Artificial Intelligence in Inherited Retinal Diseases: Imaging-Based Applications and Emerging Trends.

INTRODUCTION: Inherited retinal diseases (IRDs) represent a heterogeneous group of progressive disorders that lead to visual impairment and blindness while posing significant diagnostic and management challenges. Advances in retinal imaging have improved clinical assessment; however, early diagnosis, disease monitoring, and prognostic evaluation remain difficult, particularly given the genetic and phenotypic complexity of IRDs. Artificial intelligence (AI), especially deep learning (DL), has emerged as a promising tool for addressing these challenges in ophthalmology. METHODS: This narrative review summarizes recent peer-reviewed studies investigating the application of AI to retinal imaging modalities in IRDs. We focused on AI-based approaches for disease detection, segmentation, classification, and prognostic assessment using imaging techniques such as optical coherence tomography (OCT), fundus autofluorescence (FAF), and color fundus photography. Peer-reviewed studies published between 2019 and 2025 were reviewed. RESULTS: The reviewed literature demonstrates that AI models can achieve high performance in automated image analysis tasks, particularly in retinitis pigmentosa (RP) and Stargardt disease (STGD). DL-based methods showed strong results in retinal layer and lesion segmentation, disease classification, and structure-function analysis. Emerging applications include prognostic modeling and identification of retinal regions with preserved functional or therapeutic potential, although these remain less well explored. Key limitations across studies include small datasets, limited external validation, lack of longitudinal analyses, and minimal integration of explainable AI techniques. DISCUSSIONS: AI has substantial potential to support imaging-based diagnosis and management of IRDs, particularly through automated image analysis and quantitative assessment of retinal biomarkers. However, current evidence remains limited by small datasets, scarce external validation, lack of longitudinal analysis, and variability in imaging protocols across studies. Future research should emphasize collaborative multicenter efforts, integration of multimodal imaging data, longitudinal monitoring, independent external validation, and the development of interpretable AI systems to improve robustness and support broader clinical adoption.