Assessing the Impact of Image Quality on Deep Learning Classification of Infectious Keratitis.

Objective: To investigate the impact of corneal photograph quality on convolutional neural network (CNN) predictions. Design: A CNN trained to classify bacterial and fungal keratitis was evaluated using photographs of ulcers labeled according to 5 corneal image quality parameters: eccentric gaze direction, abnormal eyelid position, over/under-exposure, inadequate focus, and malpositioned light reflection. Participants: All eligible subjects with culture and stain-proven bacterial and/or fungal ulcers presenting to Aravind Eye Hospital in Madurai, India, between January 1, 2021 and December 31, 2021. Methods: Convolutional neural network classification performance was compared for each quality parameter, and gradient class activation heatmaps were generated to visualize regions of highest influence on CNN predictions. Main Outcome Measures: Area under the receiver operating characteristic and precision recall curves were calculated to quantify model performance. Bootstrapped confidence intervals were used for statistical comparisons. Logistic loss was calculated to measure individual prediction accuracy. Results: Individual presence of either light reflection or eyelids obscuring the corneal surface was associated with significantly higher CNN performance. No other quality parameter significantly influenced CNN performance. Qualitative review of gradient class activation heatmaps generally revealed the infiltrate as having the highest diagnostic relevance. Conclusions: The CNN demonstrated expert-level performance regardless of image quality. Future studies may investigate use of smartphone cameras and image sets with greater variance in image quality to further explore the influence of these parameters on model performance. Financial Disclosures: Proprietary or commercial disclosure may be found after the references.

GCL loss in BRAO.

PURPOSE: Our recent publication used optical coherence tomography (OCT) to follow thinning of the retinal ganglion cell layer (GCL) in central retinal artery occlusion (CRAO). Thinning of the inner layers also occurs in patients with branch retinal artery occlusion (BRAO). The mechanism for such thinning may be partially due to proteolysis by a calcium-activated protease called calpain. Calpain inhibitor SNJ-1945 ameliorated the proteolysis in a past series of model experiments. The purposes of the present retrospective study were to: 1) use segmentation analysis of OCT images to follow the loss of retinal layers in BRAO compared to CRAO patients, and 2) predict the number of patients and days of observation needed for a clinical trial of a calpain inhibitor against BRAO. METHODS: A retrospective, case control study was conducted by computer-aided search in a medical records database for BRAO (ICD10 Code H34.239) with at least one OCT procedure (CPT: 92134). Non-proliferative, co-morbid eye diseases were allowed in the patient data base, and manual correction of auto-segmentation errors was performed. GCL thickness changes were followed over time and Cohen-d/sample size statistics were used to predict minimal patients needed for drug trials. RESULTS: The thickness of the GCL layer in BRAO decreased rapidly with time as in CRAO, but in more limited quadrants. The data, as fit to a single-phase decay curve, showed that GCL thickness could be used to provide sample size statistics in a clinical trial to test a calpain inhibitor. For example, a 60-day trial with a 60% effective inhibitor would need a minimum of 29 patients. CONCLUSIONS: Using thickness changes in the GCL layer to monitor the efficacy of potential inhibitors against BRAO and CRAO is practical in human trials requiring a reasonable number of patients and relatively short trial period. TRANSLATIONAL RELEVANCE: Measurement of GCL thickness would be a useful indicator of amelioration of BRAO and CRAO progression in a clinical trial of a putative inhibitor.

Three-Year Safety Results of SAR422459 (EIAV-ABCA4) Gene Therapy in Patients With ABCA4-Associated Stargardt Disease: An Open-Label Dose-Escalation Phase I/IIa Clinical Trial, Cohorts 1-5.

PURPOSE: To report on the safety of the first 5 cohorts of a gene therapy trial using recombinant equine infectious anemia virus expressing ABCA4 (EIAV-ABCA4) in adults with Stargardt dystrophy due to mutations in ABCA4. DESIGN: Nonrandomized multicenter phase I/IIa clinical trial. METHODS: Patients received a subretinal injection of EIAVABCA4 in the worse-seeing eye at 3 dose levels and were followed for 3 years after treatment. MAIN OUTCOME MEASURES: The primary end point was ocular and systemic adverse events. The secondary end points were best-corrected visual acuity, static perimetry, kinetic perimetry, total field hill of vision, full field electroretinogram, multifocal ERG, color fundus photography, short-wavelength fundus autofluorescence, and spectral domain optical coherence tomography. RESULTS: The subretinal injections were well tolerated by all 22 patients across 3 dose levels. There was 1 case of a treatment-related ophthalmic serious adverse event in the form of chronic ocular hypertension. The most common adverse events were associated with the surgical procedure. In 1 patient treated with the highest dose, there was a significant decline in the number of macular flecks as compared with the untreated eye. However, in 6 patients, hypoautofluorescent changes were worse in the treated eye than in the untreated eye. Of these, 1 patient had retinal pigment epithelium atrophy that was characteristic of tissue damage likely associated with bleb induction. No patients had any clinically significant changes in best-corrected visual acuity, static perimetry, kinetic perimetry, total field hill of vision, full field electroretinogram, or multifocal ERG attributable to the treatment. CONCLUSIONS: Subretinal treatment with EIAV-ABCA4 was well tolerated with only 1 case of ocular hypertension. No clinically significant changes in visual function tests were found to be attributable to the treatment. However, 27% of treated eyes showed exacerbation of retinal pigment epithelium atrophy on fundus autofluorescence. There was a significant reduction in macular flecks in 1 treated eye from the highest dose cohort. Additional follow-up and continued investigation in more patients will be required to fully characterize the safety and efficacy of EIAV-ABCA4.

Segmenting OCT for detecting drug efficacy in CRAO.

PURPOSE: Thinning of the inner layers of the retina occurs in patients with central retinal artery occlusion (CRAO). The mechanism for such thinning may be partially due to proteolysis by a calcium-activated protease called calpain. Calpain inhibitor SNJ-1945 ameliorated the proteolysis in a past series of model experiments. The purposes of the present retrospective study were to: 1) use segmentation analysis of optical coherence tomography (OCT) images to mathematically model the loss of specific retinal layers in CRAO patients, and 2) predict the number of patients and days of observation needed for clinical trials of inhibitors against CRAO. METHODS: A retrospective case control study was conducted by computer-aided search for CRAO (ICD10 H43.1) with at least one OCT procedure (CPT: 92134) in the OHSU Epic patient data base. RESULTS: After initial swelling, thinning of the inner retinal layers, especially the ganglion cell (GCL) layer followed exponential decay curves. Using sample size statistics and GCL thickness as a marker in a 30-day clinical trial, 19 eyes/group could theoretically detect a 20% beneficial effect of an inhibitor against CRAO. Other markers, such as the whole retinal thickness and combined inner layers could also be used as less-specific markers. CONCLUSIONS: Using thickness changes in the GCL layer to monitor the efficacy of potential inhibitors against CRAO is practical in human trials requiring a reasonable number of patients and relatively short trial period. TRANSLATIONAL RELEVANCE: Measurement of GCL thickness would be a useful indicator of CRAO progression in a clinical trial of putative inhibitors.

Reliability of Spectral-Domain OCT Ellipsoid Zone Area and Shape Measurements in Retinitis Pigmentosa.

PURPOSE: We investigate the ellipsoid zone (EZ) area and EZ boundary shape measurement reliability and the operability characteristics of two methods of EZ boundary delineation in spectral-domain optical coherence tomography (SD-OCT). METHODS: EZ boundaries in SD-OCT scans of 122 eyes from 64 subjects with autosomal dominant retinitis pigmentosa were delineated by three raters using two methods, termed the profile and en face methods. For each method, we determined the measurement reliabilities for boundary area (EZ area) and boundary shape, percentage of eyes with measurable EZ (measurability), time required, and effect of rater experience. RESULTS: With expert raters, inter- and intrarater area intraclass correlation coefficients (ICCs) were 0.986 and 0.980 (profile) and 0.959 and 0.976 (en face), respectively. In comparison, the corresponding shape ICCs were 0.906 and 0.891 (profile) and 0.845 and 0.885 (en face), indicating lower reliability for the raw measurements (P ≤ 0.01). Only profile method interrater reliability depended on experience. Average measurement times per eye were 8.2 (profile) and 4.1 (en face) minutes. Measurability percentages were 99.2% (profile) and 73.0% (en face). CONCLUSIONS: The slower profile method had better measurability, and with expert raters yielded the best area and shape reliabilities. The faster, but less sensitive, en face method still showed excellent reliability, and was less dependent on experience. Shape analysis reveals the boundary measurements underpinning EZ area have lower reliability than suggested by area analysis. TRANSLATIONAL RELEVANCE: This study provides new reliability perspectives and logistical considerations for the manual measurement procedures that generate EZ area outcome measures.

Test-Retest Variability of Functional and Structural Parameters in Patients with Stargardt Disease Participating in the SAR422459 Gene Therapy Trial.

PURPOSE: The goal of this analysis was to determine the test-retest variability of functional and structural measures from a cohort of patients with advanced forms of Stargardt Disease (STGD) participating in the SAR422459 (NCT01367444) gene therapy clinical trial. METHODS: Twenty-two participants, aged 24 to 66, diagnosed with advanced forms of STGD, with at least one pathogenic ABCA4 mutation on each chromosome participating in the SAR422459 (NCT01367444) gene therapy clinical trial, were screened over three visits within 3 weeks or less. Functional visual evaluations included: best-corrected visual acuity (BCVA) Early Treatment Diabetic Retinopathy Study (ETDRS) letter score, semiautomated kinetic perimetry (SKP) using isopters I4e, III4e, and V4e, hill of vision (HOV) calculated from static visual fields (SVF) by using a 184n point centrally condensed grid with the stimulus size V test target. Retinal structural changes such as central macular thickness and macular volume were assessed by spectral-domain optical coherence tomography (SD-OCT). Repeatability coefficients (RC) and 95% confidential intervals (CI) were calculated for each parameter using a hierarchical mixed-effects model and bootstrapping. RESULTS: Criteria for statistically significant changes for various parameters were found to be the following: BCVA letter score (8 letters), SKP isopters I4e, III4e, and V4e (3478.85; 2488.02 and 2622.46 deg2, respectively), SVF full volume HOV (VTOT, 14.62 dB-sr), central macular thickness, and macular volume (4.27 µm and 0.15 mm3, respectively). CONCLUSIONS: This analysis provides important information necessary to determine if significant changes are occurring in structural and functional assessments commonly used to measure disease progression in this cohort of patients with STGD. Moreover, this information is useful for future trials assessing safety and efficacy of treatments in STGD. TRANSLATIONAL RELEVANCE: Determination of variability of functional and structural measures in participants with advanced stages of the STGD is necessary to assess efficacy and safety in treatment trials involving STGD patients.

Structure-Function Modeling of Optical Coherence Tomography and Standard Automated Perimetry in the Retina of Patients with Autosomal Dominant Retinitis Pigmentosa.

PURPOSE: To assess relationships between structural and functional biomarkers, including new topographic measures of visual field sensitivity, in patients with autosomal dominant retinitis pigmentosa. METHODS: Spectral domain optical coherence tomography line scans and hill of vision (HOV) sensitivity surfaces from full-field standard automated perimetry were semi-automatically aligned for 60 eyes of 35 patients. Structural biomarkers were extracted from outer retina b-scans along horizontal and vertical midlines. Functional biomarkers were extracted from local sensitivity profiles along the b-scans and from the full visual field. These included topographic measures of functional transition such as the contour of most rapid sensitivity decline around the HOV, herein called HOV slope for convenience. Biomarker relationships were assessed pairwise by coefficients of determination (R2) from mixed-effects analysis with automatic model selection. RESULTS: Structure-function relationships were accurately modeled (conditional R(2)>0.8 in most cases). The best-fit relationship models and correlation patterns for horizontally oriented biomarkers were different than vertically oriented ones. The structural biomarker with the largest number of significant functional correlates was the ellipsoid zone (EZ) width, followed by the total photoreceptor layer thickness. The strongest correlation observed was between EZ width and HOV slope distance (marginal R(2) = 0.85, p<10(-10)). The mean sensitivity defect at the EZ edge was 7.6 dB. Among all functional biomarkers, the HOV slope mean value, HOV slope mean distance, and maximum sensitivity along the b-scan had the largest number of significant structural correlates. CONCLUSIONS: Topographic slope metrics show promise as functional biomarkers relevant to the transition zone. EZ width is strongly associated with the location of most rapid HOV decline.