[Paper Review] Current State of Community-Driven Radiological AI Deployment in Medical Imaging

Vikas Gupta, Barbaros S. Erdal+25 more|arXiv (Cornell University)|Dec 29, 2022

Radiomics and Machine Learning in Medical Imaging4 citations

TL;DR

This paper proposes the MONAI Consortium’s open-source framework to bridge the gap between radiological AI research and clinical deployment, enabling seamless integration of AI models into hospital workflows through standardized tools, APIs, and community-driven development. It presents a taxonomy of AI use cases, identifies key deployment barriers, and demonstrates real-world implementations that reduce radiologist workload and improve efficiency through interoperable, clinically viable AI systems.

ABSTRACT

Artificial Intelligence (AI) has become commonplace to solve routine everyday tasks. Because of the exponential growth in medical imaging data volume and complexity, the workload on radiologists is steadily increasing. We project that the gap between the number of imaging exams and the number of expert radiologist readers required to cover this increase will continue to expand, consequently introducing a demand for AI-based tools that improve the efficiency with which radiologists can comfortably interpret these exams. AI has been shown to improve efficiency in medical-image generation, processing, and interpretation, and a variety of such AI models have been developed across research labs worldwide. However, very few of these, if any, find their way into routine clinical use, a discrepancy that reflects the divide between AI research and successful AI translation. To address the barrier to clinical deployment, we have formed MONAI Consortium, an open-source community which is building standards for AI deployment in healthcare institutions, and developing tools and infrastructure to facilitate their implementation. This report represents several years of weekly discussions and hands-on problem solving experience by groups of industry experts and clinicians in the MONAI Consortium. We identify barriers between AI-model development in research labs and subsequent clinical deployment and propose solutions. Our report provides guidance on processes which take an imaging AI model from development to clinical implementation in a healthcare institution. We discuss various AI integration points in a clinical Radiology workflow. We also present a taxonomy of Radiology AI use-cases. Through this report, we intend to educate the stakeholders in healthcare and AI (AI researchers, radiologists, imaging informaticists, and regulators) about cross-disciplinary challenges and possible solutions.

Motivation & Objective

• To address the critical gap between AI model development in research labs and actual clinical deployment in radiology.
• To reduce technical and operational barriers to integrating AI into existing hospital imaging workflows and IT infrastructures.
• To establish open standards and community-driven tools that enable interoperable, scalable, and reproducible AI deployment across diverse healthcare institutions.
• To provide a practical, end-to-end framework for labeling, training, deploying, and measuring AI models in clinical radiology settings.
• To democratize access to AI in medical imaging by supporting smaller and underserved healthcare facilities through open-source, low-overhead solutions.

Proposed method

• The MONAI Consortium was formed as a nonprofit, community-driven initiative to standardize AI deployment in radiology using open-source tools and shared best practices.
• The framework supports end-to-end AI workflows, from data labeling and model training to deployment and monitoring, using interoperable standards like DICOM, IHE, and HL7.
• MONAI Deploy enables deployment of AI models in clinical environments via containerized, modular, and portable applications compatible with PACS, RIS, and EMR systems.
• The consortium developed a model zoo for sharing pre-trained models and a taxonomy of radiology AI use cases to guide implementation across clinical scenarios.
• Integration with IHE Standard Log of Events (SOLE) allows objective measurement of AI impact on radiologist workflow efficiency and decision support.
• Collaborations with vendors and participation in events like IHE Connectathon and RSNA have validated the framework’s real-world feasibility and interoperability.

Experimental results

Research questions

• RQ1 What are the primary technical and operational barriers to deploying radiological AI models in clinical settings?
• RQ2 How can open-source, community-driven frameworks improve the translation of AI research into clinical practice?
• RQ3 What standards and architectural patterns enable seamless integration of AI into existing radiology IT ecosystems?
• RQ4 How can AI deployment be measured for clinical impact, particularly in terms of radiologist workflow efficiency and decision support?
• RQ5 What role do shared tools, APIs, and model repositories play in accelerating adoption across diverse healthcare institutions?

Key findings

• The MONAI framework enables end-to-end AI deployment in radiology with minimal overhead, supporting labeling, training, and clinical integration through open-source, interoperable tools.
• Real-world deployments at multiple institutions demonstrate that AI models can be successfully integrated into clinical workflows with measurable impact on radiologist efficiency.
• The use of standardized logging via IHE SOLE enables objective measurement of AI performance and workflow impact, supporting evidence-based evaluation.
• The MONAI model zoo and community-driven development model have accelerated sharing and reuse of AI models across institutions, reducing duplication and increasing reproducibility.
• Collaborations with vendors and participation in IHE events have validated the framework’s interoperability and clinical feasibility in real-world settings.
• The consortium’s open, community-driven model enables wider adoption of AI in radiology, particularly in underserved and smaller healthcare institutions.