Herramienta para la segmentación de microcalcificaciones e identificación de riegos de carcinoma en imágenes mamográficas haciendo uso de inteligencia artificial.

Abstract

In this research project explains and details the construction of a CAD that by means of Mask R-CNN semantically segments microcalcifications present in mammograms, additionally based on the spatial distribution of these lesions wit hin the breast, the severity of the injury classifying it as benign or malignant. In order to reduce the computational crap of the system, the RestNet50 pre-trained model was used, with resized and compressed images obtained from the CBIS DSSM database.