EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Operationalizing Federated Healthcare AI: Design Patterns, Benchmarks, and Policy

Sadhasivam Mohanadas

International Journal of Computing and Engineering · 23.06.2026 · 8 (3) · 1-19

DOI: 10.47941/ijce.3797 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Computing and Engineering, online erschienen am 23.06.2026. Verfasst von Sadhasivam Mohanadas. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Purpose

In healthcare, federated AI can be developed and implemented in a privacy-preserving, clinically relevant, and policy-compliant manner. In this paper, we explore the design patterns, infrastructure requirements, and performance benchmarks for shared model development for genomic analysis, medical imaging, natural language processing, and sepsis detection using data from connected health devices and critical care monitoring systems. Methodology: The method used in this paper is a narrative review and framework synthesis of studies of the implementation of Federated AI in healthcare. Several design patterns for shared model development in distributed clinical environments for Genomics, Imaging, Language Processing, Sepsis prediction, and other applications, as well as several connected health devices, were analyzed. In addition, the required infrastructure, edge-based inference, and current benchmarks for model performance, privacy, latency, fairness, auditability, and deployment readiness of several AI applications in healthcare were reviewed and discussed.

Findings

There are existing studies and designs that have applied the shared model development approach to learning in distributed clinical settings, such as genomics, imaging, and language processing, using data from connected health devices and systems. These studies have the potential to improve patient care while keeping patient data locally within their respective clinical institutions. The existing approaches have limitations, however. The major limitations include inconsistent data quality across institutions, inadequate infrastructure to support distributed learning, and insufficient explainability. Furthermore, there are uncertainties in AI governance in healthcare. There is a fundamental trust issue in healthcare institutions that are tasked with implementing learning systems. Unique Contribution to Theory, Practice, and Policy: The paper outlines a framework to assist in implementing distributed health care AI by combining shared model learning, edge-based inference, privacy-protected synthetic data generation, explainability, and health care governance. The paper changes the way health care AI is perceived, from a centralized learning method to a distributed intelligent system. Deployment of the distributed health care AI system, using appropriate benchmarks (accuracy, latency, fairness, auditability, site preparedness, etc.) and corresponding (informed) consent management, explainability by default, audit trails, cross-border health care governance, etc., also supports low-resource health care sites.

Zitieren

APA

Mohanadas, S. (2026). Operationalizing Federated Healthcare AI: Design Patterns, Benchmarks, and Policy. International Journal of Computing and Engineering, 8(3), 1-19. https://doi.org/10.47941/ijce.3797

Vancouver

Mohanadas S. Operationalizing Federated Healthcare AI: Design Patterns, Benchmarks, and Policy. International Journal of Computing and Engineering. 2026;8(3):1-19. doi:10.47941/ijce.3797

BibTeX
@article{Mohanadas2026,
  author = {Mohanadas, Sadhasivam},
  title = {Operationalizing Federated Healthcare AI: Design Patterns, Benchmarks, and Policy},
  journal = {International Journal of Computing and Engineering},
  year = {2026},
  volume = {8},
  number = {3},
  pages = {1--19},
  doi = {10.47941/ijce.3797}
}
Kostenlos starten →