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

IF 4.6 Q1 OPHTHALMOLOGY

Ophthalmology science Pub Date : 2025-06-04 DOI:10.1016/j.xops.2025.100845

Gunjan Naik PhD , Saoud Al-Khuzaei MD, PhD , Ismail Moghul PhD , Thales A.C. de Guimaraes PhD, MD , Sagnik Sen MD , Malena Daich Varela MD, PhD , Yichen Liu MSci , Pallavi Bagga PhD , Vincent Rocco PGCERT , Dun Jack Fu MD, PhD , Mariya Moosajee MD, PhD , Savita Madhusudhan MD , Andrew R. Webster MD , Samantha De Silva MD, PhD , Praveen J. Patel MD , Omar A. Mahroo MD, PhD , Susan M. Downes MD , Michel Michaelides MD , Konstantinos Balaskas MD , Nikolas Pontikos PhD , William Woof A. PhD

{"title":"Retinograd-AI: An Open-Source Automated Fundus Autofluorescence Retinal Image Gradability Assessment for Inherited Retinal Diseases","authors":"Gunjan Naik PhD , Saoud Al-Khuzaei MD, PhD , Ismail Moghul PhD , Thales A.C. de Guimaraes PhD, MD , Sagnik Sen MD , Malena Daich Varela MD, PhD , Yichen Liu MSci , Pallavi Bagga PhD , Vincent Rocco PGCERT , Dun Jack Fu MD, PhD , Mariya Moosajee MD, PhD , Savita Madhusudhan MD , Andrew R. Webster MD , Samantha De Silva MD, PhD , Praveen J. Patel MD , Omar A. Mahroo MD, PhD , Susan M. Downes MD , Michel Michaelides MD , Konstantinos Balaskas MD , Nikolas Pontikos PhD , William Woof A. PhD","doi":"10.1016/j.xops.2025.100845","DOIUrl":null,"url":null,"abstract":"<div><h3>Purpose</h3><div>To develop an automated system for assessing the quality of fundus autofluorescence (FAF) images in patients with inherited retinal diseases (IRDs).</div></div><div><h3>Design</h3><div>A retrospective study of imaging data.</div></div><div><h3>Participants</h3><div>Patients with a confirmed molecular diagnosis of IRD who have undergone FAF imaging at Moorfields Eye Hospital.</div></div><div><h3>Methods</h3><div>A dataset of 2445 FAF images from patients with IRD was marked by 3 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 (AI) algorithm, Retinograd-AI, which was then applied to predict the gradability label of our entire dataset of 136 631 FAF images.</div></div><div><h3>Main Outcome Measures</h3><div>Fundus autofluorescence gradability of FAF images as predicted and validated against human assessment.</div></div><div><h3>Results</h3><div>Retinograd-AI achieves 91% accuracy on our held-out dataset of 133 images with an area under the receiver operator characteristic curve 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, <em>P</em> < 0.001). Excluding X-linked conditions, 77.1% of images were rated as gradable in men and 82.3% in women (odds ratio = 1.43, <em>P</em> < 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 <em>PRPH2</em> (93.1%), while the lowest was in <em>RDH12</em> (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 <span><span>https://github.com/Eye2Gene/retinograd-ai</span><svg><path></path></svg></span>.</div></div><div><h3>Conclusions</h3><div>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 the 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.</div></div><div><h3>Financial Disclosure(s)</h3><div>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</div></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"5 6","pages":"Article 100845"},"PeriodicalIF":4.6000,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Ophthalmology science","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2666914525001435","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"OPHTHALMOLOGY","Score":null,"Total":0}

引用次数: 0

Abstract

Purpose

To develop an automated system for assessing the quality of fundus autofluorescence (FAF) images in patients with inherited retinal diseases (IRDs).

Design

A 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 2445 FAF images from patients with IRD was marked by 3 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 (AI) algorithm, Retinograd-AI, which was then applied to predict the gradability label of our entire dataset of 136 631 FAF images.

Main Outcome Measures

Fundus autofluorescence gradability of FAF images as predicted and 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 curve 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 men and 82.3% in women (odds ratio = 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 the 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.

Financial Disclosure(s)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

查看原文本刊更多论文

Retinograd-AI：一种用于遗传性视网膜疾病的开源眼底自动荧光视网膜图像可分级性评估

目的建立一套用于遗传性视网膜疾病（IRDs）患者眼底自身荧光（FAF）图像质量评估的自动化系统。设计：影像学资料的回顾性研究。参与者：在Moorfields眼科医院接受FAF成像的确诊分子诊断为IRD的患者。方法根据严格的评分方案，对来自IRD患者的2445张FAF图像进行数据集评分，由3名专家评分，分为可分级（质量可接受）和不可分级（质量差）。该数据集用于训练人工智能（AI）算法Retinograd-AI，然后应用该算法预测136 631张FAF图像的整个数据集的可分级标签。FAF图像的眼底自身荧光可分级性预测和验证了人类评估。结果retina - grad- ai在133张图像的hold out数据集上达到了91%的准确率，接收算子特征曲线下的面积为0.94，表明在区分可分级和不可分级图像方面具有很高的性能。将Retinograd-AI应用于我们的整个数据集，发现可分级性与年龄存在小但显著的正相关(ß = 0.002, P <；0.001)。排除x连锁条件，77.1%的男性和82.3%的女性图像被评为可分级(优势比= 1.43,P <；0.001)。根据基因型，从我们数据集中30个最常见的遗传诊断中，可分级图像比例最高的是PRPH2致病变异患者（93.1%），而最低的是RDH12（27.1%）。将Retinograd-AI应用于过滤图像将基因预测分类器的准确率从33.8%提高到68.9%。Retinograd-AI是开源的，可在https://github.com/Eye2Gene/retinograd-ai.ConclusionsRetinograd-AI上获得，是一个开源的AI模型，用于自动评估ird中FAF图像的视网膜图像质量。通过Retinograd-AI进行自动分级评估，可以对视网膜图像进行大规模分析，并开发强大的分析管道。质量评估对于将Eye2Gene等人工智能算法部署到临床环境中至关重要。由于IRD病理的多样性，Retinograd-AI将以目前的形式或通过迁移学习和微调扩展到其他条件。财务披露专有或商业披露可在本文末尾的脚注和披露中找到。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊