Faculty Bibliography

PURPOSE/UNASSIGNED:We aim to study a case of pathologic myopia with visible light OCT. DESIGN/UNASSIGNED:Case report. SUBJECTS/UNASSIGNED:We recruit 1 patient with pathologic myopia presenting with an acute lacquer crack with submacular hemorrhage (SMH) in the right eye and a chronic lacquer crack in the left eye. METHODS/UNASSIGNED:We acquire visible light OCT images with 1 μm axial resolution. Images are processed to depict spectral centroid shift, and spectral fitting is performed to determine oxygen saturation. Results are compared to clinical imaging. MAIN OUTCOME MEASURES/UNASSIGNED:Visible light OCT images, spectral centroid shift (redshift), oximetry, and spectral fitting. RESULTS/UNASSIGNED:with spectral evidence of an overlying RPE deficit (deficient red shift) and photoreceptor abnormalities. CONCLUSIONS/UNASSIGNED:As visible light OCT technology advances, its application toward well-characterized human retinal pathology can clarify its utility. We describe the first case of visible light OCT applied to pathologic myopia, a primary RPE-BM-choriocapillaris interface disease. We confirm that extravascular hemoglobin can be subject to spectral fitting, and we quantify the oxygen saturation of acute SMH. We further show that structural changes in chronic lacquer cracks can be characterized with this new technology. FINANCIAL DISCLOSURES/UNASSIGNED:Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

BACKGROUND AND OBJECTIVE/UNASSIGNED:This study aimed to develop machine learning algorithms to predict visual and anatomic outcomes and treatment frequency in patients with macular edema following central retinal vein occlusion (MEfCRVO) after intravitreal aflibercept injections (IAI). PATIENTS AND METHODS/UNASSIGNED:= 91) trials treated with monthly IAI 2 mg for 24 weeks then pro re nata (PRN) through week 52 were used to develop machine learning algorithms. RESULTS/UNASSIGNED:= 0.76). PRN injection frequency from week 24 to 52 was predicted with high accuracy (AUC = 0.83). Univariate analyses confirmed all predictive factors. CONCLUSION/UNASSIGNED:Machine learning algorithms predicted outcomes and dosing frequency with high accuracy and may help inform patients' and clinicians' expectations during MEfCRVO management.

OBJECTIVE:Macular retinal nerve fibre layer (mRNFL) and ganglion cell-inner plexiform layer thickness (GC-IPL) measurements are important markers of diabetic retinal neurodegeneration (DRN). In this cross-sectional study, we aimed to quantify the contribution of total retinal thickness (TRT) and other factors in the variation of mRNFL and GC-IPL thickness among participants with diabetes. METHODS AND ANALYSIS/METHODS:We used macular-centred spectral domain-optical coherence tomography scans from participants with diabetes in the UK Biobank. Two multiple linear regression models (prior to and after adjusting for TRT) were used to determine factors associated with mRNFL and GC-IPL thicknesses. A p value of less than 0.05 was considered statistically significant. RESULTS:A total of 3832 eyes from 3832 participants with diabetes were analysed. Factors that explained the greatest variation in thickness were TRT (20.9% for mRNFL and 57.2% for GC-IPL), followed by spherical equivalent (8.0% for mRNFL only), gender (2.2% for mRNFL only) and age (1.4% for GC-IPL only). Other factors significantly associated with mRNFL and/or GC-IPL thickness explained less than 1% of the variation in their thicknesses. Self-reported ancestral background was not significantly associated with mRNFL thickness after accounting for TRT. CONCLUSIONS:Although many factors were significantly associated with mRNFL and GC-IPL thickness in participants with diabetes, they accounted for a fraction of the variation in the thickness of both layers. TRT explained most of the variation in these measurements, hence accounting for TRT is needed when using these metrics to evaluate DRN.

The availability of big data can transform the studies in biomedical research to generate greater scientific insights if expert labeling is available to facilitate supervised learning. However, data annotation can be labor-intensive and cost-prohibitive if pixel-level precision is required. Weakly supervised semantic segmentation (WSSS) with image-level labeling has emerged as a promising solution in medical imaging. However, most existing WSSS methods in the medical domain are designed for single-class segmentation per image, overlooking the complexities arising from the co-existence of multiple classes in a single image. Additionally, the multi-class WSSS methods from the natural image domain cannot produce comparable accuracy for medical images, given the challenge of substantial variation in lesion scales and occurrences. To address this issue, we propose a novel anomaly-guided mechanism (AGM) for multi-class segmentation in a single image on retinal optical coherence tomography (OCT) using only image-level labels. AGM leverages the anomaly detection and self-attention approach to integrate weak abnormal signals with global contextual information into the training process. Furthermore, we include an iterative refinement stage to guide the model to focus more on the potential lesions while suppressing less relevant regions. We validate the performance of our model with two public datasets and one challenging private dataset. Experimental results show that our approach achieves a new state-of-the-art performance in WSSS for lesion segmentation on OCT images.

PURPOSE:To assess the association of a novel spectral domain optical coherence tomography biomarker with 6-month visual acuity in in the Study of COmparative Treatments for REtinal Vein Occlusion 2. METHODS:Spectral domain optical coherence tomography volume scans were evaluated for inner retinal hyperreflectivity, quantified by optical intensity ratio (OIR) and OIR variation. Baseline visual acuity letter score (VALS), baseline OCT biomarkers, and month 1 OIR were correlated with VALS at month 6. Regression trees, a machine learning technique yielding readily interpretable models, were used to assess for variable interaction. RESULTS:Only baseline VALS correlated positively with month 6 VALS in multivariate regression. Regression trees detected a novel functional and anatomical interaction in a subgroup. Among patients with a baseline VALS worse than 43, those with an OIR variation at month 1 of more than 0.09 had a mean of 13 fewer letters of vision at 6 months compared with patients with an OIR variation of 0.09 or less. CONCLUSIONS:Baseline VALS was the strongest predictor of month 6 VALS. Regression tree analysis detected an interaction effect, in which higher OIR variation at month 1 predicted worse 6-month VALS in patients with low VALS at baseline. OIR variation may serve as a predictor for poor visual outcome despite treatment of macular edema secondary to retinal vein occlusion in patients with poor vision at baseline. TRANSLATIONAL RELEVANCE:Pixel heterogeneity in three-dimensional OCT data may serve as measure of disruption of the retinal laminations, and this factor may carry visually prognostic value.

PURPOSE OF REVIEW/OBJECTIVE:In this review, we focus on artificial intelligence (AI) algorithms for diabetic retinopathy (DR) screening and risk stratification and factors to consider when implementing AI algorithms in the clinic. RECENT FINDINGS/RESULTS:AI algorithms have been adopted, and have received regulatory approval, for automated detection of referable DR with clinically acceptable diagnostic performance. While these metrics are an important first step, performance metrics that go beyond measures of technical accuracy are needed to fully evaluate the impact of AI algorithm on patient outcomes. Recent advances in AI present an exciting opportunity to improve patient care. Using DR as an example, we have reviewed factors to consider in the implementation of AI algorithms in real-world clinical practice. These include real-world evaluation of safety, efficacy, and equity (bias); impact on patient outcomes; ethical, logistical, and regulatory factors.

PURPOSE/UNASSIGNED:To describe delayed detection of pericentral hydroxychloroquine (HCQ) toxicity. METHODS/UNASSIGNED:67-year-old Dominican woman with rheumatoid arthritis on HCQ presented for examination. RESULTS/UNASSIGNED:Spectral-domain optical coherence tomography (SD-OCT) demonstrated bilateral cystoid macular edema with parafoveal attenuation of the external limiting membrane (ELM) and the ellipsoid zone (EZ). ELM and EZ disruption was present in inferior macula. While subtle superior defects were present on 10-2 visual fields, superior pericentral defects were noted on 24-2 testing. Hyperautofluorescence along inferior arcades corresponded to SD-OCT and visual fields. Examination 2 years prior demonstrated nonspecific points of depression on 10-2 visual fields and normal central SD-OCT findings. EZ and ELM disruption was present in the perifoveal inferior macula. CONCLUSIONS/UNASSIGNED:Early pericentral distribution of HCQ toxicity is not limited to Asian patients. Detecting pericentral HCQ toxicity involves reviewing entire macular cube on OCT. When OCT changes are suspected on parafoveal OCT B-scans, visual field testing with 24-2 may be more sensitive than 10-2.

PURPOSE/OBJECTIVE:To describe the vascular anatomy and intraluminal flow characteristics of segmental retinal arteritis (SRA) using structural and angiographic optical coherence tomography (OCT). METHODS:Retrospective case series of consecutive patients presenting with SRA. All patients were evaluated at presentation with fundus photography, spectral domain OCT, and OCT angiography. One patient was imaged with dense B-scan OCT angiography. RESULTS:Three eyes of three male patients were evaluated. All examinations were consistent with reactivation of ocular toxoplasmosis with an area of active retinochoroiditis adjacent to a focal chorioretinal scar. Spectral domain OCT through areas of SRA noted on clinical examination demonstrated areas of hyperreflectivity circumscribing the affected vessel with a normoreflective lumen. Optical coherence tomography angiography and dense B-scan OCT angiography demonstrated narrowing of the intraluminal flow signal that correlated with areas of segmental hyperreflectivity on spectral domain OCT. Vascular sections proximal and distal to areas of SRA showed normal flow signal. CONCLUSION/CONCLUSIONS:Vessels with SRA demonstrated hyperreflectivity highlighting the vessel wall on spectral domain OCT. Optical coherence tomography angiography showed narrowing of the flow signal within these segments suggesting reduced lumen diameter. Coupling these finding with previous indocyanine green imaging findings in SRA, the collective data suggest the plaques are localized within the vessel wall to either the endothelium or the muscular tunica media without occlusion of the vessel lumen.