Page 45 - AIH-2-3

Artificial Intelligence in Health

PERSPECTIVE ARTICLE
The role of Omnichain in advancing federated

learning for artificial intelligence training in
healthcare

Dongfang Wu * , and Yichen Wang 1,2
1,2
1 Global Health Research Center, Duke Kunshan University, Kunshan, Jiangsu, China
2 Global Health Institute, Duke University, Durham, North Carolina, United States of America

Abstract

Health data serves as a crucial foundation for artificial intelligence (AI) training in
the healthcare sector. The pivotal procedure for acquiring numerous and effective
health data lies in incentivizing participants to contribute their health data while
adhering to privacy regulations like the General Data Protection Regulation.
Federated learning achieves privacy protection by transmitting only parameters
rather than data to the model. When integrated with blockchain smart contracts, this
approach facilitates the automation of incentives according to health data quality,
thereby mitigating human’s subjective intervention. Consequently, the synergy
of these two methodologies offers new promise for the training of AI models in
healthcare. However, this advantage encounters performance degradation due
*Corresponding author:
Dongfang Wu to the heterogeneity among diverse blockchains. This article posits the concept
(dongfang.wu@duke.edu) of Omnichain as a potential solution to this challenge by analyzing its operational
Citation: Wu D, Wang Y. The role of mechanisms and future developmental trajectories and providing potential
Omnichain in advancing federated perspectives for its combination with hybrid federal learning solutions such as
learning for artificial intelligence differential privacy and secure multi-party computation to promote its application in
training in healthcare. Artif Intell the sphere of AI in healthcare.
Health. 2025;2(3):39-43.
doi: 10.36922/aih.5753
Received: November 4, 2024 Keywords: Omnichain; Federated learning; Artificial intelligence training; Healthcare;
1st revised: December 23, 2024 Training performance
2nd revised: January 24, 2025
Accepted: February 25, 2025
Published online: March 7, 2025 1. Introduction
Copyright: © 2025 Author(s). Health data is a vital asset in healthcare and thoroughly exploiting it for artificial
This is an Open-Access article intelligence (AI) training yields substantial value in diagnosis, medication, and patient
distributed under the terms of the care. Nonetheless, two concurrent challenges deserve consideration: (i) how to ensure
Creative Commons Attribution
License, permitting distribution, adherence of health data collection to privacy regulatory frameworks such as General
and reproduction in any medium, Data Protection Regulation; and (ii) how to incentivize individuals or institutions to
1
provided the original work is
properly cited. share their health data, with the aim to ensure continuity and reliability of AI training
quality. 2
Publisher’s Note: AccScience
Publishing remains neutral with Federated learning, an emergent paradigm in machine learning for AI, allows
regard to jurisdictional claims in
published maps and institutional algorithms to be trained across multiple distributed devices or servers with local data
affiliations. samples. In contrast to traditional machine learning, participants in federated learning

Volume 2 Issue 3 (2025) 39 doi: 10.36922/aih.5753