Joint Breast Neoplasm Detection and Subtyping using Multi-Resolution Network Trained on Large-Scale H&E Whole Slide Images with Weak Labels

Adam Casson, Siqi Liu, Ran A Godrich, Hamed Aghdam, Donghun Lee, Kasper Malfroid, Brandon Rothrock, Christopher Kanan, Juan Retamero, Matt Hanna, Ewan Millar, David Klimstra, Thomas Fuchs

Keywords: Digital pathology, Deep learning, Breast neoplasm, Breast cancer, Weakly supervised learning

Abstract: Breast cancer is the most commonly diagnosed cancer and the use of artificial intelligence (AI) to help diagnose the disease from digital pathology images has the potential to greatly improve patient outcomes. However, current methods for detecting, segmenting, and sub-typing breast neoplasms and other proliferative lesions often rely on costly and time-consuming manual annotation efforts, which can be impractical for large-scale datasets. In this work, we propose an annotation-free learning framework to jointly detect, segment, and subtype breast neoplasms. Our approach leverages top-k multiple instance learning to train an initial neoplasm detection backbone network from weakly-labeled whole slide images, which is then used to automatically generate pixel-level pseudo-labels for whole slides with only one subtype. A second network is trained using these pseudo-labels, and slide-level classification is performed by training an aggregator network that fuses the embeddings from both backbone networks. We trained and validated our framework on large-scale datasets with more than 100k whole slide images and demonstrate its effectiveness on tasks including breast neoplasms detection, segmentation, and subtyping.