Structured crowdsourcing enables convolutional segmentation of histology images

• Type of Material: Article
• Authors: Mohamed Amgad, Emory University; Habiba Elfandy, Cairo University; Hagar Hussein, Cairo University; Lamees A. Atteya, Egyptian Ministry of Health; Mai A T Elsebaie, Ain Shams University; Lamia S. Abo Elnasr, Menoufia University; Rokia A. Sakr, Menoufia University; Hazem S E Salem, Ain Shams University; Ahmed F Ismail, Alexandria University; Anas M. Saad, Ain Shams University; Joumana Ahmed, Cairo University; Maha A T Elsebaie, Ain Shams University; Mustafijur Rahman, University of Chittagong; Inas A. Ruhban, Damascus University; Nada M. Elgazar, Mansoura University; Yahya Alagha, Cairo University; Mohamed H. Osman, Zagazig University; Ahmed M. Alhusseiny, Mansoura University; Mariam M. Khalaf, Batterjee Medical College; Abo-Alela F. Younes, Ain Shams University; Ali Abdulkarim, Cairo University; Dua M. Younes, Ain Shams University; Ahmed M. Gadallah, Ain Shams University; Ahmad M. Elkashash, Cairo University; Salma Y. Fala, Suez Canal University; Basma M. Zaki, Suez Canal University; Jonathan Beezley, Kitware Inc; Deepak R. Chittajallu, Kitware Inc; David Manthey, Kitware Inc; David Gutman, Emory University; Lee Cooper, Emory University
• Language: English
• Date: 2019-09-15
• Publisher: Oxford University Press
• Publication Version: Final Published Version
• Copyright Statement: © 2019 The Author(s) 2019. Published by Oxford University Press.
• License: Creative Commons Attribution 4.0 International
• Final Published Version (URL): http://dx.doi.org/10.1093/bioinformatics/btz083
• Title of Journal or Parent Work: Bioinformatics
• Volume: 35
• Issue: 18
• Start Page: 3461
• End Page: 3467
• Grant/Funding Information: This work was supported by the U.S. National Institutes of Health; and National Cancer Institute grants [U01CA220401, U24CA194362].
• Keywords: DIGITAL SLIDE ARCHIVE; Computer Science, Interdisciplinary Applications; Statistics & Probability; Physical Sciences; Life Sciences & Biomedicine; Mathematical & Computational Biology; Biochemical Research Methods; Science & Technology; Technology; Biotechnology & Applied Microbiology; Mathematics; Biochemistry & Molecular Biology; Computer Science
• Research Categories: Biology, Bioinformatics

Abstract

While deep-learning algorithms have demonstrated outstanding performance in semantic image segmentation tasks, large annotation datasets are needed to create accurate models. Annotation of histology images is challenging due to the effort and experience required to carefully delineate tissue structures, and difficulties related to sharing and markup of whole-slide images. Results: We recruited 25 participants, ranging in experience from senior pathologists to medical students, to delineate tissue regions in 151 breast cancer slides using the Digital Slide Archive. Inter-participant discordance was systematically evaluated, revealing low discordance for tumor and stroma, and higher discordance for more subjectively defined or rare tissue classes. Feedback provided by senior participants enabled the generation and curation of 20 000+ annotated tissue regions. Fully convolutional networks trained using these annotations were highly accurate (mean AUC=0.945), and the scale of annotation data provided notable improvements in image classification accuracy. Availability and Implementation: Dataset is freely available at: https://goo.gl/cNM4EL. Supplementary information: Supplementary data are available at Bioinformatics online.