Implementing Medical AI in Resource-Limited Environments: Obstacles and Approaches
Overview
This scoping review synthesizes evidence from 44 studies on deploying medical AI in low-resource settings, highlighting key barriers such as unreliable infrastructure, poor data quality, limited local expertise, and governance gaps. Effective strategies emphasize resilient infrastructure, interoperable data standards, capacity building, ethical governance, and sustainable policy frameworks.
Background
Artificial intelligence has transformative potential in healthcare by improving diagnostics and clinical decision-making. However, its implementation in low- and middle-income countries (LMICs) faces challenges including weak digital infrastructure, fragmented health data, and limited local skills. Sustainable AI deployment requires addressing these systemic constraints through human-centered, context-aware approaches that integrate technology with local workflows and governance. This review explores barriers and enabling strategies across four domains: infrastructure, data quality and capacity, ethics and governance, and policy sustainability.
Data Highlights
A total of 44 peer-reviewed studies published between 2015 and 2026 were included. Common barriers identified include unreliable electricity and internet access, incomplete or messy data, limited AI familiarity among healthcare workers, and lack of clear regulatory frameworks. Enabling strategies reported involve investments in resilient digital infrastructure, adoption of interoperable data standards such as HL7/FHIR, continuous capacity-building programs, fairness and bias auditing mechanisms, and integration of AI governance within national digital health policies supported by sustainable financing.
Key Findings
Unstable electricity and intermittent internet connectivity significantly hinder AI system reliability in low-resource settings.
Data quality issues, including incomplete and fragmented health records, limit AI effectiveness and require robust electronic health record systems.
Limited local technical capacity and familiarity with AI among healthcare workers impede adoption and sustainable use.
Ethical concerns such as transparency, accountability, privacy, and equity must be embedded throughout the AI lifecycle to ensure trustworthiness.
Governance frameworks aligned with WHO and UNESCO AI ethics guidelines are essential for equitable and sustainable AI integration.
Successful deployment strategies focus on resilient infrastructure, interoperable data standards (e.g., HL7/FHIR), continuous training, and sustainable financing models.
Clinical Implications
Clinicians and health system leaders in LMICs should prioritize strengthening digital infrastructure and data systems to support AI tools. Capacity-building initiatives are critical to enhance local expertise and ensure AI solutions are contextually appropriate. Embedding ethical principles and governance mechanisms will foster trust and equitable use of AI, ultimately augmenting rather than replacing human clinical judgment.
Conclusion
Sustainable medical AI deployment in resource-limited environments depends on a human-centered approach that integrates resilient infrastructure, quality data, capacity building, and ethical governance. Aligning AI innovation with local contexts and values is key to realizing its potential to improve healthcare equity and outcomes in LMICs.
