Clinical Report: AI Achieves High Accuracy in Diagnosing Inherited Retinal Diseases
Overview
A systematic review and meta-analysis demonstrated that artificial intelligence (AI), particularly deep learning models, can detect inherited retinal diseases (IRDs) such as retinitis pigmentosa (RP) and Stargardt disease with very high accuracy. AI models showed pooled sensitivities ranging from 85% to 96% and specificities up to 99%, with some accuracies reaching 99.9%.
Background
Inherited retinal diseases (IRDs) are a group of genetic disorders that cause progressive vision loss and blindness. Early and accurate diagnosis is critical for patient management and potential therapeutic interventions. Traditional diagnostic methods rely on clinical examination and imaging modalities such as optical coherence tomography (OCT), fundus photography, and fundus autofluorescence (FAF). Recent advances in artificial intelligence (AI), especially deep learning convolutional neural networks (CNNs), have shown promise in enhancing diagnostic accuracy for IRDs.
Data Highlights
| Disease | Sensitivity | Specificity | Accuracy | Imaging Modalities |
|---|---|---|---|---|
| Retinitis Pigmentosa (RP) | 94% | 99% | Up to 99.9% | OCT, FAF, Widefield Imaging, Fundus Photography |
| Stargardt Disease | 96% | 99% | High (AUC up to 0.998) | FAF, OCT |
| Familial Exudative Vitreoretinopathy (FEVR) | 85% | 99% | 89% to 94% | Wide-angle Retinal Imaging |
Key Findings
- Deep learning models, especially those using CNN architectures like ResNet, Inception, and Xception, achieved high diagnostic performance for IRDs.
- Retinitis pigmentosa detection showed pooled sensitivity of 94%, specificity of 99%, and accuracies up to 99.9% across multiple imaging modalities.
- Stargardt disease detection had similarly high pooled sensitivity (96%) and specificity (99%), with FAF and OCT showing excellent diagnostic metrics.
- Widefield and ultrawidefield imaging significantly improved specificity for RP detection compared to other imaging techniques (99% vs 98%, P=0.03).
- False positives were often due to high myopia mimicking RP, while false negatives were frequently caused by media opacities and reduced image contrast.
- Transfer learning enhanced model performance, particularly in studies with smaller datasets, and the Xception model outperformed other CNNs in capturing complex retinal patterns.
Clinical Implications
AI-based diagnostic tools can substantially improve the accuracy and early detection of inherited retinal diseases, potentially enabling more personalized patient care. Incorporating widefield imaging and advanced CNN models such as Xception may optimize diagnostic performance. Clinicians should be aware of potential false positives related to myopia and false negatives due to media opacities when interpreting AI results.
Conclusion
AI models demonstrate high diagnostic accuracy for IRDs, particularly retinitis pigmentosa and Stargardt disease, across various imaging modalities. Continued research is necessary to validate these tools across diverse clinical settings and to address current limitations for broader clinical implementation.
References
- Shahid Beheshti University of Medical Sciences, 2024 -- AI in IRD Diagnosis: Systematic Review and Meta-analysis
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.
