Artificial Intelligence in Optometry: Current and Future Perspectives.

Area of Science:

• Ophthalmology and Optometry
• Artificial Intelligence in Healthcare
• Medical Informatics

Background:

• Global eye care faces challenges due to a shortage of professionals and increasing vision impairment.
• Artificial intelligence (AI) has the potential to enhance efficiency and decision-making in optometric workflows.
• Innovative solutions are needed to bridge gaps in eye care services, particularly in underserved regions.

Purpose of the Study:

• To review the application of AI in optometry, focusing on its potential to optimize patient care.
• To analyze AI models used in optometry, considering model selection, data size, evaluation metrics, and explainability.
• To identify the strengths and weaknesses of current AI models in optometric settings.

Main Methods:

• Systematic review of AI models applied across key optometric areas.
• Analysis of model selection (CNN, transfer learning, RF, SVM, XGBoost) and dataset parameters.
• Evaluation of model performance using accuracy metrics and assessment of overfitting and explainability.

Main Results:

• Image-based AI studies predominantly used CNN or transfer learning; clinical data studies used RF, SVM, and XGBoost.
• Larger datasets generally led to higher accuracy, but many optometry AI models suffered from insufficient training and validation sample sizes.
• Overfitting was observed in some studies, and 20% reported accuracy below 80%, impacting clinical utility.

Conclusions:

• AI holds promise for improving optometry, but current models require larger datasets and robust validation to enhance accuracy and reliability.
• Addressing limitations such as small sample sizes and overfitting is crucial for the successful clinical implementation of AI in eye care.
• This review offers insights for optometrists on the capabilities and limitations of AI, guiding informed adoption in practice.