Automatic detection and segmentation of breast lesions on MR images using 3D level set technique

Magnetic Resonance Imaging (MRI) is an often used imaging technique for the detection of breast cancer. This imaging technique offers the possibility to acquire three-dimensional data set and to highlight the tissues that have a characteristic pathological behaviour by using a contrast medium agent. We propose a method to overcome the time consuming operation needed to analyze these types of dataset. By using a region based level set algorithm we segment in the 3D space the regions of the images that have a time-enhancement curve compatible with the presence of a malignant lesion. By using this segmentation technique it is also possible to get volumetric information such as spatial position, mass volume and shape. The analysis performed on 12 unselected patients showed good results, all the lesions were detected and correctly segmented. In conclusion we propose a fast, automatic and accurate method for the detection and the segmentation of breast lesion in three-dimensional breast MR dataset.