Date

This work was published as an extended abstract for the IWBI2020 workshop. It was carried out during a three month MSc internship at Screenpoint Medical under supervision of Jaap Kroes and Michiel Kallenberg.

Read the paper here. See the presentation here (2nd presentation in session).

Abstract

Semantic segmentation of breast images is typically performed as a preprocessing step for breast cancer detection by Computer Aided Diagnosis (CAD) systems. While most literature on region segmentation is based on conventional techniques like line estimation, thresholding and atlas-based approaches, such methods may have problems with generalisation. This paper investigates a robust multi-vendor breast region segmentation system for full field digital mammograms (FFDM) and digital breast tomography (DBT) using a U-Net neural network. Additionally, the effect of adding attention gates to the U-Net architecture was analysed. The proposed networks were trained and tested in a cross-validation setting on in-house FFDM/DBT data and the public INbreast datasets, comprising over 10.000 FFDM and 3.500 DBT images from five different vendors. Dice scores were obtained in the range 0.978 -- 0.985, with slightly higher scores for the architecture that includes attention gates.