Filtering, segmentation and feature extraction in ultrasound evaluation of breast lesions

Abstract

The early diagnosis of breast cancer depends in many cases on the analysis of medical imaging, mainly mammography, ultrasonography and MRI. This work deals with ultrasound images in order to reduce speckle noise, identify the contour of the nodules and analyze a wide range of criteria which allow distinguishing between benign and malignant tumors. We try to automatize the different phases of the process and extract some objective parameters for a robust and reproducible diagnosis. We provide the physicians with both graphical as well as numerical results for the features which have been analyzed.