2020/06/25 by Saad Nadeem, Nadeem, Saad, Travis J. Hollmann +3 · 1 citation
Biochemistry, Genetics and Molecular Biology · Computer Science · #AI in cancer detection #Computer Vision and Pattern Recognition (cs.CV) #Digital Imaging for Blood Diseases #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Molecular Biology Techniques and Applications #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2006.14566
openalex publication_date 2020/06/25 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Variations in hematoxylin and eosin (H&E) stained images (due to clinical lab\nprotocols, scanners, etc) directly impact the quality and accuracy of clinical\ndiagnosis, and hence it is important to control for these variations for a\nreliable diagnosis. In this work, we present a new approach based on the\nmultimarginal Wasserstein barycenter to normalize and augment H&E stained\nimages given one or more references. Specifically, we provide a mathematically\nrobust way of naturally incorporating additional images as intermediate\nreferences to drive stain normalization and augmentation simultaneously. The\npresented approach showed superior results quantitatively and qualitatively as\ncompared to state-of-the-art methods for stain normalization. We further\nvalidated our stain normalization and augmentations in the nuclei segmentation\ntask on a publicly available dataset, achieving state-of-the-art results\nagainst competing approaches.\n