Application of artificial neural networks models on the classification of nodules in breast ultrasound images

Abstract

The increase in breast cancer cases has taken attention of many specialists in several fields. One of the consequences of this feature is the development of CAD schemes designed to aid in obtaining more accurate diagnosis in mammography evaluation. In this field the efforts in aiding the detection and characterization of nodules and/or microcalcifications in mammography images has been great. The ultrasound exam has been the main complementary exam to the X-ray mammography, often as way to provide a significant reduction in the number of unnecessary biopsies. As the breast ultrasound images have the aim of evaluate suspect masses which are hard to be characterized in the conventional mammography evaluation, this work purpose is to aid the detection and classification - according to the BI-RADS standards - of such masses on ultrasound digital images. The methodology is based in automated processing techniques involving computer vision and artificial neural networks. It will test phantoms, followed by analysis of the effect of the application on actual images.