An optimization approach to segment breast lesions in ultra-sound images using clinically validated visual cues

摘要: As long as breast cancer remains the leading cause of deaths among female population world wide, developing tools to assist radiologists during diagnosis process is necessary. However, most technologies developed in imaging laboratories are rarely integrated this assessing process, they based on information cues differing from those used by clinicians. In order grant Computer Aided Diagnosis (CAD) systems with these when performing non-aided diagnosis, better segmentation strategies needed automatically produce accurate delineations structures. This paper proposes a highly modular and flexible framework for segmenting tissues lesions present Breast Ultra-Sound (BUS) images. relies an optimization strategy high-level de-scriptors designed analogously visual radiologists. The methodology comprehensively compared other sixteen published methodologies BUS proposed achieves similar results than reported state-of-the-art.