Initial Insights Into an Institutional Secure Large Language Model for Magnetic Resonance Imaging Examination Requests: Retrospective Study

Key Points

• The study aims to evaluate the use of secure large language models (sLLMs) in enhancing the accuracy and context of MRI examination requests.
• Conducted a retrospective analysis across various MRI types—body, musculoskeletal, and neuroradiology.
• Implemented sLLMs to augment examination requests.
• Assessed the accuracy of sLLM-enhanced requests compared to general radiologists.
• sLLM integration improved clinical context and contrast selection in MRI requests.
• Demonstrated accuracy comparable to general radiologists regarding region or coverage.
• Potential reduction in manual workload for protocol selection was noted.

Abstract