Skip to content

Success stories

Developing machine-learning tools through collaborative annotation studies

The primary determinant in machine-learning algorithm performance is the availability of abundant labeled data for training. Collecting data for applications such as tissue-region segmentation and cell classification is challenging, given the limited availability of experts and of tools to collect and review annotations.

The DSA web platform enables users to collaborate across the world and has been used to generate more than 120,000 human markups of histopathology in multiple annotation studies.

Its API and user roles allow collaborative annotation study teams that engage users with different expertise levels, from pathologists to medical students, to generate large and extensively reviewed annotation datasets. In one study with more than 25 participants from over 5 countries, the team generated more than 25,000 annotated tissue regions and produced the richest public dataset of annotated breast-cancer tissues at the time. Programmatic monitoring and management of these studies through the API is key to their success.

Data on this scale supports accurate machine-learning models for tasks such as tissue segmentation, cell classification, and cell detection, and helps measure the limits of concordance among human experts. The data from the collaborative annotation study is available on a demo instance of the Digital Slide Archive.

View the dataset visualization

References

  • Chandradevan, Ramraj, Ahmed A. Aljudi, Bradley R. Drumheller, Nilakshan Kunananthaseelan, Mohamed Amgad, David A. Gutman, Lee A. D. Cooper, and David L. Jaye. "Machine-Based Detection and Classification for Bone Marrow Aspirate Differential Counts: Initial Development Focusing on Nonneoplastic Cells." Laboratory Investigation 100, no. 1 (September 30, 2019): 98–109. doi:10.1038/s41374-019-0325-7 · PMID 31570774
  • Amgad, Mohamed, Habiba Elfandy, Hagar Hussein, Lamees A Atteya, Mai A T Elsebaie, Lamia S Abo Elnasr, Rokia A Sakr, et al. "Structured Crowdsourcing Enables Convolutional Segmentation of Histology Images." Edited by Robert Murphy. Bioinformatics 35, no. 18 (February 6, 2019): 3461–67. doi:10.1093/bioinformatics/btz083 · PMID 30726865
  • Mobadersany, Pooya, Safoora Yousefi, Mohamed Amgad, David A. Gutman, Jill S. Barnholtz-Sloan, José E. Velázquez Vega, Daniel J. Brat, and Lee A. D. Cooper. "Predicting Cancer Outcomes from Histology and Genomics Using Convolutional Networks." Proceedings of the National Academy of Sciences 115, no. 13 (March 12, 2018): E2970–79. doi:10.1073/pnas.1717139115 · PMID 29531073

The Digital Slide Archive and TCGA

The Cancer Digital Slide Archive (CDSA) hosts whole-slide images of pathologic specimens associated with The Cancer Genome Atlas (TCGA) studies. At more than 10,000 images and 20 terabytes, the CDSA has enabled a number of TCGA Analysis Working Group studies, as well as other investigator-driven studies that use the digital pathology images.

The Digital Slide Archive builds on that experience, creating a software platform that lets users run their own digital pathology servers for data management and hosting in their own investigations.

View the CDSA demo site

References

  • Abeshouse, Adam, Clement Adebamowo, Sally N. Adebamowo, Rehan Akbani, Teniola Akeredolu, Adrian Ally, Matthew L. Anderson, et al. "Comprehensive and Integrated Genomic Characterization of Adult Soft Tissue Sarcomas." Cell 171, no. 4 (November 2017): 950–965.e28. doi:10.1016/j.cell.2017.10.014 · PMID 29100075
  • Gutman, David A., Mohammed Khalilia, Sanghoon Lee, Michael Nalisnik, Zach Mullen, Jonathan Beezley, Deepak R. Chittajallu, David Manthey, and Lee A.D. Cooper. "The Digital Slide Archive: A Software Platform for Management, Integration, and Analysis of Histology for Cancer Research." Cancer Research 77, no. 21 (October 31, 2017): e75–78. doi:10.1158/0008-5472.CAN-17-0629 · PMID 29092945
  • Gutman, David A, Jake Cobb, Dhananjaya Somanna, Yuna Park, Fusheng Wang, Tahsin Kurc, Joel H Saltz, Daniel J Brat, Lee A D Cooper, and Jun Kong. "Cancer Digital Slide Archive: An Informatics Resource to Support Integrated in Silico Analysis of TCGA Pathology Data." Journal of the American Medical Informatics Association 20, no. 6 (November 2013): 1091–98. doi:10.1136/amiajnl-2012-001469 · PMID 23893318
Source and version

These stories are consolidated from the upstream project website (docs/_success_stories/) on 2026-11-01. Wording is reproduced from the upstream pages.