Segmentació de mamografies utilitzant tècniques d'aprenentatge profund

Abstract

[en] CADe and CADx (computer-aided detection and computer-aided diagnosis) systems are designed to assist medical professionals in quickly analyzing and evaluating information obtained through X-rays, magnetic resonance imaging (MRI), ultrasounds, among others. These systems combine elements of computer vision and artificial intelligence with medical imaging techniques. An important field of work for these systems is the analysis of mammograms to aid in the diagnosis of breast cancer. The objective of this work is to develop a mammogram segmentation system using deep learning, specifically the U-Net neural network architecture. To accomplish this, the publicly available CBIS-DDSM dataset is utilized, which is one of the largest and widely employed datasets in the field of mammography to validate new automatic segmentation methods.