AI-Assisted Optical Coherence Tomography Segmentation for Enhanced Diagnosis of Inherited Retinal Diseases.

PURPOSE: Inherited retinal diseases (IRDs) are rare and diverse, posing a diagnostic challenge in ophthalmology. This study aimed to determine whether artificial intelligence (AI)-assisted image processing can improve IRD diagnosis and provide insights into disease characteristics. We used an optical coherence tomography (OCT) segmentation algorithm to characterize retinal features in IRDs. Two control groups were included to enhance the contextual understanding of these features: healthy eyes and eyes with age-related macular degeneration (AMD). An AI-driven classification model was then used to classify the data into disease and control groups.

METHODS

We analyzed 327 images from 181 patients with IRD and 146 control individuals, including healthy subjects and patients with AMD. IRD cases were stratified into macular and retinal dystrophies. Automated segmentation of six retinal layers and detection of nine biomarkers were performed on retinal OCT images using the AI-based RetinAI Discovery tool. A random forest classifier differentiated macular IRD, retinal IRD, and controls.

RESULTS

The model detected IRD with 91% accuracy and achieved 91% accuracy in differentiating macular from retinal IRD. Key OCT features for differentiation included reduced perifoveal photoreceptor and outer nuclear layer thicknesses and increased retinal nerve fiber layer thickness in retinal IRD. Macular IRD featured significant foveal photoreceptor and outer nuclear layer thinning.

CONCLUSIONS

This study shows that standardized OCT image analysis combined with AI-based classification can accurately detect and stratify IRDs. The model's high accuracy highlights its potential as a reliable diagnostic tool in ophthalmology.

TRANSLATIONAL RELEVANCE: This AI-assisted OCT evaluation approach enhances ophthalmic diagnostics by improving IRD detection and classification.