Phenotyping of ABCA4 Retinopathy by Machine Learning Analysis of Full-Field Electroretinography

Abstract

Purpose Biallelic pathogenic variants in ABCA4 are the commonest cause of monogenic retinal disease. The full-field electroretinogram (ERG) quantifies severity of retinal dysfunction. We explored application of machine learning in ERG interpretation and in genotype–phenotype correlations. Methods International standard ERGs in 597 cases of ABCA4 retinopathy were classified into three functional phenotypes by human experts: macular dysfunction alone (group 1), or with additional generalized cone dysfunction (group 2), or both cone and rod dysfunction (group 3). Algorithms were developed for automatic selection and measurement of ERG components and for classification of ERG phenotype. Elastic-net regression was used to quantify severity of specific ABCA4 variants based on effect on retinal function. Results Of the cohort, 57.6%, 7.4%, and 35.0% fell into groups 1, 2, and 3 respectively. Compared with human experts, automated classification showed overall accuracy of 91.8% (SE, 0.169), and 96.7%, 39.3%, and 93.8% for groups 1, 2, and 3. When groups 2 and 3 were combined, the average holdout group accuracy was 93.6% (SE, 0.142). A regression model yielded phenotypic severity scores for the 47 commonest ABCA4 variants. Conclusions This study quantifies prevalence of phenotypic groups based on retinal function in a uniquely large single-center cohort of patients with electrophysiologically characterized ABCA4 retinopathy and shows applicability of machine learning. Novel regression-based analyses of ABCA4 variant severity could identify individuals predisposed to severe disease. Translational Relevance Machine learning can yield meaningful classifications of ERG data, and data-driven scoring of genetic variants can identify patients likely to benefit most from future therapies.

Distinct Clinical Effects of Two RP1L1 Hotspots in East Asian Patients With Occult Macular Dystrophy (Miyake Disease): EAOMD Report 4

“…AI-guided genotype/variant-based characterization based on ffERGs has also been developed. 26 Thus, an AI-guided combination of retinal imaging and functional assessment can provide a more comprehensive machine diagnosis. AI has the potential to facilitate prognostic predictions in patients with Miyake disease at the earliest stages of genetic testing, which can significantly influence their life plans.…”

“… 20 – 22 Deep learning techniques have proven successful in ophthalmology, 23 with machine learning–assisted diagnosis gaining widespread recognition in retinal disease management. 24 – 26 Accurate diagnosis of inherited retinal disease (IRD), particularly OMD, is challenging because of limited access to multidisciplinary specialist teams. Artificial intelligence (AI)–based diagnostic platforms that rely on color fundus photographs (CFPs), fundus autofluorescence (FAF) images, and SD-OCT images have been developed based on retinal images of IRDs.…”

Visual electrophysiology and “the potential of the potentials”

“…With development of newer devices (including portable and multimodal technology), and more refined, including more rapid, testing protocols, combined with novel, AI-assisted analyses, it is likely that tests will become more accessible and continue to yield valuable clinical and scientific information. This will have relevance to common and rare diseases of the eye and visual pathway, and also, given similarities between retinal and brain circuitry, potentially to wider neurological and neuropsychiatric disease [71,85,88,[92][93][94].…”

“…Artificial intelligence (AI) is being applied to many areas of healthcare showing high levels of accuracy, equivalent to experts. There have been investigations applying AI or machine-learning techniques to electrophysiology data [34,[85][86][87][88][89][90]. These include studies of ERG data that may have applicability in conditions including glaucoma [88], hydroxychloroquine retinopathy [87], and even autism spectrum disorder [86] and depression [89], well as studies of VEP data to improve estimation of acuity [90].…”

Prognostication in Stargardt Disease Using Fundus Autofluorescence: Improving Patient Care

“…These findings are supported by the association with genotype grouping (e.g., group 1 harboring milder variants, whereas group 3 is associated with a greater prevalence of null variants). 13 , 20 , 21 Further analysis demonstrated that those with abnormal ffERG results also showed decreased BCVA and higher rate of scotoma and atrophy enlargement than those with normal ffERG results. 15 , 22 …”