Lab-AI: Using Retrieval Augmentation to Enhance Language Models for Personalized Lab Test Interpretation in Clinical...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Pathology

Background:

• Patient portals often use universal normal ranges for lab results.
• This overlooks critical conditional factors such as age and gender.
• Accurate interpretation of lab results is vital in clinical medicine.

Purpose of the Study:

• To introduce Lab-AI, an interactive system for personalized lab result interpretation.
• To leverage retrieval-augmented generation (RAG) for accessing credible health information.
• To provide personalized normal ranges based on patient-specific data.

Main Methods:

• Developed Lab-AI with two modules: factor retrieval and normal range retrieval.
• Utilized GPT-4-turbo with RAG for system implementation.
• Evaluated the system on 122 lab tests, including 40 with conditional factors.

Main Results:

• GPT-4-turbo with RAG achieved a 0.948 F1 score for factor retrieval.
• Achieved 0.995 accuracy for normal range retrieval.
• Outperformed non-RAG systems significantly in both factor and normal range retrieval.

Conclusions:

• Lab-AI demonstrates significant potential for enhancing patient comprehension of lab results.
• Personalized normal ranges improve the accuracy of lab result interpretation.
• RAG integrated with AI offers a powerful approach for clinical decision support.