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

Artificial intelligence for detection of retinal toxicity in chloroquine and hydroxychloroquine therapy using multifocal electroretinogram waveforms

Chloroquine and hydroxychloroquine, while effective in rheumatology, pose risks of retinal toxicity, necessitating regular screening to prevent visual disability. The gold standard for screening includes retinal imaging and automated perimetry, with multifocal electroretinography (mfERG) being a recognized but less accessible method. This study explores the efficacy of Artificial Intelligence (AI) algorithms for detecting retinal damage in patients undergoing (hydroxy-)chloroquine therapy. We analyze the mfERG data, comparing the performance of AI models that utilize raw mfERG time-series signals against models using conventional waveform parameters. Our classification models aimed to identify maculopathy, and regression models were developed to predict perimetric sensitivity. The findings reveal that while regression models were more adept at predicting non-disease-related variation, AI-based models, particularly those utilizing full mfERG traces, demonstrated superior predictive power for disease-related changes compared to linear models. This indicates a significant potential to improve diagnostic capabilities, although the unbalanced nature of the dataset may limit some applications.

Generating Synthetic Light-Adapted Electroretinogram Waveforms Using Artificial Intelligence to Improve Classification of Retinal Conditions in Under-Represented Populations

Visual electrophysiology is often used clinically to determine the functional changes associated with retinal or neurological conditions. The full-field flash electroretinogram (ERG) assesses the global contribution of the outer and inner retinal layers initiated by the rods and cone pathways depending on the state of retinal adaptation. Within clinical centers, reference normative data are used to compare clinical cases that may be rare or underpowered within a specific demographic. To bolster either the reference dataset or the case dataset, the application of synthetic ERG waveforms may offer benefits to disease classification and case-control studies. In this study and as a proof of concept, artificial intelligence (AI) to generate synthetic signals using generative adversarial networks is deployed to upscale male participants within an ISCEV reference dataset containing 68 participants, with waveforms from the right and left eye. Random forest classifiers further improved classification for sex within the group from a balanced accuracy of 0.72-0.83 with the added synthetic male waveforms. This is the first study to demonstrate the generation of synthetic ERG waveforms to improve machine learning classification modelling with electroretinogram waveforms.

Phenotyping and genotyping inherited retinal diseases: Molecular genetics, clinical and imaging features, and therapeutics of macular dystrophies, cone and cone-rod dystrophies, rod-cone dystrophies, Leber congenital amaurosis, and cone dysfunction syndromes

Inherited retinal diseases (IRD) are a leading cause of blindness in the working age population and in children. The scope of this review is to familiarise clinicians and scientists with the current landscape of molecular genetics, clinical phenotype, retinal imaging and therapeutic prospects/completed trials in IRD. Herein we present in a comprehensive and concise manner: (i) macular dystrophies (Stargardt disease (ABCA4), X-linked retinoschisis (RS1), Best disease (BEST1), PRPH2-associated pattern dystrophy, Sorsby fundus dystrophy (TIMP3), and autosomal dominant drusen (EFEMP1)), (ii) cone and cone-rod dystrophies (GUCA1A, PRPH2, ABCA4, KCNV2 and RPGR), (iii) predominant rod or rod-cone dystrophies (retinitis pigmentosa, enhanced S-Cone syndrome (NR2E3), Bietti crystalline corneoretinal dystrophy (CYP4V2)), (iv) Leber congenital amaurosis/early-onset severe retinal dystrophy (GUCY2D, CEP290, CRB1, RDH12, RPE65, TULP1, AIPL1 and NMNAT1), (v) cone dysfunction syndromes (achromatopsia (CNGA3, CNGB3, PDE6C, PDE6H, GNAT2, ATF6), X-linked cone dysfunction with myopia and dichromacy (Bornholm Eye disease; OPN1LW/OPN1MW array), oligocone trichromacy, and blue-cone monochromatism (OPN1LW/OPN1MW array)). Whilst we use the aforementioned classical phenotypic groupings, a key feature of IRD is that it is characterised by tremendous heterogeneity and variable expressivity, with several of the above genes associated with a range of phenotypes.

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

Purpose

To characterize the clinical effects of two RP1L1 hotspots in patients with East Asian occult macular dystrophy (OMD).

Methods

Fifty-one patients diagnosed with OMD harboring monoallelic pathogenic RP1L1 variants (Miyake disease) from Japan, South Korea, and China were enrolled. Patients were classified into two genotype groups: group A, p.R45W, and group B, missense variants located between amino acids (aa) 1196 and 1201. The clinical parameters of the two genotypes were compared, and deep learning based on spectral-domain optical coherence tomographic (SD-OCT) images was used to distinguish the morphologic differences.

Results

Groups A and B included 29 and 22 patients, respectively. The median age of onset in groups A and B was 14.0 and 40.0 years, respectively. The median logMAR visual acuity of groups A and B was 0.70 and 0.51, respectively, and the survival curve analysis revealed a 15-year difference in vision loss (logMAR 0.22). A statistically significant difference was observed in the visual field classification, but no significant difference was found in the multifocal electroretinographic classification. High accuracy (75.4%) was achieved in classifying genotype groups based on SD-OCT images using machine learning.

Conclusions

Distinct clinical severities and morphologic phenotypes supported by artificial intelligence-based classification were derived from the two investigated RP1L1 hotspots: a more severe phenotype (p.R45W) and a milder phenotype (1196-1201 aa). This newly identified genotype-phenotype association will be valuable for medical care and the design of therapeutic trials.

Prognostication in Stargardt Disease Using Fundus Autofluorescence: Improving Patient Care

PURPOSE

To explore fundus autofluorescence (FAF) imaging as an alternative to electroretinography as a noninvasive, quick, and readily interpretable method to predict disease progression in Stargardt disease (STGD).

DESIGN

Retrospective case series of patients who attended Moorfields Eye Hospital (London, United Kingdom).

PARTICIPANTS

Patients with STGD who met the following criteria were included: (1) biallelic disease-causing variants in ABCA4, (2) electroretinography testing performed in house with an unequivocal electroretinography group classification, and (3) ultrawidefield (UWF) FAF imaging performed up to 2 years before or after the electroretinography.

METHODS

Patients were divided into 3 electroretinography groups based on retinal function and 3 FAF groups according to the extent of hypoautofluorescence and retinal background appearance. Fundus autofluorescence images of 30° and 55° were reviewed subsequently.

MAIN OUTCOME MEASURES

Electroretinography and FAF concordance and its association with baseline visual acuity (VA) and genetics.

RESULTS

Two hundred thirty-four patients were included in the cohort. One hundred seventy patients (73%) were in electroretinography and FAF groups of the same severity, 33 (14%) were in a milder FAF than electroretinography group, and 31 (13%) were in a more severe FAF than electroretinography group. Children < 10 years of age (n = 23) showed the lowest electroretinography and FAF concordance at 57% (9 of the 10 with discordant electroretinography and FAF showed milder FAF than electroretinography), and adults with adult onset showed the highest (80%). In 97% and 98% of patients, 30° and 55° FAF imaging, respectively, matched with the group defined by UWF FAF.

CONCLUSIONS

We demonstrated that FAF imaging is an effective method to determine the extent of retinal involvement and thereby inform prognostication by comparing FAF with the current gold standard of electroretinography. In 80% of patients in our large molecularly proven cohort, we were able to predict if the disease was confined to the macula or also affected the peripheral retina. Children assessed at a young age, with at least 1 null variant, early disease onset, poor initial VA, or a combination thereof may have wider retinal involvement than predicted by FAF alone, may progress to a more severe FAF phenotype over time, or both.

FINANCIAL DISCLOSURE(S)

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

Enhancing Electroretinogram Classification with Multi-Wavelet Analysis and Visual Transformer

The electroretinogram (ERG) is a clinical test that records the retina's electrical response to light. Analysis of the ERG signal offers a promising way to study different retinal diseases and disorders. Machine learning-based methods are expected to play a pivotal role in achieving the goals of retinal diagnostics and treatment control. This study aims to improve the classification accuracy of the previous work using the combination of three optimal mother wavelet functions. We apply Continuous Wavelet Transform (CWT) on a dataset of mixed pediatric and adult ERG signals and show the possibility of simultaneous analysis of the signals. The modern Visual Transformer-based architectures are tested on a time-frequency representation of the signals. The method provides 88% classification accuracy for Maximum 2.0 ERG, 85% for Scotopic 2.0, and 91% for Photopic 2.0 protocols, which on average improves the result by 7.6% compared to previous work.

Group comparisons of the individual electroretinogram time trajectories for the ascending limb of the b-wave using a raw and registered time series

OBJECTIVES

The electroretinogram is a clinical test commonly used in the diagnosis of retinal disorders with the peak time and amplitude of the a- and b-waves used as the main indicators of retinal function. However, subtle changes that affect the shape of the electroretinogram waveform may occur in the early stages of disease or in conditions that have a neurodevelopmental or neurodegenerative origin. In such cases, we introduce a statistical approach to mathematically model the shape of the electroretinogram waveform that may aid clinicians and researchers using the electroretinogram or other biological signal recordings to identify morphological features in the waveforms that may not be captured by the time or time-frequency domains of the waveforms. We present a statistical graphics-based analysis of the ascending limb of the b-wave (AL-b) of the electroretinogram in children with and without a diagnosis of autism spectrum disorder (ASD) with a narrative explanation of the statistical approach to illustrate how different features of the waveform based on location and scale derived from raw and registered time series can reveal subtle differences between the groups.

RESULTS

Analysis of the raw time trajectories confirmed findings of previous studies with a reduced and delayed b-wave amplitude in ASD. However, when the individual time trajectories were registered then group differences were visible in the mean amplitude at registered time ~ 0.6 suggesting a novel method to differentiate groups using registration of the ERG waveform.

The Value of Electroretinography in Identifying Candidate Genes for Inherited Retinal Dystrophies: A Diagnostic Guide

Inherited retinal dystrophies (IRDs) are a group of heterogeneous diseases caused by genetic mutations that specifically affect the function of the rod, cone, or bipolar cells in the retina. Electroretinography (ERG) is a diagnostic tool that measures the electrical activity of the retina in response to light stimuli, and it can help to determine the function of these cells. A normal ERG response consists of two waves, the a-wave and the b-wave, which reflect the activity of the photoreceptor cells and the bipolar and Muller cells, respectively. Despite the growing availability of next-generation sequencing (NGS) technology, identifying the precise genetic mutation causing an IRD can be challenging and costly. However, certain types of IRDs present with unique ERG features that can help guide genetic testing. By combining these ERG findings with other clinical information, such as on family history and retinal imaging, physicians can effectively narrow down the list of candidate genes to be sequenced, thereby reducing the cost of genetic testing. This review article focuses on certain types of IRDs with unique ERG features. We will discuss the pathophysiology and clinical presentation of, and ERG findings on, these disorders, emphasizing the unique role ERG plays in their diagnosis and genetic testing.

Natural History of Visual Dysfunction in ABCA4 Retinopathy and Its Genetic Correlates

PURPOSE

To systematically assess the ability to detect change and retest reliability for a panel of visual function assessments in ABCA4 retinopathy.

DESIGN

Prospective natural history study (NCT01736293).

METHODS

Patients with at least 1 documented pathogenic ABCA4 variant and a clinical phenotype consistent with ABCA4 retinopathy were recruited from a tertiary referral center. Participants underwent longitudinal, multifaceted functional testing, including measures of function at fixation (best-corrected visual acuity, low-vision Cambridge Color Test), macular function (microperimetry), and retina-wide function (full-field electroretinography [ERG]). Two- and 5-year ability to detect change was determined based on the η2 statistic.

RESULTS

A total of 134 eyes from 67 participants with a mean follow-up of 3.65 years were included. In the 2-year interval, the microperimetry-derived perilesional sensitivity (η2 of 0.73 [0.53, 0.83]; -1.79 dB/y [-2.2, -1.37]) and mean sensitivity (η2 of 0.62 [0.38, 0.76]; -1.28 dB/y [-1.67, -0.89]) showed most change over time, but could only be recorded in 71.6% of the participants. In the 5-year interval, the dark-adapted ERG a- and b-wave amplitude showed marked change over time as well (eg, DA 30 a-wave amplitude with an η2 of 0.54 [0.34, 0.68]; -0.02 log10(µV)/y [-0.02, -0.01]). The genotype explained a large fraction of variability in the ERG-based age of disease initiation (adjusted R2 of 0.73) CONCLUSIONS: Microperimetry-based clinical outcome assessments were most sensitive to change but could only be acquired in a subset of participants. Across a 5-year interval, the ERG DA 30 a-wave amplitude was sensitive to disease progression, potentially allowing for more inclusive clinical trial designs encompassing the whole ABCA4 retinopathy spectrum.

Visual electrophysiology and "the potential of the potentials"

Visual electrophysiology affords direct, quantitative, objective assessment of visual pathway function at different levels, and thus yields information complementary to, and not necessarily obtainable from, imaging or psychophysical testing. The tests available, and their indications, have evolved, with many advances, both in technology and in our understanding of the neural basis of the waveforms, now facilitating more precise evaluation of physiology and pathophysiology. After summarising the visual pathway and current standard clinical testing methods, this review discusses, non-exhaustively, several developments, focusing particularly on human electroretinogram recordings. These include new devices (portable, non-mydiatric, multimodal), novel testing protocols (including those aiming to separate rod-driven and cone-driven responses, and to monitor retinal adaptation), and developments in methods of analysis, including use of modelling and machine learning. It is likely that several tests will become more accessible and useful in both clinical and research settings. In future, these methods will further aid our understanding of common and rare eye disease, will help in assessing novel therapies, and will potentially yield information relevant to neurological and neuro-psychiatric conditions.