Faculty Bibliography

SIGNIFICANCE/CONCLUSIONS:Optimal meibography utilization and interpretation are hindered due to poor lid presentation, blurry images, or image artifacts and the challenges of applying clinical grading scales. These results, using the largest image dataset analyzed to date, demonstrate development of algorithms that provide standardized, real-time inference that addresses all of these limitations. PURPOSE/OBJECTIVE:This study aimed to develop and validate an algorithmic pipeline to automate and standardize meibomian gland absence assessment and interpretation. METHODS:A total of 143,476 images were collected from sites across North America. Ophthalmologist and optometrist experts established ground-truth image quality and quantification (i.e., degree of gland absence). Annotated images were allocated into training, validation, and test sets. Convolutional neural networks within Google Cloud VertexAI trained three locally deployable or edge-based predictive models: image quality detection, over-flip detection, and gland absence detection. The algorithms were combined into an algorithmic pipeline onboard a LipiScan Dynamic Meibomian Imager to provide real-time clinical inference for new images. Performance metrics were generated for each algorithm in the pipeline onboard the LipiScan from naive image test sets. RESULTS:Individual model performance metrics included the following: weighted average precision (image quality detection: 0.81, over-flip detection: 0.88, gland absence detection: 0.84), weighted average recall (image quality detection: 0.80, over-flip detection: 0.87, gland absence detection: 0.80), weighted average F1 score (image quality detection: 0.80, over-flip detection: 0.87, gland absence detection: 0.81), overall accuracy (image quality detection: 0.80, over-flip detection: 0.87, gland absence detection: 0.80), Cohen κ (image quality detection: 0.60, over-flip detection: 0.62, and gland absence detection: 0.71), Kendall τb (image quality detection: 0.61, p<0.001, over-flip detection: 0.63, p<0.001, and gland absence detection: 0.67, p<001), and Matthews coefficient (image quality detection: 0.61, over-flip detection: 0.63, and gland absence detection: 0.62). Area under the precision-recall curve (image quality detection: 0.87 over-flip detection: 0.92, gland absence detection: 0.89) and area under the receiver operating characteristic curve (image quality detection: 0.88, over-flip detection: 0.91 gland absence detection: 0.93) were calculated across a common set of thresholds, ranging from 0 to 1. CONCLUSIONS:Comparison of predictions from each model to expert panel ground-truth demonstrated strong association and moderate to substantial agreement. The findings and performance metrics show that the pipeline of algorithms provides standardized, real-time inference/prediction of meibomian gland absence.

BACKGROUND:Identification and referral of at-risk patients from primary care practitioners (PCPs) to eye care professionals remain a challenge. Approximately 1.9 million Americans suffer from vision loss as a result of undiagnosed or untreated ophthalmic conditions. In ophthalmology, artificial intelligence (AI) is used to predict glaucoma progression, recognize diabetic retinopathy (DR), and classify ocular tumors; however, AI has not yet been used to triage primary care patients for ophthalmology referral. OBJECTIVE:This study aimed to build and compare machine learning (ML) methods, applicable to electronic health records (EHRs) of PCPs, capable of triaging patients for referral to eye care specialists. METHODS:Accessing the Optum deidentified EHR data set, 743,039 patients with 5 leading vision conditions (age-related macular degeneration [AMD], visually significant cataract, DR, glaucoma, or ocular surface disease [OSD]) were exact-matched on age and gender to 743,039 controls without eye conditions. Between 142 and 182 non-ophthalmic parameters per patient were input into 5 ML methods: generalized linear model, L1-regularized logistic regression, random forest, Extreme Gradient Boosting (XGBoost), and J48 decision tree. Model performance was compared for each pathology to select the most predictive algorithm. The area under the curve (AUC) was assessed for all algorithms for each outcome. RESULTS:XGBoost demonstrated the best performance, showing, respectively, a prediction accuracy and an AUC of 78.6% (95% CI 78.3%-78.9%) and 0.878 for visually significant cataract, 77.4% (95% CI 76.7%-78.1%) and 0.858 for exudative AMD, 79.2% (95% CI 78.8%-79.6%) and 0.879 for nonexudative AMD, 72.2% (95% CI 69.9%-74.5%) and 0.803 for OSD requiring medication, 70.8% (95% CI 70.5%-71.1%) and 0.785 for glaucoma, 85.0% (95% CI 84.2%-85.8%) and 0.924 for type 1 nonproliferative diabetic retinopathy (NPDR), 82.2% (95% CI 80.4%-84.0%) and 0.911 for type 1 proliferative diabetic retinopathy (PDR), 81.3% (95% CI 81.0%-81.6%) and 0.891 for type 2 NPDR, and 82.1% (95% CI 81.3%-82.9%) and 0.900 for type 2 PDR. CONCLUSIONS:The 5 ML methods deployed were able to successfully identify patients with elevated odds ratios (ORs), thus capable of patient triage, for ocular pathology ranging from 2.4 (95% CI 2.4-2.5) for glaucoma to 5.7 (95% CI 5.0-6.4) for type 1 NPDR, with an average OR of 3.9. The application of these models could enable PCPs to better identify and triage patients at risk for treatable ophthalmic pathology. Early identification of patients with unrecognized sight-threatening conditions may lead to earlier treatment and a reduced economic burden. More importantly, such triage may improve patients' lives.

BACKGROUND AND OBJECTIVE/OBJECTIVE:To quantify fractal dimension (FD) by mass-radius method in optical coherence tomography angiography (OCTA) images and characterize microvascular differences in eyes with and without diabetic retinopathy (DR). PATIENTS AND METHODS/METHODS:A retrospective study was conducted using OCTA images of 3 mm × 3 mm and 6 mm × 6 mm scans for superficial and deep capillary plexuses from 49 control eyes and 58 eyes with DR. RESULTS:< .05). In the 3 mm × 3 mm superficial plexus, the FD of severe nonproliferative DR (NPDR) and proliferative DR (PDR) were significantly lower compared to control. The scans of the deep plexus showed only severe NPDR was significantly reduced in the 6 mm × 6 mm scan, whereas moderate NPDR, severe NPDR, and PDR were significantly lower in the 3 mm × 3 mm scan. CONCLUSION/CONCLUSIONS:.

PURPOSE/OBJECTIVE:To report a case of bilateral toxic corneal and conjunctival epitheliopathy secondary to administration of filter paper impregnated with lysergic acid diethylamide (LSD) in the inferior conjunctival fornices. METHODS:This is a single case report of an 18-year-old man who presented to the emergency department with acute, bilateral eye pain and redness of 24 hours. The patient admitted to placing folded strips of blotting paper impregnated with LSD into the inferior fornices of his eyes the previous night. RESULTS:The patient was found to have localized bilateral corneal and conjunctival abrasions with underlying subconjunctival hemorrhage. Conjunctival abrasion was "kissing," involving the bulbar and palpebral conjunctiva, corresponding to the presumed location of the filter paper. There was no corneal stromal opacification. He was lost to follow up within 1 week of initial presentation but stated that his symptoms improved. CONCLUSIONS:To the best of our knowledge, this is the first reported case of bilateral hemorrhagic conjunctival abrasion and corneal abrasion secondary to LSD. "Kissing" conjunctival lesions, which have been previously reported with heroin use, should raise suspicion for drug abuse.

PURPOSE: To use fractal dimensional analysis to investigate retinal vascular disease patterns in patients with diabetic retinopathy using spectral domain optical coherence tomography angiography. METHODS: A retrospective study was conducted which included 49 eyes from 26 control subjects and 58 eyes from 35 patients known to have diabetic retinopathy. Of the 58 eyes with known retinopathy, 31 were categorized as nonproliferative diabetic retinopathy (13 mild, 9 moderate, and 9 severe) and 27 were categorized as proliferative diabetic retinopathy. Optical coherence tomography angiography images were acquired using the RTVue XR Avanti (Optovue, Inc). Automated segmentation was obtained through both the superficial and deep capillary plexuses for each eye. Grayscale optical coherence tomography angiography images were standardized and binarized using ImageJ (National Institutes of Health). Fractal box-counting analyses were conducted using Fractalyse (TheMA). Fractal dimensions (FDs) and correlation coefficient of the superficial and deep capillary plexuses were compared between control eyes and those in various stages of diabetic retinopathy. RESULTS: The superficial and deep capillary plexuses from diabetic and control eyes were analyzed. The average FD for diabetic eyes was significantly lower than in control eyes in the superficial plexus (P = 2.4 x 10) and in the deep capillary plexus (P = 1.87 x 10 ) with a more statistically significant difference noted in the deep capillary plexus. When analyzing diabetic patients without edema noted on optical coherence tomography, the FD was significantly reduced in the superficial (P = 0.001) and deep (P = 1.49 x 10) plexuses. When analyzing diabetic patients with edema noted on optical coherence tomography, the FD was significantly reduced in the superficial (P = 2.0 x 10) and deep (P = 1.85 x 10) plexuses. CONCLUSION: The optical coherence tomography angiography FD is significantly lower in both superficial and deep capillary plexuses in eyes with all stages studied of diabetic retinopathy. The results were more often significant for the deep capillary plexus. The use of fractal analysis provides an objective criterion to assess microvascular disease burden in diabetic retinopathy.

Purpose/UNASSIGNED:To investigate whether preoperative corneal topographic and biomechanical parameters (CTBPs) predict postoperative residual refractive error (RRE). Methods/UNASSIGNED:We retrospectively included 151 eyes from 151 patients of small-incision lenticule extraction (SMILE) with target RRE of plano and 3-month measurements of refractive error from Tianjin Eye Hospital. Multivariate linear/logistic regressions were performed to associate age, gender, preoperative refractive error, lenticule thickness, and CTBPs with postoperative RRE/the occurrence of myopic RRE ≤ -0.25 diopter (D). Stepwise regression was used for feature selection. Leave-one-cross-validation was used for model evaluation by the area under the receiver operating characteristic curve (AUC). Results/UNASSIGNED:< 0.001) improved the AUC performance to 0.771 from 0.615. Conclusions/UNASSIGNED:Postoperative outcomes of SMILE can be predicted by individual CTBPs. Translational Relevance/UNASSIGNED:Our findings could be used to customize a refractive nomogram based on individual corneal properties improving outcomes and patient satisfaction.

Purpose: To assess whether microvascular dropout as measured by fractal self-similarity breakdown of optical coherence tomography angiography (OCTA) occurs in a sectoral fashion in eyes with primary open angle glaucoma (POAG) compared to control patients.
Method(s): A retrospective study using OCTA images obtained on 40 eyes with POAG, and 14 control eyes. OCTA images with peripapillary scans of 4.5mm x 4.5mm diameters were obtained using RTVue XR Avanti (Optovue Inc., Fremont, CA, USA), and standardized and binarized using ImageJ (National Institutes of Health, Bethesda, Maryland, USA). Fractal dimension by means of box-counting algorithm using box sizes with increasing exponential factor of two with grid algorithm on Fractalyse (TheMA, Besancon Cedex, France) was plotted against linear box dimension and first local peak (FLP) representing smallest box size resolution before breakdown of self-similarity was recorded. The ratio of superior to inferior (SI Ratio) of FLP was the main outcome measure.
Result(s): There is significant difference between control and POAG eyes (p = 0.01), with POAG having larger superior to inferior (SI Ratio) of FLP (1.40 +/-0.93) compared to control (1.00 +/-0.0). The increased SI FLP ratio of the POAG eyes reflects a preferential loss of self-similarity in the superior quadrant (i.e. higher FLP values) rather than increased self-similarity in the inferior quadrant (FLP inferior = 8.0 +/-0 and FLP superior = 11.2 +/-7.4 in POAG, FLP inferior = 8.0 +/-0 and FLP superior = 8.0 +/-0 in controls).
Conclusion(s): The measurement of box size of the first local maximum of fractal dimension as a function of increasing box size represents the smallest box size resolution prior to loss of uniformity of the vascular pattern's fractal dimension. While this loss of uniformity may be related to resolution in normal eyes, the preferential loss of microvascular complexity inferior to the optic disc in POAG eyes suggests asymmetric small vessel dropout in the POAG. This accords well with the asymmetric loss demonstrated in a prior OCTA study. Our study was not designed to determine whether microvascular loss occurred as a consequence or as a cause of axonal loss. We propose FLP SI Ratio as a useful measure of POAG-associated microvascular dropout