Glaucoma Detection for Automated Referral System: Using OCT Data and Fine-tuning LLM Models.

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Glaucoma is a primary cause of irreversible blindness.
• Early detection and timely referrals are critical for effective glaucoma management.
• Current referral processes can be administratively burdensome.

Purpose of the Study:

• To develop and evaluate an AI-powered automated system for generating glaucoma referral letters.
• To utilize large language models (LLMs) like GPT-4o with OCT image data for clinical decision support.
• To streamline the referral pathway for glaucoma-suspicious cases.

Main Methods:

• An AI system was developed using GPT-4o and GPT-4o-mini.
• The system processed 220 Optical Coherence Tomography (OCT) images.
• It extracted clinical data, interpreted results, and generated structured referral letters.

Main Results:

• The AI system achieved 100% recall (sensitivity) in identifying glaucoma cases.
• GPT-4o demonstrated high performance with 91% accuracy, 88% precision, and 93% F1-score.
• GPT-4o outperformed GPT-4o-mini in the evaluated metrics.

Conclusions:

• The AI system effectively automates the generation of glaucoma referral letters, reducing administrative workload.
• This technology streamlines referrals from OCT devices, facilitating early intervention.
• Integrating AI enhances diagnostic accuracy and optimizes resource utilization in glaucoma care.