Retinograd-AI: An Open-Source Automated Fundus Autofluorescence Retinal Image Gradability Assessment for Inherited Retinal Diseases

Le résumé fourni par la source

Purpose To develop an automated system for assessing the quality of Fundus Autofluorescence (FAF) images in patients with inherited retinal diseases (IRD). Design Retrospective study of imaging data. Participants Patients with a confirmed molecular diagnosis of IRD who have undergone FAF imaging at Moorfields Eye Hospital. Methods A dataset of 2,445 FAF images from patients with IRD were marked by three expert graders, as either gradable (acceptable quality) or ungradable (poor quality), following a strict grading protocol. This dataset was used to train an artificial intelligence algorithm, Retinograd-AI, which was then applied to predict the gradability label of our entire dataset of 136,631 FAF images. Main outcome measures FAF gradability of FAF images as predicted validated against human assessment. Results Retinograd-AI achieves 91% accuracy on our held-out dataset of 133 images with an Area Under the Receiver Operator Characteristic (AUROC) of 0.94, indicating high performance in distinguishing between gradable and ungradable images. Applying Retinograd-AI to our entire dataset, a small but significant positive association of gradability with age was found (ß=0.002, p<0.001). Excluding X-linked conditions, 77.1% of images were rated as gradable in males, and 82.3% in females (OR=1.43, p < 0.001). By genotype, from the 30 most common genetic diagnoses in our dataset, the highest proportion of gradable images was in patients with disease causing variants in PRPH2 (93.1%), while the lowest was in RDH12 (27.1%). Applying Retinograd-AI to filter images improved the accuracy of a gene prediction classifier from 33.8% to 68.9%. Retinograd-AI is open-sourced, and available at https://github.com/Eye2Gene/retinograd-ai Conclusions Retinograd-AI is an open-source AI model for automated retinal image quality assessment of FAF images in IRDs. Automated gradability assessment through Retinograd-AI enables large scale analysis of retinal images and the development of robust analysis pipelines. Quality assessment is essential for deployment of AI algorithms, such as Eye2Gene, into clinical settings. Due to the diverse nature of IRD pathologies, Retinograd-AI will be extended to other conditions, either in its current form or through transfer learning and fine-tuning.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.