Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Review
. 2025 May 19:12:1544501.
doi: 10.3389/fmed.2025.1544501. eCollection 2025.

Ethical framework for responsible foundational models in medical imaging

Debesh Jha  1 , Gorkem Durak  1 , Abhijit Das  1 , Jasmer Sanjotra  1 , Onkar Susladkar  1 , Suramyaa Sarkar  1 , Ashish Rauniyar  2 , Nikhil Kumar Tomar  1 , Linkai Peng  1 , Sirui Li  1 , Koushik Biswas  1 , Ertugrul Aktas  1 , Elif Keles  1 , Matthew Antalek  1 , Zheyuan Zhang  1 , Bin Wang  1 , Xin Zhu  1   3 , Hongyi Pan  1 , Deniz Seyithanoglu  1 , Alpay Medetalibeyoglu  1 , Vanshali Sharma  1 , Vedat Cicek  1 , Amir A Rahsepar  1 , Rutger Hendrix  1   4 , A Enis Cetin  3 , Bulent Aydogan  5 , Mohamed Abazeed  6 , Frank H Miller  1 , Rajesh N Keswani  1   7 , Hatice Savas  1 , Sachin Jambawalikar  8 , Daniela P Ladner  9 , Amir A Borhani  1 , Concetto Spampinato  1   4 , Michael B Wallace  1   10 , Ulas Bagci  1
Affiliations
Review

Ethical framework for responsible foundational models in medical imaging

Debesh Jha et al. Front Med (Lausanne). .

Abstract

The emergence of foundational models represents a paradigm shift in medical imaging, offering extraordinary capabilities in disease detection, diagnosis, and treatment planning. These large-scale artificial intelligence systems, trained on extensive multimodal and multi-center datasets, demonstrate remarkable versatility across diverse medical applications. However, their integration into clinical practice presents complex ethical challenges that extend beyond technical performance metrics. This study examines the critical ethical considerations at the intersection of healthcare and artificial intelligence. Patient data privacy remains a fundamental concern, particularly given these models' requirement for extensive training data and their potential to inadvertently memorize sensitive information. Algorithmic bias poses a significant challenge in healthcare, as historical disparities in medical data collection may perpetuate or exacerbate existing healthcare inequities across demographic groups. The complexity of foundational models presents significant challenges regarding transparency and explainability in medical decision-making. We propose a comprehensive ethical framework that addresses these challenges while promoting responsible innovation. This framework emphasizes robust privacy safeguards, systematic bias detection and mitigation strategies, and mechanisms for maintaining meaningful human oversight. By establishing clear guidelines for development and deployment, we aim to harness the transformative potential of foundational models while preserving the fundamental principles of medical ethics and patient-centered care.

Keywords: ethical AI; fairness; foundational models; medical imaging; responsible AI.

PubMed Disclaimer

Conflict of interest statement

UB acknowledges the following COI: Ther-AI LLC. MW acknowledges the following COIs: Boston Scientific, ClearNote Health, Cosmo Pharmaceuticals, Endostart, Endiatix, Fujifilm, Medtronic, Surgical Automations, Ohelio Ltd, Venn Bioscience, Virgo Inc., Surgical Automation, and Microtek. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The author(s) declared that they were an editorial board member of Frontiers, at the time of submission. This had no impact on the peer review process and the final decision.

Figures

Figure 1
Figure 1
Comprehensive workflow of a foundational model in medical imaging: from multi-modal data acquisition and curation to deployment and downstream tasks for higher order skills and vision tasks.
Figure 2
Figure 2
Illustration of the federated learning paradigm, in which multiple institutions collaboratively train deep learning models in a decentralized framework without exchanging raw patient data. Each institution independently updates its local model using private datasets and transmits only model parameters to a central aggregation server. The server securely integrates these updates to refine a global model, which is then redistributed to participating sites. This iterative process preserves data privacy, maintains data locality, and mitigates risks related to bias and fairness while ensuring robust model generalization across diverse clinical settings.
Figure 3
Figure 3
Framework of ethical AI governance components.
See this image and copyright information in PMC

References

    1. Medetalibeyoglu A, Velichko YS, Hart EM, Bagci U. Foundational artificial intelligence models and modern medical practice. BJR| Artif Intellig. (2025) 2:ubae018. 10.1093/bjrai/ubae018 - DOI - PMC - PubMed
    1. Brown T, Mann B, Ryder N, Subbiah M, Kaplan JD, Dhariwal P, et al. Language models are few-shot learners. In: Advances in Neural Information Processing Systems. (2020). p. 33.
    1. STAT News. Epic Overhauls Sepsis Algorithm After Finding Biases in its Training Data. (2022).
    1. Rasmy L, Xiang Y, Xie Z, Tao C, Zhi D. Med-BERT: pretrained contextualized embeddings on large-scale structured electronic health records for disease prediction. NPJ Digit Med. (2021) 4:455. 10.1038/s41746-021-00455-y - DOI - PMC - PubMed
    1. Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez I. Attention is all you need. In: Advances in Neural Information Processing Systems. Long Beach, CA. (2017). p. 30.

LinkOut - more resources