Related Experiment Video
Updated: Jun 7, 2025
03:14
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
504
MRI spine request form enhancement and auto protocoling using a secure institutional large language model.
James Thomas Patrick Decourcy Hallinan1, Naomi Wenxin Leow2, Wilson Ong3
1Department of Diagnostic Imaging, National University Hospital, 5 Lower Kent Ridge Rd, Singapore 119074, Singapore; Department of Diagnostic Radiology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore.
The Spine Journal : Official Journal of the North American Spine Society
|November 13, 2024
Summary
A secure large language model (LLM) significantly improved clinical information on MRI spine request forms. The LLM also accurately suggested correct MRI protocols in 78.4% of cases, enhancing radiologist workflow.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Healthcare
- Clinical Workflow Optimization
Background:
- Secure institutional large language models (LLMs) offer potential to alleviate noninterpretative tasks for radiologists.
- Streamlining diagnostic processes is crucial for improving healthcare efficiency.
Purpose of the Study:
- To evaluate the effectiveness of a secure institutional LLM in enhancing MRI spine request forms.
- To assess the LLM's capability in auto-protocoling MRI spine examinations.
Main Methods:
- A retrospective analysis of 250 spine MRI request forms from 218 patients was conducted.
- A secure institutional LLM (Claude 2.0) was employed to augment clinical information and suggest MRI protocols.
- The adequacy of information and accuracy of protocols were rated by experienced musculoskeletal radiologists and compared to a consensus standard.
Main Results:
- LLM-augmented forms showed significantly higher adequacy of clinical information (93.6-96.0%) compared to original forms (46.8-58.8%).
- The LLM suggested the correct MRI protocol in 78.4% of cases, demonstrating substantial accuracy.
- The LLM excelled at identifying spinal instrumentation (95.1%), outperforming junior radiologists.
Conclusions:
- Utilizing a secure institutional LLM demonstrably improves the quality of clinical information on MRI spine request forms.
- The LLM's ability to accurately suggest MRI protocols holds promise for optimizing the MRI workflow and reducing radiologist workload.
