End-to-End Breast Mass Classification on Digital Breast Tomosynthesis

Chao Zheng, TianLan Mo

Keywords: Breast, Mammography, Deep Learning, Neural Network, Classification

TL;DR: A End-to-End multi-scale multi-level features fusion Network model for breast mass classication using DBT

Track: short paper

Abstract: Automatic classification of the masses in digital breast tomosynthesis(DBT) is still a big challenge and play a crucial role to assist radiologists for accurate diagnosis.In this paper,we develop a End-to-End multi-scale multi-level features fusion Network (EMMFFN) model for breast mass classication using DBT. Three multifaceted representations of the breast mass (gross mass, overview, and mass background) are extracted from the ROIs and then, fed into the EMMFFN model at the same time to generated three sets of feature map.The three feature maps are finally fused at the feature level to generate the final prediction.Our result show that the EMMFFN model achieves a breast mass classification AUC of 85.09%,which was superior to the single submodel who only use one aspect of patch.

Paper Type: well-validated application