering algorithm. Curvelet-based textural features are extracted from 3D segmented lesions and classified by SVM classifier. The results achieved the accuracy of 75% and AUC of 0.75 for non-mass enhancing breast lesions which provides comparable results to other recent methods.