From Policy to Practice: A Governance Framework for Trustworthy Artificial Intelligence in Medical Imaging in Low and Middle-Income Countries: Ghana as an Anchoring Case

Author (s)

Selorm Kweku Dzokoto, Saio Alusine Marrah, Sofo Tanko Rashid Computer, Karl Mark Arhin, Frank Korbla Dzokoto, Shahid Nawaz, & Emefa Arisiya

Abstract

Artificial intelligence (AI) is transforming medical imaging by improving diagnostic accuracy, streamlining workflows, and expanding access to specialist interpretation. These advances hold particular promise for low- and middle-income countries (LMICs) such as Ghana, where severe shortages of radiologists constrain timely diagnosis. Yet the introduction of AI also introduces substantial risks related to algorithmic bias, weak data governance, cybersecurity vulnerabilities, unclear clinical accountability, and dependence on external vendors. International guidance on trustworthy AI, while valuable, frequently presupposes institutional capabilities that many LMICs do not yet possess, creating a persistent policy-to-practice gap. This conceptual framework paper addresses that gap by proposing a seven-pillar governance framework for trustworthy AI in medical imaging, anchored in the Ghanaian context and designed for transferability across LMICs. The pillars comprise: (1) Governance and Leadership, (2) Ethical and Legal Compliance, (3) Data Governance and Security, (4) Clinical Validation and Safety, (5) Human Capacity Development, (6) Infrastructure and Digital Readiness, and (7) Monitoring, Auditing, and Continuous Improvement. The framework is operationalized through a five-level governance maturity model, measurable indicators, and a phased implementation roadmap. Drawing on international instruments (WHO, IMDRF, OECD, FUTURE-AI), comparative experiences from Rwanda, Kenya, Nigeria, South Africa, India, and other LMICs, and insights from institutional and technology governance theory, the paper translates high-level principles into practical institutional machinery. The contribution is both theoretical linking health policy, institutional, and technology governance perspectives and practical, offering decision-makers a capacity-proportionate pathway from aspirational policy to accountable practice.

Key words: Artificial intelligence, Medical imaging governance, Trustworthy AI, Low- and middle-income countries, Ghana, Health policy, Conceptual framework.

Download PDF

Cite This Article:

Dzokoto, S. K., Marrah, S. A., Rashid Computer, S. T., Arhin, K. M., Dzokoto, F. K., Nawaz, S., & Arisiya, E. (2026). From Policy to Practice: A Governance Framework for Trustworthy Artificial Intelligence in Medical Imaging in Low and Middle-Income Countries: Ghana as an Anchoring Case, Journal of Scientific Reports, 15(1), 36-51.  DOI: https://doi.org/10.58970/JSR.1238

About Author (s)

Selorm Kweku Dzokoto (Corresponding Author), PhD Researcher (School of Information and Software Engineering), University of Electronic Science and Technology of China, Sichuan P.R., 611731 Chengdu, China. ORCID: https://orcid.org/0009-0003-1230-0414

Saio Alusine Marrah, PhD Researcher (School of Information and Software Engineering), University of Electronic Science and Technology of China, Sichuan P.R., 611731 Chengdu, China. ORCID: https://orcid.org/0009-0008-4727-2480

Sofo Tanko Rashid Computer, PhD Researcher, Department of Humanities, Accra Metropolitan University, Accra, Ghana.

Karl Mark Arhin, Doctor of Philosophy (Political Science Department), Kaaf University, Kasoa, Ghana. ORCID: https://orcid.org/0009-0007-4362-5800

Frank Korbla Dzokoto, Associate Dean (Pearson Construction Management Faculty), Global Banking School, U.K. ORCID: https://orcid.org/0009-0007-8033-2468

Shahid Nawaz, Masters’ Student (School of Information and Software Engineering), University of Electronic Science and Technology of China, Sichuan P.R., 611731 Chengdu, China. ORCID: https://orcid.org/0009-0009-6004-9897

Emefa Arisiya, Health Professional, University of Wisconsin, United States of America.

Download PDF

DOI: https://doi.org/10.58970/JSR.1238

Summarize this article with: