Objective:

To examine the real-world clinical value of AI in ophthalmology, moving beyond technical novelty to assess its impact on eye care delivery.

Approach:

• Focus on Clinical Value: The review emphasizes the need to evaluate whether AI improves access to care, reduces clinician workload, and enhances treatment timeliness and health equity.
• Case Studies: The review discusses specific applications of AI, particularly in diabetic retinopathy screening and OCT triage, to illustrate the current state of AI in clinical practice.

Key Findings:

• Diabetic retinopathy screening has shown significant progress, with autonomous AI systems achieving regulatory approval and demonstrating operational benefits, as evidenced by studies such as those by Gulshan et al. and Abramoff et al.
• AI for OCT interpretation has potential but faces challenges in generalization and implementation across different settings, as highlighted in the research by De Fauw et al.
• AI can enhance outreach and community screening, particularly in underserved areas, by providing specialist-grade triage, as indicated by recent utilization data.

Interpretation:

The distinction between technical capability and clinical usefulness is crucial, as many AI tools remain underutilized despite their potential benefits being recognized in various studies.

Limitations:

• Adoption of AI tools in ophthalmology is still limited despite regulatory approvals and positive trial results, with many applications lacking sufficient prospective multicenter implementation evidence.
• Many promising AI applications lack sufficient prospective multicenter implementation evidence, which hinders their widespread adoption.

Conclusion:

The review advocates for a shift in focus from merely assessing AI's accuracy to evaluating its real-world impact on eye care delivery.

Sources:

• Gulshan et al. Study
• Abramoff et al. Trial
• Ipp et al. Study
• De Fauw et al. Research
• Cui et al. Study