CoChat

Swarm learning for decentralized artificial intelligence in cancer histopathology

Open access

Published version of a preprint

The DOI matches its Crossref record; 58 of the 58 references it cites were found in a registry or an index; 13 of the 25 sources its claims cite were checked for support.

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.

Overview

CoChat Summary

TL;DR

This retrospective multicentric study shows that swarm learning (SL) can train AI models on decentralized gigapixel histopathology images to predict BRAF mutational status and microsatellite instability in colorectal cancer without transferring patient data. SL models generally outperformed local models and performed on par with centrally merged models, while being data efficient.

Key findings

  • For BRAF prediction in QUASAR, local models achieved AUROCs of 0.7358 ± 0.0162 (Epi700), 0.7339 ± 0.0107 (DACHS) and 0.7071 ± 0.0243 (TCGA); the merged model reached 0.7567 ± 0.0139. SL models b-chkpt1, b-chkpt2 and weighted w-chkpt achieved 0.7634 ± 0.0047, 0.7621 ± 0.0045 and 0.7736 ± 0.0057, respectively, with w-chkpt significantly better than the merged model (P = 0.0374).
  • For MSI prediction in QUASAR, b-chkpt1 and b-chkpt2 achieved AUROCs of 0.8001 ± 0.0073 and 0.8151 ± 0.0071, outperforming local Epi700 (0.7884 ± 0.0043) and TCGA (0.7639 ± 0.0162); w-chkpt was on par with the merged model (0.8326 ± 0.0089 vs 0.8308 ± 0.0190, P = 0.8650). In YCR BCIP, merged and w-chkpt were also on par (0.8943 ± 0.0161 vs 0.8882 ± 0.0084, P = 0.4647).
  • SL was data efficient: for BRAF, w-chkpt maintained AUROCs of 0.7000 ± 0.0260 at 100 patients, 0.7139 ± 0.0149 at 200 and 0.7438 ± 0.0093 at 300 patients per cohort, not significantly different from the merged model; for MSI, w-chkpt was comparable to merged down to 300 patients per cohort, but merged outperformed swarm models at ≤200 patients.
  • In a blinded reader study, for BRAF models trained on 300 patients per cohort, TILs (P = 0.019), poor differentiation (P = 0.017) and signet ring cells (P = 0.019) were significantly more frequent in tiles selected by swarm-trained models than by locally trained models.

Methods

  • Retrospective multicentric study using five colorectal cancer cohorts: training cohorts Epi700 (n = 661, Northern Ireland), DACHS (n = 2,448, Germany) and TCGA (n = 632, largely United States); test cohorts QUASAR (clinical trial cohort) and YCR BCIP (n = 889, United Kingdom). Ground-truth BRAF and MSI/dMMR status was determined by cohort-specific methods including PCR, immunohistochemistry and sequencing.
  • Whole-slide images were tessellated into 512 × 512 × 3 patches, color-normalized, and filtered for blur/no tissue; ResNet-18 extracted a 512 × 1 feature vector from 150 randomly selected patches per patient, followed by a fully connected classifier. SL used the HPE implementation on three physically separate servers with blockchain-based coordination; basic and weighted SL checkpoints were compared. AUROC was the primary metric, experiments were repeated five times, and two-sided unpaired t-tests with P ≤ 0.05 were used.

Limitations

  • Stated: differential privacy was not explicitly investigated; model performance needs further improvement before clinical implementation; previous work suggests approximately 10,000 patients may be needed for further gains; embedding SL servers in existing healthcare infrastructure across countries would require substantial practical efforts; large-scale international validation is needed.
  • Evident: all statistical comparisons used two-sided t-tests without correction for multiple testing; only one model and one hyperparameter set were evaluated; synchronization interval optimization was tested only on the QUASAR MSI/dMMR task; BRAF status was not assessed in YCR BCIP because BRAF testing there was performed only in MSI/dMMR cases.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41591-022-01768-5

Cites

58 of 58 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Figures

Fig. 1 Schematic of the deep learning and SL workflows. a, Histology image analysis workflow for training. b, Histology image analysis workflow for model deployment (inference). c, SL workflow and training cohorts included in this study. On three physically separate bare-metal servers (dashed line), three different sets of clinical data reside. Each server runs an AI process (a program that trains a model on the data) and a network process (a program that handles communication with peers via blockchain). d, Test cohorts included in this study. e, Schematic of the basic SL experiment. For basic SL, the number of epochs is equal for all cohorts, and weights are equal for all cohorts. f, Schematic of the weighted SL experiment. For weighted SL, the number of epochs is larger for small cohorts, and weights are smaller for small cohorts (wE = weight for the Epi700 cohort, wD = weight for the DACHS cohort, wT = weight for the TCGA cohort). Icon credits: a, OpenMoji (CC BY-SA 4.0); c,d, Twitter Twemoji (CC-BY 4.0).
Fig. 1 Schematic of the deep learning and SL workflows. a, Histology image analysis workflow for training. b, Histology image analysis workflow for model deployment (inference). c, SL workflow and training cohorts included in this study. On three physically separate bare-metal servers (dashed line), three different sets of clinical data reside. Each server runs an AI process (a program that trains a model on the data) and a network process (a program that handles communication with peers via blockchain). d, Test cohorts included in this study. e, Schematic of the basic SL experiment. For basic SL, the number of epochs is equal for all cohorts, and weights are equal for all cohorts. f, Schematic of the weighted SL experiment. For weighted SL, the number of epochs is larger for small cohorts, and weights are smaller for small cohorts (wE = weight for the Epi700 cohort, wD = weight for the DACHS cohort, wT = weight for the TCGA cohort). Icon credits: a, OpenMoji (CC BY-SA 4.0); c,d, Twitter Twemoji (CC-BY 4.0).CC BY · PubMed Central
Fig. 2 AI-based prediction of molecular alterations by local, merged and swarm models. a, Classification performance (AUROC) for prediction of BRAF mutational status at the patient level in the QUASAR dataset. Total cohort sizes (number of patients, for BRAF mutational status) in the training set are 642 for Epi700, 2,075 for DACHS and 500 for TCGA. Total cohort size (number of patients, for BRAF mutational status) in the test set is 1,477 for QUASAR. b, AUROC for prediction of MSI status in QUASAR. Total cohort sizes (number of patients, for MSI/dMMR) in the training sets are 594 for Epi700, 2,039 for DACHS and 426 for TCGA. Total cohort size (number of patients, for MSI/dMMR status) in the test set is 1,774 for QUASAR. c, AUROC for prediction of MSI status in the YCR BCIP dataset. Total cohort sizes (number of patients, for MSI/dMMR status) in the training sets are identical to those in b. Total cohort size (number of patients, for MSI/dMMR status) in the test set is 805 for YCR BCIP. In a–c, the boxes show the median values and quartiles, the whiskers show the rest of the distribution (except for points identified as outliers), and all original data points are shown. d, Model examination through slide heatmaps of tile-level predictions for representative cases in the QUASAR cohort. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant (P > 0.05). Exact P values are available in Supplementary Table 1 (for a), Supplementary Table 2 (for b) and Supplementary Table 3 (for c). All statistical comparisons were made using two-sided t-tests without correction for multiple testing.
Fig. 2 AI-based prediction of molecular alterations by local, merged and swarm models. a, Classification performance (AUROC) for prediction of BRAF mutational status at the patient level in the QUASAR dataset. Total cohort sizes (number of patients, for BRAF mutational status) in the training set are 642 for Epi700, 2,075 for DACHS and 500 for TCGA. Total cohort size (number of patients, for BRAF mutational status) in the test set is 1,477 for QUASAR. b, AUROC for prediction of MSI status in QUASAR. Total cohort sizes (number of patients, for MSI/dMMR) in the training sets are 594 for Epi700, 2,039 for DACHS and 426 for TCGA. Total cohort size (number of patients, for MSI/dMMR status) in the test set is 1,774 for QUASAR. c, AUROC for prediction of MSI status in the YCR BCIP dataset. Total cohort sizes (number of patients, for MSI/dMMR status) in the training sets are identical to those in b. Total cohort size (number of patients, for MSI/dMMR status) in the test set is 805 for YCR BCIP. In a–c, the boxes show the median values and quartiles, the whiskers show the rest of the distribution (except for points identified as outliers), and all original data points are shown. d, Model examination through slide heatmaps of tile-level predictions for representative cases in the QUASAR cohort. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant (P > 0.05). Exact P values are available in Supplementary Table 1 (for a), Supplementary Table 2 (for b) and Supplementary Table 3 (for c). All statistical comparisons were made using two-sided t-tests without correction for multiple testing.CC BY · PubMed Central
Fig. 3 SL models are data efficient. a, Classification performance (AUROC) for prediction of BRAF mutational status at the patient level in the QUASAR cohort. Total cohort sizes (number of patients, for BRAF mutational status) in the training sets are 642 for Epi700, 2,075 for DACHS and 500 for TCGA. Total cohort size (number of patients, for BRAF mutational status) in the test set is 1,477 for QUASAR. b, Classification performance (AUROC) for prediction of MSI/dMMR status at the patient level in the QUASAR cohort. Total cohort sizes (number of patients, for MSI/dMMR) in the training sets are 594 for Epi700, 2,039 for DACHS and 426 for TCGA. Total cohort size (number of patients, for MSI/dMMR status) in the test set is 1,774 for QUASAR. c, Classification performance (AUROC) for prediction of MSI/dMMR status at the patient level in the YCR BCIP cohort. Total cohort sizes (number of patients, for MSI/dMMR status) in the training sets are identical to those in b. Total cohort size (number of patients, for MSI/dMMR status) in the test set is 805 for YCR BCIP. In a–c, the boxes show the median values and quartiles, the whiskers show the rest of the distribution (except for points identified as outliers), and all original data points are shown.
Fig. 3 SL models are data efficient. a, Classification performance (AUROC) for prediction of BRAF mutational status at the patient level in the QUASAR cohort. Total cohort sizes (number of patients, for BRAF mutational status) in the training sets are 642 for Epi700, 2,075 for DACHS and 500 for TCGA. Total cohort size (number of patients, for BRAF mutational status) in the test set is 1,477 for QUASAR. b, Classification performance (AUROC) for prediction of MSI/dMMR status at the patient level in the QUASAR cohort. Total cohort sizes (number of patients, for MSI/dMMR) in the training sets are 594 for Epi700, 2,039 for DACHS and 426 for TCGA. Total cohort size (number of patients, for MSI/dMMR status) in the test set is 1,774 for QUASAR. c, Classification performance (AUROC) for prediction of MSI/dMMR status at the patient level in the YCR BCIP cohort. Total cohort sizes (number of patients, for MSI/dMMR status) in the training sets are identical to those in b. Total cohort size (number of patients, for MSI/dMMR status) in the test set is 805 for YCR BCIP. In a–c, the boxes show the median values and quartiles, the whiskers show the rest of the distribution (except for points identified as outliers), and all original data points are shown.CC BY · PubMed Central
Fig. 4 Highly predictive image patches for BRAF prediction. All patches are from the QUASAR test set and were obtained using the median model (out of five repetitions) trained on 300 randomly selected patients per training cohort. a–f, Model trained on Epi700 (a), model trained on DACHS (b), model trained on TCGA (c), model trained on all three datasets (d), swarm chkpt1 (e), swarm chkpt2 (f). Tiles with red borders contain artifacts or more than 50% nontumor tissue.
Fig. 4 Highly predictive image patches for BRAF prediction. All patches are from the QUASAR test set and were obtained using the median model (out of five repetitions) trained on 300 randomly selected patients per training cohort. a–f, Model trained on Epi700 (a), model trained on DACHS (b), model trained on TCGA (c), model trained on all three datasets (d), swarm chkpt1 (e), swarm chkpt2 (f). Tiles with red borders contain artifacts or more than 50% nontumor tissue.CC BY · PubMed Central
Extended Data Fig. 1 Workflow details and effect of synchronization interval. (a) Schematic of the structure of the Swarm Learning network in HPE Swarm Learning which was used in this study. (b) Schematic of the training procedure in Swarm Learning. (c) Evaluation of synchronization (sync) interval on the model performance. (d) Pairwise (two-sided) t-tests yielded non-significant (p > 0.05) p-values for all pairwise comparisons of the AUROCs obtained with 1, 4, 8, 16, 32 and 64 iterations between sync events. (e) Time for training with different sync internal. Abbreviations: WSI = whole slide images, MSI = microsatellite instability, SL = swarm learning, SN = swarm network, SPIRE = SPIFFE Runtime Environment. All statistical comparisons were made with two-sided t-tests without correction for multiple testing.
Extended Data Fig. 1 Workflow details and effect of synchronization interval. (a) Schematic of the structure of the Swarm Learning network in HPE Swarm Learning which was used in this study. (b) Schematic of the training procedure in Swarm Learning. (c) Evaluation of synchronization (sync) interval on the model performance. (d) Pairwise (two-sided) t-tests yielded non-significant (p > 0.05) p-values for all pairwise comparisons of the AUROCs obtained with 1, 4, 8, 16, 32 and 64 iterations between sync events. (e) Time for training with different sync internal. Abbreviations: WSI = whole slide images, MSI = microsatellite instability, SL = swarm learning, SN = swarm network, SPIRE = SPIFFE Runtime Environment. All statistical comparisons were made with two-sided t-tests without correction for multiple testing.CC BY · PubMed Central
Extended Data Fig. 2 CONSORT chart for Epi700. Initial patient number in this dataset, exclusions and missing values, and final patient number.
Extended Data Fig. 2 CONSORT chart for Epi700. Initial patient number in this dataset, exclusions and missing values, and final patient number.CC BY · PubMed Central
Extended Data Fig. 3 CONSORT chart for DACHS. Initial patient number in this dataset, exclusions and missing values, and final patient number.
Extended Data Fig. 3 CONSORT chart for DACHS. Initial patient number in this dataset, exclusions and missing values, and final patient number.CC BY · PubMed Central
Extended Data Fig. 4 CONSORT chart for TCGA. Initial patient number in this dataset, exclusions and missing values, and final patient number.
Extended Data Fig. 4 CONSORT chart for TCGA. Initial patient number in this dataset, exclusions and missing values, and final patient number.CC BY · PubMed Central
Extended Data Fig. 5 CONSORT chart for QUASAR. Initial patient number in this dataset, exclusions and missing values, and final patient number.
Extended Data Fig. 5 CONSORT chart for QUASAR. Initial patient number in this dataset, exclusions and missing values, and final patient number.CC BY · PubMed Central
Extended Data Fig. 6 CONSORT chart for YCR-BCIP. Initial patient number in this dataset, exclusions and missing values, and final patient number.
Extended Data Fig. 6 CONSORT chart for YCR-BCIP. Initial patient number in this dataset, exclusions and missing values, and final patient number.CC BY · PubMed Central
Extended Data Fig. 7 Results of the blinded reader study. For each model (seven model types, BRAF and MSI/dMMR prediction tasks, positive and negative class), a blinded observer scored the presence of five relevant histopathological patterns or structures in the highly scoring image tiles. (a) Presence of relevant patterns or structures in highly scoring tiles in the BRAF mutated class for BRAF prediction models trained on 300 patients per cohort, as scored by the blinded observer. P-values indicate a two-sided comparison between the three local models and the three Swarm-trained models for each feature. (b) Presence of relevant patterns or structures in highly scoring tiles in the MSI/dMMR for MSI status prediction models trained on 300 patients per cohort, as scored by the blinded observer. (c) Same experiment as panel (A), but for the models which were trained on all patients in all cohorts. (d) Same experiment as panel (B), but for the models which were trained on all patients in all cohorts. Abbreviations: MSI = mismatch repair deficiency, B-Chkpt = basic Swarm Learning experiment checkpoint, W-Chkpt = weighted Swarm Learning experiment checkpoint, TILs = tumor-infiltrating lymphocytes, Poor Diff. = poor differentiation, Crohn’s like = Crohn’s like lymphoid reaction, N/A = not applicable (division by zero). All statistical comparisons were made with two-sided t-tests without correction for multiple testing.
Extended Data Fig. 7 Results of the blinded reader study. For each model (seven model types, BRAF and MSI/dMMR prediction tasks, positive and negative class), a blinded observer scored the presence of five relevant histopathological patterns or structures in the highly scoring image tiles. (a) Presence of relevant patterns or structures in highly scoring tiles in the BRAF mutated class for BRAF prediction models trained on 300 patients per cohort, as scored by the blinded observer. P-values indicate a two-sided comparison between the three local models and the three Swarm-trained models for each feature. (b) Presence of relevant patterns or structures in highly scoring tiles in the MSI/dMMR for MSI status prediction models trained on 300 patients per cohort, as scored by the blinded observer. (c) Same experiment as panel (A), but for the models which were trained on all patients in all cohorts. (d) Same experiment as panel (B), but for the models which were trained on all patients in all cohorts. Abbreviations: MSI = mismatch repair deficiency, B-Chkpt = basic Swarm Learning experiment checkpoint, W-Chkpt = weighted Swarm Learning experiment checkpoint, TILs = tumor-infiltrating lymphocytes, Poor Diff. = poor differentiation, Crohn’s like = Crohn’s like lymphoid reaction, N/A = not applicable (division by zero). All statistical comparisons were made with two-sided t-tests without correction for multiple testing.CC BY · PubMed Central

Verification

Claims the paper makes

7 Valid · 1 Not valid · 8 Not checked · 56 Own results and interpretations

27 more claims

Create a free account to see all 72 claims and to verify the 12 references still open.

See all claims — free account
  • Probably not a substantial claim. 21
  • It may state two separate claims as one. 23
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

3 of 3

Citations

Cites

Show all 58 references

Research works printed but not linked: 1

  1. McMahan, B., Moore, E., Ramage, D., Hampson, S. & Arcas, B. A. Y. Communication-efficient learning of deep networks from decentralized data. In Proc. 20th Int. Conf. Artif. Intell. Stat. Vol. 54 (Eds. Singh, A. & Zhu, J.) 1273–1282 (PMLR, 2017).

Cited by

See all 256

Metadata

Authors
Saldanha, O. L., Quirke, P., West, N. P., James, J. A., Loughrey, M. B., Grabsch, H. I., Salto-Tellez, M., Alwers, E., Cifci, D., Ghaffari Laleh, N., Seibel, T., Gray, R., Hutchins, G. G. A., Brenner, H., van Treeck, M., Yuan, T., Brinker, T. J., Chang-Claude, J., Khader, F., Schuppert, A., Luedde, T., Trautwein, C., Muti, H. S., Foersch, S., Hoffmeister, M., Truhn, D., & Kather, J. N.
Affiliations
  • Universitätsklinikum Aachen (DE)
  • University of Leeds (GB)
  • Queen's University Belfast (GB)
  • German Cancer Research Center (DE)
  • University of Oxford (GB)
  • Universität Hamburg (DE)
  • Düsseldorf University Hospital (DE)
  • University Medical Center of the Johannes Gutenberg University Mainz (DE)
Corresponding author
Kather, Jakob Nikolas
Venue
Nature Medicine
Published
2022-04-25
Volume
28
Issue
6
Pages
1232-1239
ISSN
  • 1078-8956 (ISSN-L)
  • 1546-170X
OpenAlex
W4225746601
Cited by
227
References
54
Citations by year
2026: 41 · 2025: 52 · 2024: 58 · 2023: 56 · 2022: 20
FWCI
15.31
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Medicine (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Research Publications (Maastricht University) (publishedVersion, cc-by)
  • White Rose Research Online (University of Leeds, The University of Sheffield, University of York) (acceptedVersion)
  • RWTH Publications (RWTH Aachen) (submittedVersion, other-oa)
  • Research Portal (Queen's University Belfast) (submittedVersion, cc-by)
  • Research Publications (Maastricht University) (submittedVersion, other-oa)
  • RWTH Publications (RWTH Aachen)
Funders
  • Deutsches Krebsforschungszentrum
  • Friends of the Cancer Centre
  • Cancer Research UK (A15333, C37703/A15333, C50104/A17592)
  • National Institute for Health and Care Research (NIHR200162)
  • Public Health Agency (SPI/5151/15, EAT/4905/13)
  • Bundesministerium für Bildung und Forschung (BR 1704/6-3, 01KH0404, 01ER0814, 01ER0815, 01ER1505A and 01ER1505B, BR 1704/6-4, BR 1704/6-1, 01ER1505B, 01ER1505A, 01ER0814, ZMVI1-2520DAT111, 01ER0815, CH 117/1-1, 01KH0404)
  • Bundesministerium für Gesundheit (ZMVI1-2520DAT111, 2520DAT111, DEEP LIVER ZMVI1-2520DAT111)
  • Deutsche Krebshilfe (#70113864, 2520DAT111)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › AI in cancer detection
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
Keywords
swarm learning, decentralized artificial intelligence, tumor histology, BRAF mutation status, microsatellite instability, colorectal cancer, H&E-stained slides, data privacy, data governance, data efficiency
MeSH terms
Artificial Intelligence, United Kingdom, Humans, Image Processing, Computer-Assisted, Neoplasms, Staining and Labeling

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Swarm learning for decentralized artificial intelligence in cancer histopathology. https://app.cochat.ai/papers/10.1038/s41591-022-01768-5
The paper
Saldanha, O. L., Quirke, P., West, N. P., James, J. A., Loughrey, M. B., Grabsch, H. I., Salto-Tellez, M., Alwers, E., Cifci, D., Ghaffari Laleh, N., Seibel, T., Gray, R., Hutchins, G. G. A., Brenner, H., van Treeck, M., Yuan, T., Brinker, T. J., Chang-Claude, J., Khader, F., . . . Kather, J. N. (2022). Swarm learning for decentralized artificial intelligence in cancer histopathology. Nature Medicine. https://doi.org/10.1038/s41591-022-01768-5

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.