Federated Computing in Orthopaedic Surgery: A Paradigm Shift in Collaborative Multicenter Research.

Compared with other medical fields, orthopaedics lags in multicenter research. A recent study found that only 9.9% of publications in leading orthopaedic journals were based on multicenter data, compared with 64.4% in general medicine and 26.9% in other surgical subspecialties 1 . The issue is well recognized: 78% of members attending the 125th American Orthopaedic Association (AOA) annual meeting in 2012 agreed that multicenter studies were a critical issue 2 . Ironically, despite the early-twenty-first-century optimism about the potential of big data in medicine, orthopaedic research has only seen modest gains in cross-institutional collaboration, rising from 2.7% in 2009 3 to 9.9% in 2021 1 . Most progress to date has come from centralized registries and databases (e.g., centralized computing, which refers to any system in which data from multiple sites are pooled into a single location [physical or virtual] for storage, curation, and analysis), but these systems bring their own limitations and logistical burdens 4 . Over the past decade, a new approach has emerged that promises to lower several barriers to multicenter collaboration: federated computing (FC). However, FC remains absent from the orthopaedic landscape. In this review, we outline persistent challenges to multicenter studies, introduce FC, explain how it addresses key barriers, and highlight landmark applications in other medical fields. Finally, we discuss how federated approaches could be leveraged in orthopaedic research—and remaining obstacles.

The Problem

The Need for Multicenter Research

Multicenter studies offer clear advantages across medicine (Table 1 ). Single-center studies are more likely to yield findings later contradicted in multicenter settings 5 and often overestimate treatment effects 6 , 7 . In orthopaedics, multicenter collaboration is particularly important given the complexity of measuring surgical outcomes. Because outcome results (e.g., pain) are influenced by biological, psychological, and social factors 8 , they are inherently noisy, requiring larger and more demographically diverse populations to detect meaningful effects. Single-center designs often lack sufficient variation to assess surgeon-specific or hospital-specific effects 9 - 11 , key sources of outcome variability, particularly in trauma research. Research on complex or heterogeneous diseases is also limited by the constraints of single-center studies. Although rare-disease research faces obvious sample size limitations, common conditions such as osteoarthritis present a different challenge. Driven by both genetic predisposition and mechanically induced tissue damage 12 , osteoarthritis likely encompasses multiple undefined subtypes 13 . Without standardized phenotypes 14 , detecting treatment effects or patient subgroups requires large, diverse data sets, making multicenter collaboration essential for uncovering latent patterns and building clinically meaningful disease models.

Data Stewardship and the Multicenter Research Deadlock

FC offers a way to overcome a persistent barrier in orthopaedic research: the difficulty of conducting multicenter studies due to data-sharing challenges. Many hospitals regard their clinical data as “digital gold.” 15 , 16 Fearing a loss of competitive advantage or of future opportunities to monetize intellectual property, even simple collaborations can be mired in lengthy legal negotiations. However, this analogy is flawed. Gold is a fungible commodity with intrinsic value. Unlike gold, the value of clinical data depends on context-specific use and the tools to analyze it, resources typically developed by industry.

Under the U.S. Health Insurance Portability and Accountability Act (HIPAA), hospitals are stewards of patient data, with ethical and legal duties to protect privacy. These obligations understandably lead to a cautious stance toward data-sharing. However, stewardship does not need to mean exclusivity; patients benefit when their information is used safely and anonymously to improve care.

Meanwhile, specialized research laboratories and industry have the expertise to turn real-world data into actionable insights, but progress is stalled by limited access. The result is a data-sharing deadlock: hospitals retain data that are often underutilized relative to their potential, whereas development lacks the access required to innovate. FC offers a promising path forward: preserving institutional control over data, aligning with ethical stewardship, and enabling collaborative research without compromising patient privacy.

More Barriers

Prior to the 2018 Common Rule 17 , one of the most substantial barriers to multicenter research was the cumbersome and inconsistent process of obtaining approval across all local institutional review boards 18 - 20 . Ambiguities in federal regulations 21 led to wide variation in how individual institutional review boards interpreted and implemented these rules 19 , 20 , creating a Kafkaesque approval process.

Although recent reforms now allow a single institutional review board to oversee multisite studies, ambiguity in regulatory interpretation persists 21 . As a result, delays remain common, particularly when institutional review boards engage in “mission creep” 21 - 23 : extending their oversight into low-risk studies involving limited or de-identified data 21 , 22 , 24 .

In parallel, data-use agreement negotiations present another major hurdle. Typically handled by institutional attorneys or contract officers, these reviews aim to manage legal, regulatory, and reputational risks, but often introduce considerable delays 24 . A 2020 study of 59 leading U.S. research universities identified common points of contention, including indemnification language and definitions of confidential information 24 .

Together, these administrative and legal barriers continue to slow and, in some cases, prevent multicenter scientific collaboration, despite modest improvement in regulatory frameworks.

A Solution: FC

Origins

The origins of FC begin with federated learning. First described by Google in 2015, federated learning enables algorithm development using personal data without directly accessing it 25 - 27 . Local computers analyze private data and then send only results, not raw data, to the federated server, where they are aggregated to reflect the input from all participating users (see the Appendix for an example calculation). As a founder said, this “brings the code to the data, instead of the data to the code and addresses the fundamental problems of privacy, ownership, and locality of data.” 27 Although initially developed for deep learning, the broader concept, FC, applies to any analyses that can be partitioned into local computations and combined into a global analysis (see Table 2 for some recently adapted traditional analytic methods). In this article, we use the term “FC” to emphasize this broader applicability.

The key advantage to FC is sharing insights without sharing data. Hospitals retain sole custody of patient records, ensuring privacy, security, and regulatory compliance. In a typical multicenter federation, each site maintains its own standardized patient data and a federated server aggregates results for global analysis (Fig. 1 ). Since its introduction, multiple variations of FC and federal learning have emerged to tackle increasingly complex research problems 28 , 29 .

The Benefits of FC

The ability to collaborate without sharing patient data reduces many barriers to multicenter studies, particularly in observational and retrospective research. FC simplifies regulatory compliance by ensuring that patient-level data never leave their originating institution 30 . This streamlined approach can greatly reduce the complexity and duration of multisite negotiations. Several FC studies have received expedited, or have even been deemed exempt from, review from the institutional review board 31 , 32 , and, in many institutions, data-use agreements are unnecessary when no patient information is shared 33 . For retrospective studies, using an honest broker often reduces approval requirements to the confirmation of proper protected health information (PHI) cleaning and handling. Simplifying collaborative bureaucracy can shorten the timeline from data availability to actionable insights, sometimes from months to weeks. Computational costs are distributed across existing institutional computing resources, minimizing additional infrastructure needs 34 . Lastly, cybersecurity risk is also reduced, as only encrypted, aggregated outputs reach the federated server, not raw patient data.

In modern research, many projects rely on advanced, cross-disciplinary skills, such as expertise in machine learning and software engineering, that are often scarce in clinical settings. In single-center studies, participation in these projects is limited to institutions with access to these high-demand skill sets. FC changes this dynamic. Within a federation, only 1 site needs advanced software development expertise; participating sites simply need the information technology (IT) capacity to connect and manage their local data. This multiplies the value of the technical experts within the network and allows researchers at other sites to contribute to projects that they otherwise might not join.

In the place of true multicenter studies, orthopaedics research often consists of several single-center investigations that might later be combined in a meta-analysis. For meta-analyses to yield valid conclusions, the underlying studies must be methodologically comparable, a condition rarely met when independent researchers design and conduct their own studies. Nonetheless, meta-analyses are often performed out of necessity, as they are one of the few ways to synthesize available evidence 35 . Researchers can attempt to account for inter-study methodological differences, but some differences are subtle, poorly documented, or unknowable, introducing bias into the results. By contrast, FC facilitates true multicenter studies with shared analytic code, ensuring that outcomes are measured “with the same ruler” across sites. This standardization produces higher-quality, less-biased evidence than a patchwork of independent studies.

FC Is Not a Registry

To be clear, FC is not a medical data registry. Registry data reflect the careful selection of data that are consistent across medical centers and are easily collected, to maximize institutional participation. Although invaluable, these national and international databases are not enough. By design, registries often lack the data needed for emerging questions. Furthermore, participation requires long-term institutional commitment. Once data are shared, they cannot be rescinded, raising the stakes for privacy and security concerns. Establishing such databases requires sustained funding, hosting, staffing, and data collection infrastructure.

In contrast, FC enables study-specific, time-limited collaborations in which researchers define a focused research question, specify study variables, secure institutional review board approval or exemption, and run analyses locally using federated software. Participating sites still require personnel capable of collecting and extracting study variables and ensuring local data quality. Once data are prepared locally, the execution of federated analyses is typically automated and requires only standard IT support at most sites. When analyses are complete, the federation can be dissolved without any ongoing transfer or custody of patient-level data.

Landmark Studies

Informing Orthopaedics Through Other Fields

Since 2015, more than 220 health-care studies have applied FC 29 , demonstrating its value across scales, data types, and clinical domains. Table 3 contains a representative subset. Consortia range from 2 hospitals 36 to >70 sites across 6 continents 37 , with cohorts ranging from hundreds of thousands of patients 36 to rare-disease data sets expanded from <100 cases per site to >6,000 cases through a federation 37 . The latter would be an otherwise unattainable scale for the very rare condition being studied. This is a key FC strength: whether connecting a few hospitals or dozens across the globe, it generates data sets far larger and more diverse than any single center could achieve alone 36 , 37 - 39 .

Clinical use cases include disease and mortality prediction 40 - 44 , image-based diagnosis 32 , 37 , 45 - 49 , postoperative complication prediction 36 , wearable sensor data analysis 50 , histopathological diagnosis 51 , genetic association studies 52 , and quantitative structure-activity relationship (QSAR) analysis used in drug discovery 38 , 39 . Study data span electronic health record (EHR) data 36 , 53 , genetics data 52 , medical imaging (including radiography 32 , 49 , magnetic resonance imaging [MRI] 36 , 45 , 46 , and computed tomography [CT] 47 , 54 , 55 ), and even output from surgical robots 56 . Although deep learning on images is most common, FC also supports logistic and linear regression 57 , 58 , generalized linear mixed modeling 52 , and survival analysis 40 - 44 , 59 , 60 , among many other techniques (Table 2 ).

Early work testing FC methods found them robust to challenges such as skewed data distributions, high dimensionality, multiclass problems, and complex models 40 . Several prior health-care studies with public data sets have been replicated with data split across FC sites, yielding results consistent with the original analysis 61 . This finding has been consistently confirmed in numerous studies 31 , 36 , 43 , 48 , 49 , 54 . In some cases, federated models even outperformed centrally trained models (i.e., models developed using multicenter data that have all been collected on a single computer), particularly in generalizing across diverse patient populations and institutional settings 61 . In EHR-based predictive modeling, federated models have matched the diagnostic performance of centralized approaches. For example, Vaid et al. reported area under the operating characteristic curve (AUC) scores within 0.01 point of centralized models across multiple hospital systems 53 , a finding that has been consistently replicated across other clinical domains 37 , 54 . Similar parity extends to medical imaging, with methods yielding diagnostic accuracy comparable with centralized models in chest radiograph classification and generalizing to external hospital data sets 49 . Comparable results have been reported in numerous single-modality imaging studies 54 , 62 - 64 . In complex tasks such as multimodal MRI tumor segmentation, FC Dice scores (0.858) matched those of centralized models (0.863) 45 . FC has proven robust in realistic video settings with noisy or sparse data, even outperforming centralized approaches in surgical video segmentation 56 . Comparable performance has also been reported in applications involving digital pathology and multimodal clinical data sets 37 , 65 .

Across studies, FC substantially outperformed single-site models, with AUC gains of +0.04 to +0.15 in EHR and radiographic tasks 49 , 53 , Dice similarity improvements of +0.12 in imaging segmentation 45 , and intersection over union gains of approximately 15 points in surgical video analysis 56 . These outcomes reinforce a key strength of FC: its ability to maintain stable, high-quality performance across institutions, improving generalizability over single-site models 31 , 32 , 36 , 37 , 39 , 45 .

Federations can also exist as sustained academic networks such as the German Cancer Consortium’s Joint Imaging Platform 66 , an ongoing consortium across 11 Picture Archiving and Communication Systems (PACSs). Other federations have direct clinical impact, including the 30-site Federated Tumour Segmentation (FeTS) initiative 37 , 67 , using a graphical user interface to improve tumor boundary detection of brain glioma, breast tumors, liver tumors, and bone lesions from patients with multiple myeloma. The MELLODDY project is a consortium of 10 competing pharmaceutical companies 39 . Together, they train a common predictive protein-binding model, accelerating drug discovery without sharing proprietary data.

Ideas for Orthopaedic-Specific Applications

FC potentially has broad, clinically meaningful applications in orthopaedic surgery, as illustrated in Table 4 . Across subspecialties, key diagnostic and prognostic questions, ranging from radiographic interpretation to surgical decision-making and long-term outcome prediction, require large, diverse data sets that no single institution can provide. Many relevant data sources, including serial imaging, complex operative variables beyond simple timestamps (e.g., alignment targets, implant positioning, navigation or robotic metrics, soft-tissue balancing data), wearable sensor outputs, and patient-reported outcomes, are too sensitive or heterogeneous for centralized pooling. FC overcomes these barriers by enabling each site to run standardized analyses locally for enhanced security, and by data control sharing only aggregated, de-identified results, but delivering broad cross-center measurements.

FC does not change how clinical data are generated or collected. Participating institutions still perform standard study procedures, including prospective data collection, retrospective electronic medical records (EMR) abstraction, registry extraction, or device data export, under local protocol and institutional review board oversight. FC changes only how analyses are performed across sites, automating the execution of standardized analytic code locally and sharing only intermediate statistical outputs rather than patient-level data.

This approach preserves patient privacy and institutional autonomy while unlocking collaboration on problems characterized by low event rates, rare subgroups, and variation in practice patterns. By integrating multimodal data without raw data transfer, FC supports the development of robust, generalizable models that reflect real-world diversity. The result is a scalable, privacy-preserving research framework capable of informing personalized treatment selection, optimizing surgical techniques, and improving long-term outcomes across orthopaedic care.

Limitations

Like any methodology, FC has limitations 68 . Privacy protections are not automatic; analyses must be designed to avoid inadvertent disclosure (e.g., computations returning a local maximum value are actually returning patient data, not an intermediate value). Numerous studies have explored what can and cannot be determined about an institution’s patient data from shared intermediate values 69 , 70 . These studies assumed a “rogue” participant within the federation. Although this is a risk when companies use FC across customer devices, it is clearly not a concerning risk when the consortium is a group of trusted collaborators 28 . Further techniques such as differential privacy can mitigate these risks but add complexity 40 , 71 . The encryption of all communications effectively blocks outsiders from ever seeing intermediate results, making security risks low 72 . Readers seeking deeper descriptions of software architectures 28 , 73 , algorithmic advances 29 , 34 , and software 74 solutions to these issues should consult algorithmically focused journals, as these evolve rapidly and exceed this review’s scope.

FC also requires technical expertise. At least 1 site must have collaborators capable of implementing computations and handling issues such as correct mathematical handling of non-independent and non-identically distributed data 71 . Beneficially, this expertise can be shared across the federation. Software setup involves choosing between open-source frameworks 74 , which require local IT capacity, and commercial turnkey platforms (e.g., Apheris or Rhino Federated Computing). Computations may run on on-premise or cloud hospital resources. Sites using on-premise systems must work closely with hospital IT and security. The sensitivity of the data will dictate how much additional scrutiny each center will need to give. For consortia studying problems reducible to non-human subject research through sufficient data cleaning (and using honest brokers), some sites may receive an institutional review board determination of exemption or non-human subject status and may not require data use agreements, depending on local policy and governance requirements.

When multicenter research was discussed at the 125th AOA annual meeting, another perceived barrier was limited National Institutes of Health (NIH) funding for multicenter studies 2 . Although FC does not change this, it can reduce participation costs and simplify the overhead associated with more traditional means of sharing data.

A common concern is that journals requiring public data access appear at odds with the privacy-preserving nature of FC. In practice, publishers allow exceptions when data cannot be shared directly, as seen in studies leveraging FC 32 , 37 , 55 , 65 .

Finally, although FC simplifies data governance and protects privacy, it does not make the science itself any simpler. Sites must still ensure that data are collected in comparable ways and stored in compatible formats and that coherent inclusion criteria are applied. Analysts and technical specialists remain essential to ensure valid, interpretable analyses across sites. FC does not eliminate the need for coordination; it simply enables it without sharing raw data.

Summary

FC represents a practical, timely solution to long-standing barriers in orthopaedic multicenter research. FC uniquely harnesses the statistical power, generalizability, and clinical insights possible from extensive multi-institutional data sets while reducing regulatory and contractual hurdles without compromising patient privacy. By exchanging only encrypted intermediate analysis values rather than raw patient information, FC dramatically simplifies collaborative analytics. Its success in other medical domains, across diverse data types, analytic methods, and clinical applications, demonstrates its technical feasibility and potential to generate high-quality, generalizable evidence.

For orthopaedics, FC offers a path to move beyond siloed, single-center studies and toward standardized, multi-institution networks tackling complex diagnostic, prognostic, and therapeutic questions. These networks can leverage scarce technical expertise across sites, support rapid project initiation, and improve external validity by incorporating diverse patient populations and practice environments. Although challenges remain, including the need for careful analytic design, technical infrastructure, and sustained funding, the evidence indicates that FC can achieve performance equivalent to, and sometimes exceeding, traditional centralized approaches, while substantially outperforming single-site models.

As orthopaedic surgery embraces data-informed decision-making, FC provides a scalable, cost-effective framework to accelerate research and improve patient outcomes while maintaining the privacy and autonomy that institutions and patients rightfully expect.

Appendix

Supporting material provided by the authors is posted with the online version of this article as a data supplement at jbjs.org ( http://links.lww.com/JBJS/J362 ).

Acknowledgments

N ote : The authors thank Drs. James Ames, John-Erik Bell, Frances Faro, Eric Henderson, William McKinnon, Adam Pearson, Vincent Pellegrini, and Philip Wolinsky (all from Dartmouth-Hitchcock Medical Center) for their contributions to defining questions suited to multicenter collaboration using federated learning for their respective orthopaedic subspecialties.

Footnotes

Investigation performed at Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire

Disclosure: The publishing and open access fees for this article were paid for by Johnson and Johnson MedTech Investigator Initiated Study DPS-JOINTS-2025-012. The Disclosure of Potential Conflicts of Interest forms are provided with the online version of the article ( http://links.lww.com/JBJS/J361 ).

Contributor Information

Brandon G. Hill, Email: wandering.hill@gmail.com.

Thomas E. Nyul, Email: tomnyul@gmail.com.

Wayne E. Moschetti, Email: Wayne.E.Moschetti@hitchcock.org.

References

1. Kuhn AW, Rund JM, Wolf BR, Brophy RH. Multi-investigator collaboration in orthopedic surgery research compared to other medical fields: update comparing 2021-2009. J Orthop Res. 2024. Apr;42(4):873-7.

2. Hilibrand AS, Spindler K, O’Keefe RJ. Demonstrating the value of orthopaedic surgery through multicenter trials: AOA critical issues. J Bone Joint Surg Am. 2015. Apr 1;97(7):e35.

3. Brophy RH, Smith MV, Latterman C, Jones MH, Reinke EK, Flanigan DC, Wright RW, Wolf BR. Multi-investigator collaboration in orthopaedic surgery research compared to other medical fields. J Orthop Res. 2012. Oct;30(10):1523-8.

4. Bohl DD, Singh K, Grauer JN. Nationwide databases in orthopaedic surgery research. J Am Acad Orthop Surg. 2016. Oct;24(10):673-82.

5. Bellomo R, Warrillow SJ, Reade MC. Why we should be wary of single-center trials. Crit Care Med. 2009. Dec;37(12):3114-9.

6. Dechartres A, Boutron I, Trinquart L, Charles P, Ravaud P. Single-center trials show larger treatment effects than multicenter trials: evidence from a meta-epidemiologic study. Ann Intern Med. 2011. Jul 5;155(1):39-51.

7. Unverzagt S, Prondzinsky R, Peinemann F. Single-center trials tend to provide larger treatment effects than multicenter trials: a systematic review. J Clin Epidemiol. 2013. Nov;66(11):1271-80.

8. Ayers DC, Franklin PD, Ring DC. The role of emotional health in functional outcomes after orthopaedic surgery: extending the biopsychosocial model to orthopaedics: AOA critical issues. J Bone Joint Surg Am. 2013. Nov 6;95(21):e165.

9. Shervin N, Rubash HE, Katz JN. Orthopaedic procedure volume and patient outcomes: a systematic literature review. Clin Orthop Relat Res. 2007. Apr;457(457):35-41.

10. Saulle R, Vecchi S, Cruciani F, Mitrova Z, Amato L, Davoli M. The combined effect of surgeon and hospital volume on health outcomes: a systematic review. Clin Ter. 2019. Mar-Apr;170(2):e148-61.

11. Harikumar V, Anvery N, Haq M, Christensen RE, Ahmed A, Koza E, Ma M, Shi V, Dirr MA, Worley B, Brieva JC, Yoo SS, Alam M. Risk factors for hospital-based surgical site infections. Arch Dermatol Res. 2023. Dec 13;316(1):43.

12. Berenbaum F. Osteoarthritis as an inflammatory disease (osteoarthritis is not osteoarthrosis!). Osteoarthritis Cartilage. 2013. Jan;21(1):16-21.

13. Mobasheri A, Saarakkala S, Finnilä M, Karsdal MA, Bay-Jensen AC, van Spil WE. Recent advances in understanding the phenotypes of osteoarthritis. F1000Res. 2019. Dec 12;8(December):8.

14. Deveza LA, Melo L, Yamato TP, Mills K, Ravi V, Hunter DJ. Knee osteoarthritis phenotypes and their relevance for outcomes: a systematic review. Osteoarthritis Cartilage. 2017. Dec;25(12):1926-41.

15. Consulting L.E.K.. Healthcare Providers and the Data Gold Rush: How to Get Your Share. 2023. Sep 21. Accessed 2025 Aug 14. https://www.lek.com/insights/hea/us/ei/healthcare-providers-and-data-gold-rush-how-get-your-share

16. MarketsandMarkets. How Can Healthcare Data Monetization Redefine ROI and Innovation in the Next Five Years? Accessed 2025 Aug 14. https://www.marketsandmarkets.com/blog/HC/healthcare-data-monetization-strategic-imperatives

17. U.S. Department of Health and Human Services. Federal Policy for the Protection of Human Subjects. U.S. Government Publishing Office; 2018.

18. Burman WJ, Reves RR, Cohn DL, Schooley RT. Breaking the camel’s back: multicenter clinical trials and local institutional review boards. Ann Intern Med. 2001. Jan 16;134(2):152-7.

19. Greene SM, Geiger AM. A review finds that multicenter studies face substantial challenges but strategies exist to achieve institutional review board approval. J Clin Epidemiol. 2006. Aug;59(8):784-90.

20. Stair TO, Reed CR, Radeos MS, Koski G, Camargo CA; MARC Investigators. Multicenter Airway Research Collaboration. Variation in institutional review board responses to a standard protocol for a multicenter clinical trial. Acad Emerg Med. 2001. Jun;8(6):636-41.

21. White RF. Institutional review board mission creep: the common rule, social science, and the nanny state. Indep Rev. 2007;11(4):547-64.

22. Gunsalus CK, Bruner EM, Burbules NC, Dash L, Finkin M, Goldberg JP, Greenough WT, Miller GA, Pratt MG. Mission creep in the IRB world. Science. 2006. Jun 9;312(5779):1441.

23. Musoba G, Jacob S, Robinson L. The institutional review board (IRB) and faculty: does the IRB challenge faculty professionalism in the social sciences? The Qualitative Report. 2014;19(51):1-14.

24. Mello MM, Triantis G, Stanton R, Blumenkranz E, Studdert DM. Waiting for data: barriers to executing data use agreements. Science. 2020. Jan 10;367(6474):150-52.

25. Konečný J, McMahan B, Ramage D. Federated Optimization: Distributed Optimization Beyond the Datacenter. arXiv. Preprint posted online November 11, 2015.

26. Konečný J, McMahan HB, Yu FX, Richtárik P, Suresh AT, Bacon D. Federated Learning: Strategies for Improving Communication Efficiency. arXiv. Preprint posted online October 30, 2017.

27. McMahan B, Moore E, Ramage D, Hampson S, Aguera y, Arcas A. Communication-Efficient Learning of Deep Networks from Decentralized Data. In: Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS). PMLR; 2017:1273-82.

28. Rieke N, Hancox J, Li W, Milletarì F, Roth HR, Albarqouni S, Bakas S, Galtier MN, Landman BA, Maier-Hein K, Ourselin S, Sheller M, Summers RM, Trask A, Xu D, Baust M, Cardoso MJ. The future of digital health with federated learning. NPJ Digit Med. 2020. Sep 14;3:119.

29. Zhang F, Kreuter D, Chen Y, Dittmer S, Tull S, Shadbahr T, Preller J, Rudd JHF, Aston JAD, Schönlieb CB, Gleadall N, Roberts M, BloodCounts! consortium. Recent methodological advances in federated learning for healthcare. Patterns (N Y). 2024. Jun 14;5(6):101006.

30. Casaletto J, Bernier A, McDougall R, Cline MS. Federated analysis for privacy-preserving data sharing: a technical and legal primer. Annu Rev Genomics Hum Genet. 2023. Aug 25;24:347-68.

31. Lee EH, Han M, Wright J, Kuwabara M, Mevorach J, Fu G, Choudhury O, Ratan U, Zhang M, Wagner MW, Goetti R, Toescu S, Perreault S, Dogan H, Altinmakas E, Mohammadzadeh M, Szymanski KA, Campen CJ, Lai H, Eghbal A, Radmanesh A, Mankad K, Aquilina K, Said M, Vossough A, Oztekin O, Ertl-Wagner B, Poussaint T, Thompson EM, Ho CY, Jaju A, Curran J, Ramaswamy V, Cheshier SH, Grant GA, Wong SS, Moseley ME, Lober RM, Wilms M, Forkert ND, Vitanza NA, Miller JH, Prolo LM, Yeom KW. An international study presenting a federated learning AI platform for pediatric brain tumors. Nat Commun. 2024. Sep 2;15(1):7615.

32. Dayan I, Roth HR, Zhong A, Harouni A, Gentili A, Abidin AZ, Liu A, Costa AB, Wood BJ, Tsai CS, Wang CH, Hsu CN, Lee CK, Ruan P, Xu D, Wu D, Huang E, Kitamura FC, Lacey G, de Antônio Corradi GC, Nino G, Shin HH, Obinata H, Ren H, Crane JC, Tetreault J, Guan J, Garrett JW, Kaggie JD, Park JG, Dreyer K, Juluru K, Kersten K, Rockenbach MABC, Linguraru MG, Haider MA, AbdelMaseeh M, Rieke N, Damasceno PF, E Silva PMC, Wang P, Xu S, Kawano S, Sriswasdi S, Park SY, Grist TM, Buch V, Jantarabenjakul W, Wang W, Tak WY, Li X, Lin X, Kwon YJ, Quraini A, Feng A, Priest AN, Turkbey B, Glicksberg B, Bizzo B, Kim BS, Tor-Díez C, Lee CC, Hsu CJ, Lin C, Lai CL, Hess CP, Compas C, Bhatia D, Oermann EK, Leibovitz E, Sasaki H, Mori H, Yang I, Sohn JH, Murthy KNK, Fu LC, de Mendonça MRF, Fralick M, Kang MK, Adil M, Gangai N, Vateekul P, Elnajjar P, Hickman S, Majumdar S, McLeod SL, Reed S, Gräf S, Harmon S, Kodama T, Puthanakit T, Mazzulli T, de Lavor VL, Rakvongthai Y, Lee YR, Wen Y, Gilbert FJ, Flores MG, Li Q. Federated learning for predicting clinical outcomes in patients with COVID-19. Nat Med. 2021. Oct;27(10):1735-43.

33. Rønn Hansen C, Price G, Field M, Sarup N, Zukauskaite R, Johansen J, Eriksen JG, Aly F, McPartlin A, Holloway L, Thwaites D, Brink C. Larynx cancer survival model developed through open-source federated learning. Radiother Oncol. 2022. Nov;176:179-86.

34. Kairouz P, McMahan HB, Avent B, Bellet A, Bennis M, Nitin Bhagoji A, Bonawitz K, Charles Z, Cormode G, Cummings R, D’Oliveira RGL, Eichner H, El Rouayheb S, Evans D, Gardner J, Garrett Z, Gascón A, Ghazi B, Gibbons PB, Gruteser M, Harchaoui Z, He C, He L, Huo Z, Hutchinson B, Hsu J, Jaggi M, Javidi T, Joshi G, Khodak M, Konecný J, Korolova A, Koushanfar F, Koyejo S, Lepoint T, Liu Y, Mittal P, Mohri M, Nock R, Özgür A, Pagh R, Qi H, Ramage D, Raskar R, Raykova M, Song D, Song W, Stich SU, Sun Z, Suresh AT, Tramèr F, Vepakomma P, Wang J, Xiong L, Xu Z, Yang Q, Yu FX, Yu H, Zhao S. Advances and open problems in federated learning. Found Trends®. Mach Learn. 2021;14(1-2):1-210.

35. Bhandari M, Morrow F, Kulkarni AV, Tornetta P, 3rd. Meta-analyses in orthopaedic surgery. A systematic review of their methodologies. J Bone Joint Surg Am. 2001 Jan;83(1):15-24.

36. Ren Y, Park Y, Shickel B, Guan Z, Patel A, Ma Y, Hu Z, Balch JA, Loftus TJ, Rashidi P, Ozrazgat-Baslanti T, Bihorac A. Federated learning for predicting major postoperative complications. Ann Surg Open. 2025. May 2;6(2):e573.

37. Pati S, Baid U, Edwards B, Sheller M, Wang SH, Reina GA, Foley P, Gruzdev A, Karkada D, Davatzikos C, Sako C, Ghodasara S, Bilello M, Mohan S, Vollmuth P, Brugnara G, Preetha CJ, Sahm F, Maier-Hein K, Zenk M, Bendszus M, Wick W, Calabrese E, Rudie J, Villanueva-Meyer J, Cha S, Ingalhalikar M, Jadhav M, Pandey U, Saini J, Garrett J, Larson M, Jeraj R, Currie S, Frood R, Fatania K, Huang RY, Chang K, Balaña C, Capellades J, Puig J, Trenkler J, Pichler J, Necker G, Haunschmidt A, Meckel S, Shukla G, Liem S, Alexander GS, Lombardo J, Palmer JD, Flanders AE, Dicker AP, Sair HI, Jones CK, Venkataraman A, Jiang M, So TY, Chen C, Heng PA, Dou Q, Kozubek M, Lux F, Michálek J, Matula P, Keřkovský M, Kopřivová T, Dostál M, Vybíhal V, Vogelbaum MA, Mitchell JR, Farinhas J, Maldjian JA, Yogananda CGB, Pinho MC, Reddy D, Holcomb J, Wagner BC, Ellingson BM, Cloughesy TF, Raymond C, Oughourlian T, Hagiwara A, Wang C, To MS, Bhardwaj S, Chong C, Agzarian M, Falcão AX, Martins SB, Teixeira BCA, Sprenger F, Menotti D, Lucio DR, LaMontagne P, Marcus D, Wiestler B, Kofler F, Ezhov I, Metz M, Jain R, Lee M, Lui YW, McKinley R, Slotboom J, Radojewski P, Meier R, Wiest R, Murcia D, Fu E, Haas R, Thompson J, Ormond DR, Badve C, Sloan AE, Vadmal V, Waite K, Colen RR, Pei L, Ak M, Srinivasan A, Bapuraj JR, Rao A, Wang N, Yoshiaki O, Moritani T, Turk S, Lee J, Prabhudesai S, Morón F, Mandel J, Kamnitsas K, Glocker B, Dixon LVM, Williams M, Zampakis P, Panagiotopoulos V, Tsiganos P, Alexiou S, Haliassos I, Zacharaki EI, Moustakas K, Kalogeropoulou C, Kardamakis DM, Choi YS, Lee SK, Chang JH, Ahn SS, Luo B, Poisson L, Wen N, Tiwari P, Verma R, Bareja R, Yadav I, Chen J, Kumar N, Smits M, van der Voort SR, Alafandi A, Incekara F, Wijnenga MMJ, Kapsas G, Gahrmann R, Schouten JW, Dubbink HJ, Vincent AJPE, van den Bent MJ, French PJ, Klein S, Yuan Y, Sharma S, Tseng TC, Adabi S, Niclou SP, Keunen O, Hau AC, Vallières M, Fortin D, Lepage M, Landman B, Ramadass K, Xu K, Chotai S, Chambless LB, Mistry A, Thompson RC, Gusev Y, Bhuvaneshwar K, Sayah A, Bencheqroun C, Belouali A, Madhavan S, Booth TC, Chelliah A, Modat M, Shuaib H, Dragos C, Abayazeed A, Kolodziej K, Hill M, Abbassy A, Gamal S, Mekhaimar M, Qayati M, Reyes M, Park JE, Yun J, Kim HS, Mahajan A, Muzi M, Benson S, Beets-Tan RGH, Teuwen J, Herrera-Trujillo A, Trujillo M, Escobar W, Abello A, Bernal J, Gómez J, Choi J, Baek S, Kim Y, Ismael H, Allen B, Buatti JM, Kotrotsou A, Li H, Weiss T, Weller M, Bink A, Pouymayou B, Shaykh HF, Saltz J, Prasanna P, Shrestha S, Mani KM, Payne D, Kurc T, Pelaez E, Franco-Maldonado H, Loayza F, Quevedo S, Guevara P, Torche E, Mendoza C, Vera F, Ríos E, López E, Velastin SA, Ogbole G, Soneye M, Oyekunle D, Odafe-Oyibotha O, Osobu B, Shu’aibu M, Dorcas A, Dako F, Simpson AL, Hamghalam M, Peoples JJ, Hu R, Tran A, Cutler D, Moraes FY, Boss MA, Gimpel J, Veettil DK, Schmidt K, Bialecki B, Marella S, Price C, Cimino L, Apgar C, Shah P, Menze B, Barnholtz-Sloan JS, Martin J, Bakas S. Federated learning enables big data for rare cancer boundary detection. Nat Commun. 2022. Dec 5;13(1):7346.

38. Chen S, Xue D, Chuai G, Yang Q, Liu Q. FL-QSAR: a federated learning-based QSAR prototype for collaborative drug discovery. Bioinformatics. 2021. Apr 1;36(22-23):5492-8.

39. Heyndrickx W, Mervin L, Morawietz T, Sturm N, Friedrich L, Zalewski A, Pentina A, Humbeck L, Oldenhof M, Niwayama R, Schmidtke P, Fechner N, Simm J, Arany A, Drizard N, Jabal R, Afanasyeva A, Loeb R, Verma S, Harnqvist S, Holmes M, Pejo B, Telenczuk M, Holway N, Dieckmann A, Rieke N, Zumsande F, Clevert DA, Krug M, Luscombe C, Green D, Ertl P, Antal P, Marcus D, Do Huu N, Fuji H, Pickett S, Acs G, Boniface E, Beck B, Sun Y, Gohier A, Rippmann F, Engkvist O, Göller AH, Moreau Y, Galtier MN, Schuffenhauer A, Ceulemans H. MELLODDY: Cross-pharma federated learning at unprecedented scale unlocks benefits in QSAR without compromising proprietary information. J Chem Inf Model. 2024. Apr 8;64(7):2331-44.

40. Sadilek A, Liu L, Nguyen D, Kamruzzaman M, Serghiou S, Rader B, Ingerman A, Mellem S, Kairouz P, Nsoesie EO, MacFarlane J, Vullikanti A, Marathe M, Eastham P, Brownstein JS, Arcas BAY, Howell MD, Hernandez J. Privacy-first health research with federated learning. NPJ Digit Med. 2021. Sep 7;4(1):132.

41. Jochems A, Deist TM, El Naqa I, Kessler M, Mayo C, Reeves J, Jolly S, Matuszak M, Ten Haken R, van Soest J, Oberije C, Faivre-Finn C, Price G, de Ruysscher D, Lambin P, Dekker A. Developing and validating a survival prediction model for NSCLC patients through distributed learning across 3 countries. Int J Radiat Oncol Biol Phys. 2017. Oct 1;99(2):344-52.

42. Field M, Vinod S, Delaney GP, Aherne N, Bailey M, Carolan M, Dekker A, Greenham S, Hau E, Lehmann J, Ludbrook J, Miller A, Rezo A, Selvaraj J, Sykes J, Thwaites D, Holloway L. Federated learning survival model and potential radiotherapy decision support impact assessment for non-small cell lung cancer using real-world data. Clin Oncol (R Coll Radiol). 2024. Jul;36(7):e197-208.

43. Bogowicz M, Jochems A, Deist TM, Tanadini-Lang S, Huang SH, Chan B, Waldron JN, Bratman S, O’Sullivan B, Riesterer O, Studer G, Unkelbach J, Barakat S, Brakenhoff RH, Nauta I, Gazzani SE, Calareso G, Scheckenbach K, Hoebers F, Wesseling FWR, Keek S, Sanduleanu S, Leijenaar RTH, Vergeer MR, Leemans CR, Terhaard CHJ, van den Brekel MWM, Hamming-Vrieze O, van der Heijden MA, Elhalawani HM, Fuller CD, Guckenberger M, Lambin P. Privacy-preserving distributed learning of radiomics to predict overall survival and HPV status in head and neck cancer. Sci Rep. 2020. Mar 11;10(1):4542.

44. Deist TM, Dankers FJWM, Ojha P, Scott Marshall M, Janssen T, Faivre-Finn C, Masciocchi C, Valentini V, Wang J, Chen J, Zhang Z, Spezi E, Button M, Jan Nuyttens J, Vernhout R, van Soest J, Jochems A, Monshouwer R, Bussink J, Price G, Lambin P, Dekker A. Distributed learning on 20 000+ lung cancer patients - The Personal Health Train. Radiother Oncol. 2020. Mar;144:189-200.

45. Sheller MJ, Edwards B, Reina GA, Martin J, Pati S, Kotrotsou A, Milchenko M, Xu W, Marcus D, Colen RR, Bakas S. Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data. Sci Rep. 2020. Jul 28;10(1):12598.

46. Silva S, Gutman BA, Romero E, Thompson PM, Altmann A, Lorenzi M. Federated learning in distributed medical databases: meta-analysis of large-scale subcortical brain data. In: 2019 IEEE International Symposium on Biomedical Imaging (ISBI 2019). New York: IEEE Computer Society; 2019. p 270-74.

47. Remedios S, Roy S, Blaber J, Bermudez C, Nath V, Patel MB, Butman JA, Landman BA, Pham DL. Distributed deep learning for robust multi-site segmentation of CT imaging after traumatic brain injury. Proc SPIE Int Soc Opt Eng. 2019. Mar;10949:109490A.

48. Lee H, Chai YJ, Joo H, Lee K, Hwang JY, Kim SM, Kim K, Nam IC, Choi JY, Yu HW, Lee MC, Masuoka H, Miyauchi A, Lee KE, Kim S, Kong HJ. Federated learning for thyroid ultrasound image analysis to protect personal information: validation study in a real health care environment. JMIR Med Inform. 2021. May 18;9(5):e25869.

49. Peng L, Luo G, Walker A, Zaiman Z, Jones EK, Gupta H, Kersten K, Burns JL, Harle CA, Magoc T, Shickel B, Steenburg SD, Loftus T, Melton GB, Gichoya JW, Sun J, Tignanelli CJ. Evaluation of federated learning variations for COVID-19 diagnosis using chest radiographs from 42 US and European hospitals. J Am Med Inform Assoc. 2022. Dec 13;30(1):54-63.

50. Chen Y, Qin X, Wang J, Yu C, Gao W. FedHealth: a federated transfer learning framework for wearable healthcare. IEEE Intell Syst. 2020. Jul;35(4):83-93.

51. Zhang Y, Li Z, Han X, Ding S, Li J, Wang J, Ying S, Shi J. Pseudo-data based self-supervised federated learning for classification of histopathological images. IEEE Trans Med Imaging. 2024. Mar;43(3):902-15.

52. Li W, Chen H, Jiang X, Harmanci A. FedGMMAT: federated generalized linear mixed model association tests. PLoS Comput Biol. 2024. Jul 24;20(7):e1012142.

53. Vaid A, Jaladanki SK, Xu J, Teng S, Kumar A, Lee S, Somani S, Paranjpe I, De Freitas JK, Wanyan T, Johnson KW, Bicak M, Klang E, Kwon YJ, Costa A, Zhao S, Miotto R, Charney AW, Böttinger E, Fayad ZA, Nadkarni GN, Wang F, Glicksberg BS. Federated learning of electronic health records to improve mortality prediction in hospitalized patients with COVID-19: machine learning approach. JMIR Med Inform. 2021. Jan 27;9(1):e24207.

54. Shiri I, Salimi Y, Sirjani N, Razeghi B, Bagherieh S, Pakbin M, Mansouri Z, Hajianfar G, Avval AH, Askari D, Ghasemian M, Sandoughdaran S, Sohrabi A, Sadati E, Livani S, Iranpour P, Kolahi S, Khosravi B, Bijari S, Sayfollahi S, Atashzar MR, Hasanian M, Shahhamzeh A, Teimouri A, Goharpey N, Shirzad-Aski H, Karimi J, Radmard AR, Rezaei-Kalantari K, Oghli MG, Oveisi M, Vafaei Sadr A, Voloshynovskiy S, Zaidi H. Differential privacy preserved federated learning for prognostic modeling in COVID-19 patients using large multi-institutional chest CT dataset. Med Phys. 2024. Jul;51(7):4736-47.

55. Feng B, Shi J, Huang L, Yang Z, Feng ST, Li J, Chen Q, Xue H, Chen X, Wan C, Hu Q, Cui E, Chen Y, Long W. Robustly federated learning model for identifying high-risk patients with postoperative gastric cancer recurrence. Nat Commun. 2024. Jan 25;15(1):742.

56. Zheng Z, Hayashi Y, Oda M, Kitasaka T, Mori K. Revisiting instrument segmentation: learning from decentralized surgical sequences with various imperfect annotations. Healthc Technol Lett. 2024. Jan 8;11(2-3):146-56.

57. Chang C, Bu Z, Long Q. CEDAR: communication efficient distributed analysis for regressions. Biometrics. 2023. Sep;79(3):2357-69.

58. Kim J, Li W, Bath T, Jiang X, Ohno-Machado L. VERTIcal Grid lOgistic regression with Confidence Intervals (VERTIGO-CI). AMIA Jt Summits Transl Sci Proc. 2021. May 17;2021:355-64.

59. Wang X, Zhang HG, Xiong X, Hong C, Weber GM, Brat GA, Bonzel CL, Luo Y, Duan R, Palmer NP, Hutch MR, Gutiérrez-Sacristán A, Bellazzi R, Chiovato L, Cho K, Dagliati A, Estiri H, García-Barrio N, Griffier R, Hanauer DA, Ho YL, Holmes JH, Keller MS, Klann MEng JG, L’Yi S, Lozano-Zahonero S, Maidlow SE, Makoudjou A, Malovini A, Moal B, Moore JH, Morris M, Mowery DL, Murphy SN, Neuraz A, Yuan Ngiam K, Omenn GS, Patel LP, Pedrera-Jiménez M, Prunotto A, Jebathilagam Samayamuthu M, Sanz Vidorreta FJ, Schriver ER, Schubert P, Serrano-Balazote P, South AM, Tan ALM, Tan BWL, Tibollo V, Tippmann P, Visweswaran S, Xia Z, Yuan W, Zöller D, Kohane IS, Avillach P, Guo Z, Cai T; Consortium for Clinical Characterization of COVID-19 by EHR 4CE. SurvMaximin: robust federated approach to transporting survival risk prediction models. J Biomed Inform. 2022. Oct;134:104176.

60. Zhang DK, Toni F, Williams M. A Federated Cox Model with Non-proportional Hazards. In: Shaban-Nejad A Michalowski M Bianco S, editors. Multimodal AI in Healthcare: A Paradigm Shift in Health Intelligence. New York: Springer International; 2023. p 171-85.

61. Garst S, Dekker J, Reinders M. A comprehensive experimental comparison between federated and centralized learning. Database (Oxford). 2025. Mar 19;2025:baaf016.

62. Yan B, Wang J, Cheng J, Zhou Y, Zhang Y, Yang Y, Liu L, Zhao H, Wang C, Liu B. Experiments of Federated Learning for COVID-19 Chest X-ray Images. In: Sun X, Zhang X, Xia Z, Bertino E, editors. Advances in Artificial Intelligence and Security. New York: Springer International; 2021. p 41-53.

63. Kakkar B, Johri P, Kumar Y, Park H, Son Y, Shafi J. An IoMT-based federated and deep transfer learning approach to the detection of diverse chest diseases using chest X-rays. Hum-Centric Comput Inf Sci. 2022;12(24).

64. Khan SH, Alam MGR. A Federated Learning Approach to Pneumonia Detection. In: 2021 International Conference on Engineering and Emerging Technologies (ICEET). 2021:1-6.

65. Ogier du Terrail J, Leopold A, Joly C, Béguier C, Andreux M, Maussion C, Schmauch B, Tramel EW, Bendjebbar E, Zaslavskiy M, Wainrib G, Milder M, Gervasoni J, Guerin J, Durand T, Livartowski A, Moutet K, Gautier C, Djafar I, Moisson AL, Marini C, Galtier M, Balazard F, Dubois R, Moreira J, Simon A, Drubay D, Lacroix-Triki M, Franchet C, Bataillon G, Heudel PE. Federated learning for predicting histological response to neoadjuvant chemotherapy in triple-negative breast cancer. Nat Med. 2023. Jan;29(1):135-46.

66. German Cancer Consortium. Joint Imaging Platform. Accessed 2025 Aug 12. https://jip.dktk.dkfz.de/jiphomepage/

67. FeTS Platform by FeTS-AI. Accessed 2025 Aug 12. https://fets-ai.github.io/FL-PoST/

68. Bak M, Madai VI, Celi LA, Kaissis GA, Cornet R, Maris M, Rueckert D, Buyx A, McLennan S. Federated learning is not a cure-all for data ethics. Nat Mach Intell. 2024. Mar;6(4):370-2.

69. Geiping J, Bauermeister H, Dröge H, Moeller M. Inverting gradients - how easy is it to break privacy in federated learning? In: Advances in Neural Information Processing Systems.Vol Vol 33. Curran Associates; 2020:16937-47.

70. Vero M, Balunović M, Dimitrov DI, Vechev M. TabLeak: tabular data leakage in federated learning. In: Proceedings of the 40th International Conference on Machine Learning.Vol Vol 202. 2023:35051-83.

71. Prayitno, Shyu CR, Putra KT, Chen HC, Tsai YY, Hossain KSMT, Jiang W, Shae ZY. A systematic review of federated learning in the healthcare area: from the perspective of data properties and applications. Appl Sci (Basel). 2021. Jan;11(23):11191.

72. Mothukuri V, Parizi RM, Pouriyeh S, Huang Y, Dehghantanha A, Srivastava G. A survey on security and privacy of federated learning. Future Gener Comput Syst. 2021. Feb;115:619-40.

73. Joshi M, Pal A, Sankarasubbu M. Federated learning for healthcare domain - pipeline, applications and challenges. ACM Trans Comput Healthc. 2022. Nov 3;3(4):1-36.

74. Riedel P, Schick L, von Schwerin R, Reichert M, Schaudt D, Hafner A. Comparative analysis of open-source federated learning frameworks - a literature-based survey and review. Int J Mach Learn Cybern. 2024. Nov;15(11):5257-78.

75. Kim Y, Sun J, Yu H, Jiang X. Federated tensor factorization for computational phenotyping. KDD Proc Int Conf Knowl Discov Data Min. 2017. Aug;2017:887-95.

76. Liu D, Dligach D, Miller T. Two-stage federated phenotyping and patient representation learning. Proc Conf Assoc Comput Linguist Meet. 2019. Aug;2019:283-91.

77. Choudhury O, Park Y, Salonidis T, Gkoulalas-Divanis A, Sylla I, Das AK. Predicting adverse drug reactions on distributed health data using federated learning. AMIA Annu Symp Proc. 2020. Mar 4;2019:313-22.

78. Li J, Tian Y, Zhu Y, Zhou T, Li J, Ding K, Li J. A multicenter random forest model for effective prognosis prediction in collaborative clinical research network. Artif Intell Med. 2020. Mar;103:101814.

79. Xu J, Xu Z, Walker P, Wang F. Federated patient hashing. Proc AAAI Conf Artif Intell. 2020. Apr 3;34(04):6486-93.

80. Li X, Gu Y, Dvornek N, Staib LH, Ventola P, Duncan JS. Multi-site fMRI analysis using privacy-preserving federated learning and domain adaptation: ABIDE results. Med Image Anal. 2020. Oct;65:101765.