Artificial intelligence (AI) can predict the presence of molecular alterations directly from routine histopathology slides. However, training robust AI systems requires large datasets for which data collection faces practical, ethical and legal obstacles. These obstacles could be overcome with swarm learning (SL), in which partners jointly train AI models while avoiding data transfer and monopolistic data governance. Here, we demonstrate the successful use of SL in large, multicentric datasets of gigapixel histopathology images from over 5,000 patients. We show that AI models trained using SL can predict BRAF mutational status and microsatellite instability directly from hematoxylin and eosin (H&E)-stained pathology slides of colorectal cancer. We trained AI models on three patient cohorts from Northern Ireland, Germany and the United States, and validated the prediction performance in two independent datasets from the United Kingdom. Our data show that SL-trained AI models outperform most locally trained models, and perform on par with models that are trained on the merged datasets. In addition, we show that SL-based AI models are data efficient. In the future, SL can be used to train distributed AI models for any histopathology image analysis task, eliminating the need for data transfer.
%0 Journal Article
%1 Saldanha2022-jc
%A Saldanha, Oliver Lester
%A Quirke, Philip
%A West, Nicholas P
%A James, Jacqueline A
%A Loughrey, Maurice B
%A Grabsch, Heike I
%A Salto-Tellez, Manuel
%A Alwers, Elizabeth
%A Cifci, Didem
%A Ghaffari Laleh, Narmin
%A Seibel, Tobias
%A Gray, Richard
%A Hutchins, Gordon G A
%A Brenner, Hermann
%A van Treeck, Marko
%A Yuan, Tanwei
%A Brinker, Titus J
%A Chang-Claude, Jenny
%A Khader, Firas
%A Schuppert, Andreas
%A Luedde, Tom
%A Trautwein, Christian
%A Muti, Hannah Sophie
%A Foersch, Sebastian
%A Hoffmeister, Michael
%A Truhn, Daniel
%A Kather, Jakob Nikolas
%D 2022
%I Springer Science and Business Media LLC
%J Nat. Med.
%K
%N 6
%P 1232--1239
%T Swarm learning for decentralized artificial intelligence in cancer histopathology
%V 28
%X Artificial intelligence (AI) can predict the presence of molecular alterations directly from routine histopathology slides. However, training robust AI systems requires large datasets for which data collection faces practical, ethical and legal obstacles. These obstacles could be overcome with swarm learning (SL), in which partners jointly train AI models while avoiding data transfer and monopolistic data governance. Here, we demonstrate the successful use of SL in large, multicentric datasets of gigapixel histopathology images from over 5,000 patients. We show that AI models trained using SL can predict BRAF mutational status and microsatellite instability directly from hematoxylin and eosin (H&E)-stained pathology slides of colorectal cancer. We trained AI models on three patient cohorts from Northern Ireland, Germany and the United States, and validated the prediction performance in two independent datasets from the United Kingdom. Our data show that SL-trained AI models outperform most locally trained models, and perform on par with models that are trained on the merged datasets. In addition, we show that SL-based AI models are data efficient. In the future, SL can be used to train distributed AI models for any histopathology image analysis task, eliminating the need for data transfer.
@article{Saldanha2022-jc,
abstract = {Artificial intelligence (AI) can predict the presence of molecular alterations directly from routine histopathology slides. However, training robust AI systems requires large datasets for which data collection faces practical, ethical and legal obstacles. These obstacles could be overcome with swarm learning (SL), in which partners jointly train AI models while avoiding data transfer and monopolistic data governance. Here, we demonstrate the successful use of SL in large, multicentric datasets of gigapixel histopathology images from over 5,000 patients. We show that AI models trained using SL can predict BRAF mutational status and microsatellite instability directly from hematoxylin and eosin (H\&E)-stained pathology slides of colorectal cancer. We trained AI models on three patient cohorts from Northern Ireland, Germany and the United States, and validated the prediction performance in two independent datasets from the United Kingdom. Our data show that SL-trained AI models outperform most locally trained models, and perform on par with models that are trained on the merged datasets. In addition, we show that SL-based AI models are data efficient. In the future, SL can be used to train distributed AI models for any histopathology image analysis task, eliminating the need for data transfer.},
added-at = {2024-09-10T11:56:37.000+0200},
author = {Saldanha, Oliver Lester and Quirke, Philip and West, Nicholas P and James, Jacqueline A and Loughrey, Maurice B and Grabsch, Heike I and Salto-Tellez, Manuel and Alwers, Elizabeth and Cifci, Didem and Ghaffari Laleh, Narmin and Seibel, Tobias and Gray, Richard and Hutchins, Gordon G A and Brenner, Hermann and van Treeck, Marko and Yuan, Tanwei and Brinker, Titus J and Chang-Claude, Jenny and Khader, Firas and Schuppert, Andreas and Luedde, Tom and Trautwein, Christian and Muti, Hannah Sophie and Foersch, Sebastian and Hoffmeister, Michael and Truhn, Daniel and Kather, Jakob Nikolas},
biburl = {https://puma.scadsai.uni-leipzig.de/bibtex/2fdad6e2dcabb6284fcbb97d14f3d14f9/scadsfct},
copyright = {https://creativecommons.org/licenses/by/4.0},
interhash = {317955d0cc273e1d2841bd762a01edce},
intrahash = {fdad6e2dcabb6284fcbb97d14f3d14f9},
journal = {Nat. Med.},
keywords = {},
language = {en},
month = jun,
number = 6,
pages = {1232--1239},
publisher = {Springer Science and Business Media LLC},
timestamp = {2024-09-10T15:15:57.000+0200},
title = {Swarm learning for decentralized artificial intelligence in cancer histopathology},
volume = 28,
year = 2022
}
