Lessons learned from a multi-centre implementation of an artificial intelligence algorithm to detect vertebral fractures for radiology, information technology, information governance and clinical leads

Journal article : Review

Artificial intelligence (AI) algorithms have been developed to identify vertebral fractures through reanalysis of existing CT scans. This study describes a real-world case study of the deployment of an AI solution in the NHS, from information governance (IG) and technology (IT) perspectives, to inform best practice recommendations. Five NHS hospitals were selected to deploy the Nanox AI solution to identify vertebral fractures from existing CT images. The journey to IG and IT assurance was described and used to inform recommendations. The time from contract signing to IG assurance ranged between 5 and 13 months. The period from IG assurance to the analysis of the first patient scan ranged from 7 to 12 months, excluding 1 site withdrawing from the process. Each site required different IG documents: Data Protection Impact Assessment (5/5 sites), Data Protection Agreement (2/5 sites), Digital Technology Assessment Criteria (2/5 sites). IT implementation delays included third-party supplier coordination, NHS IT staff availability, and local capability. Based on the observed challenges, 6 best practice recommendations are proposed to address current challenges to AI adoption in radiology settings in the NHS to support IG and 8 to support IT implementation services. Significant challenges remain if AI is to be routinely used to identify vertebral fractures. The proposed recommendations provide a pathway to improve effective and efficient AI deployment. This study proposes recommendations from IG and IT perspectives to improve the local deployment of AI in the NHS.

Authors: Vijjhalwar, R; Song, K; Chappell, DDG; Turton, J; Boylan, J

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Oxford University Press
• Journal: BJR Artificial Intelligence
• Volume: 2
• Issue: 1
• Pages: ubaf017
• Article number: ubaf017
• Publication date: 2025-10-24
• Acceptance date: 2025-10-13
• DOI: 10.1093/bjrai/ubaf017
• EISSN: 2976-8705
• ISSN: 2976-8705
• Language: English
• Keywords: spinal fractures; information governance; fracture liaison service; osteoporosis; information technology; artificial intelligence
• Subtype: Review