Objective:
To evaluate the accuracy of AI models in diagnosing inherited retinal diseases (IRDs) through a systematic review and meta-analysis.
Key Findings:
- AI achieved up to 99.9% accuracy for retinitis pigmentosa (RP) with 94% sensitivity and 99% specificity.
- Stargardt disease detection showed 96% sensitivity and 99% specificity.
- Familial exudative vitreoretinopathy (FEVR) had 85% sensitivity and 99% specificity.
- Widefield imaging improved specificity for RP detection compared to non-widefield studies.
- Error analysis identified false positives due to high myopia and false negatives from media opacities.
Interpretation:
AI models demonstrate high diagnostic performance for IRDs, particularly for RP and Stargardt disease, suggesting potential for revolutionizing retinal disease diagnosis.
Limitations:
- Moderate certainty for pooled sensitivity and specificity; low certainty for diagnostic odds ratio due to potential publication bias.
- Moderate heterogeneity across studies (I² = 45%).
- Need for further research to ensure robustness and applicability of AI models in diverse clinical settings.
Conclusion:
AI has significant potential to enhance early diagnosis and personalized care for patients with IRDs, but further research is necessary to address existing limitations.
This content is an AI-generated, fully rewritten summary based on a published scholarly article. It does not reproduce the original text and is not a substitute for the original publication. Readers are encouraged to consult the source for full context, data, and methodology.
