Related Experiment Video
Updated: May 28, 2025
07:45
Assessing Binocular Central Visual Field and Binocular Eye Movements in a Dichoptic Viewing Condition
Published on: July 21, 2020
4.4K
Using ChatGPT-4 in visual field test assessment
Gulsah Gumus Akgun1, Cigdem Altan1, Ali Safa Balci1
1Beyoglu Eye Training and Research Hospital, University of Health Sciences, Istanbul, Turkey.
Clinical & Experimental Optometry
|February 12, 2025
Summary
ChatGPT-4 shows high accuracy in interpreting visual field test names and patterns, but struggles with complex map interpretation and diagnosis. This highlights potential for AI in ophthalmology, with room for improvement in detailed visual field analysis.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Medical Diagnostics
Background:
- Visual field testing is crucial for diagnosing and managing ophthalmic diseases like glaucoma.
- Integrating advanced AI, such as ChatGPT-4, offers potential to enhance clinical decision-making and accessibility in eye care.
Purpose of the Study:
- To evaluate the accuracy and capability of ChatGPT-4 in interpreting various components of visual field tests.
- To compare AI interpretation with expert ophthalmologists' assessments.
Main Methods:
- 30 patient visual field printouts were analyzed by ChatGPT-4.
- Performance was assessed on identifying test elements (name, pattern, reliability, maps) and diagnostic categorization.
- Comparison was made against interpretations from two glaucoma consultants.
Main Results:
- ChatGPT-4 achieved high accuracy for test names (100%), patterns (90%), and reliability indices (93.3%).
- Interpretation of deviation and greyscale maps yielded lower accuracy (66.7% and 30%, respectively).
- Accurate 'normal' classification or diagnosis suggestion was achieved in only 33.3% of cases.
Conclusions:
- Large language models like ChatGPT-4 demonstrate significant potential in analyzing visual field tests, particularly numerical data.
- Current limitations exist in interpreting complex visual field maps and providing accurate diagnostic suggestions.
- Further development is needed for AI to fully support clinical interpretation of visual field defects.
