Harnessing the power of longitudinal medical imaging for eye disease prognosis using Transformer-based sequence modeling.

Deep learning has enabled breakthroughs in automated diagnosis from medical imaging, with many successful applications in ophthalmology. However, standard medical image classification approaches only assess disease presence at the time of acquisition, neglecting the common clinical setting of longitudinal imaging. For slow, progressive eye diseases like age-related macular degeneration (AMD) and primary open-angle glaucoma (POAG), patients undergo repeated imaging over time to track disease progression and forecasting the future risk of developing a disease is critical to properly plan treatment. Our proposed Longitudinal Transformer for Survival Analysis (LTSA) enables dynamic disease prognosis from longitudinal medical imaging, modeling the time to disease from sequences of fundus photography images captured over long, irregular time periods. Using longitudinal imaging data from the Age-Related Eye Disease Study (AREDS) and Ocular Hypertension Treatment Study (OHTS), LTSA significantly outperformed a single-image baseline in 19/20 head-to-head comparisons on late AMD prognosis and 18/20 comparisons on POAG prognosis. A temporal attention analysis also suggested that, while the most recent image is typically the most influential, prior imaging still provides additional prognostic value.

Implementation of a Teleophthalmology Screening Program for Diabetic Retinopathy in New York City.

Purpose: To examine the implementation of a teleophthalmology program for diabetic retinopathy (DR) screening at a metropolitan hospital system and identify the challenges that the clinical teams encountered using the program. Methods: The study was conducted in 2 parts. The first was a pilot retrospective chart review of 300 consecutive patients screened for DR by the teleophthalmology screening program. The baseline variables, DR capture rate and staging, and continuity of care for those diagnosed with DR were analyzed. The second was a web-based survey identifying the barriers encountered by 36 physicians and clinical staff as they participated in the teleophthalmology screening program. Results: Part 1: Of the patients evaluated, 57 (19.0%) were diagnosed with DR; 42 (73.7%) had mild nonproliferative DR (NPDR), 7 (12.3%) had moderate NPDR, none had severe NPDR, and 8 (14.0%) had PDR. Thirty-one patients (54.4%) with retinopathy diagnoses were referred for an in-person follow-up at the clinic while the rest continued monitoring via the program. Of this subset, 22 (71.0%) completed the follow-up visit. Part 2: The survey respondents comprised 28 physicians (77.8%), 6 licensed nurse practitioners (16.7%), and 2 medical assistants (5.6%). Twenty-two providers (71.0%) preferred initiating referrals for in-person annual examinations over teleophthalmology screening referrals. The most common barriers described were related to workflow interruption, time constraints, and staff shortages. Conclusions: The teleophthalmology DR screening program allowed identification of early or absent DR at clinics in an urban setting (New York City). The findings suggest areas for targeted improvement in the screening program to better complement internal referral practices' workflows.

A Proposal for the Use of a Fixed Low-Energy Selective Laser Trabeculoplasty for Open Angle Glaucoma.

Selective laser trabeculoplasty (SLT) has been in routine clinical use for over 20 years with millions of patients successfully treated and a low rate of clinically significant complications. The procedure requires the clinician to manually position the laser beam on the trabecular meshwork using a gonioscopy lens and to titrate the SLT laser energy based on the amount of pigmentation in the angle, as well as the observation of small bubbles produced by the laser effect. We propose that SLT energy titration is unnecessary either to achieve intraocular pressure (IOP) reduction or to minimize potential side effects. Ample evidence to support our proposal includes multiple clinical reports demonstrating comparable levels of IOP reduction resulting from different laser energies, a large variety of energy and other laser parameters used in commercially available SLT lasers, and the nature of the laser-induced changes in the trabecular meshwork tissue with respect to energy. Despite these variations in laser parameters, SLT consistently reduces IOP with a low complication rate. We propose that using low fixed energy for all patients will effectively and safely lower patients' IOP while reducing the complexity of the SLT procedure, potentially making SLT accessible to more patients.

Improving model fairness in image-based computer-aided diagnosis.

Deep learning has become a popular tool for computer-aided diagnosis using medical images, sometimes matching or exceeding the performance of clinicians. However, these models can also reflect and amplify human bias, potentially resulting inaccurate missed diagnoses. Despite this concern, the problem of improving model fairness in medical image classification by deep learning has yet to be fully studied. To address this issue, we propose an algorithm that leverages the marginal pairwise equal opportunity to reduce bias in medical image classification. Our evaluations across four tasks using four independent large-scale cohorts demonstrate that our proposed algorithm not only improves fairness in individual and intersectional subgroups but also maintains overall performance. Specifically, the relative change in pairwise fairness difference between our proposed model and the baseline model was reduced by over 35%, while the relative change in AUC value was typically within 1%. By reducing the bias generated by deep learning models, our proposed approach can potentially alleviate concerns about the fairness and reliability of image-based computer-aided diagnosis.

NOVEL MFRP MUTATION WITH NANOPHTHALMOS, OPTIC DISK DRUSEN, AND PERIPHERAL RETINOSCHISIS IMAGED WITH ULTRA-WIDEFIELD OPTICAL COHERENCE TOMOGRAPHY.

PURPOSE: To describe with multimodal imaging including the use of ultra-widefield optical coherence tomography imaging a distinct phenotype of autosomal recessive nanophthalmos associated with a novel mutation of the MFRP gene (membrane-type frizzled-related protein). METHODS: Case report of a single patient followed by the Weill Cornell Medicine Department of Ophthalmology Retina and Glaucoma Services, and review of the relevant literature. RESULTS: A patient with a novel homozygous mutation in the MFRP gene (c.472C>T) presented with nanophthalmos, optic disk drusen, foveal hypoplasia, and extensive peripheral retinoschisis, which was revealed to be multilevel retinoschisis on ultra-widefield optical coherence tomography. Unlike other reported cases, the findings associated with this novel mutation did not include foveoschisis or clinically obvious retinitis pigmentosa. The patient underwent prophylactic peripheral laser iridotomy in both eyes. CONCLUSION: Here, we present a patient with nanophthalmos, optic disk drusen, and foveal hypoplasia associated with extensive peripheral retinoschisis imaged by ultra-widefield optical coherence tomography, but not foveal retinoschisis or prominent retinitis pigmentosa. The findings may expand the clinical spectrum of MFRP -associated nanophthalmos.

Primary Open-Angle Glaucoma Diagnosis from Optic Disc Photographs Using a Siamese Network.

Purpose: Primary open-angle glaucoma (POAG) is one of the leading causes of irreversible blindness in the United States and worldwide. Although deep learning methods have been proposed to diagnose POAG, these methods all used a single image as input. Contrastingly, glaucoma specialists typically compare the follow-up image with the baseline image to diagnose incident glaucoma. To simulate this process, we proposed a Siamese neural network, POAGNet, to detect POAG from optic disc photographs. Design: The POAGNet, an algorithm for glaucoma diagnosis, is developed using optic disc photographs. Participants: The POAGNet was trained and evaluated on 2 data sets: (1) 37 339 optic disc photographs from 1636 Ocular Hypertension Treatment Study (OHTS) participants and (2) 3684 optic disc photographs from the Sequential fundus Images for Glaucoma (SIG) data set. Gold standard labels were obtained using reading center grades. Methods: We proposed a Siamese network model, POAGNet, to simulate the clinical process of identifying POAG from optic disc photographs. The POAGNet consists of 2 side outputs for deep supervision and uses convolution to measure the similarity between 2 networks. Main Outcome Measures: The main outcome measures are the area under the receiver operating characteristic curve, accuracy, sensitivity, and specificity. Results: In POAG diagnosis, extensive experiments show that POAGNet performed better than the best state-of-the-art model on the OHTS test set (area under the curve [AUC] 0.9587 versus 0.8750). It also outperformed the baseline models on the SIG test set (AUC 0.7518 versus 0.6434). To assess the transferability of POAGNet, we also validated the impact of cross-data set variability on our model. The model trained on OHTS achieved an AUC of 0.7490 on SIG, comparable to the previous model trained on the same data set. When using the combination of SIG and OHTS for training, our model achieved superior AUC to the single-data model (AUC 0.8165 versus 0.7518). These demonstrate the relative generalizability of POAGNet. Conclusions: By simulating the clinical grading process, POAGNet demonstrated high accuracy in POAG diagnosis. These results highlight the potential of deep learning to assist and enhance clinical POAG diagnosis. The POAGNet is publicly available on https://github.com/bionlplab/poagnet.

Effective Low-Cost Ophthalmological Screening With a Novel iPhone Fundus Camera at Community Centers.

Ophthalmologic care is inaccessible to many people due to a variety of factors, including the availability of providers, cost of equipment for ophthalmologic care, and transportation to clinics and appointments. Because many causes of blindness are both highly prevalent and preventable once identified, it is essential to address gaps in care for underserved populations. We developed a novel 3D-printed mobile retinal camera. In this study, we organized recurring student-run screening events around New York City that took place in community centers and churches, at which we utilized our device to take retinal images. Our screening events reached a diverse population of New Yorkers, disproportionately those with lower household income, many of whom had not had recent eye exams. To validate the device for use in telehealth ophthalmologic visits, we transmitted the images to a remote ophthalmologist for evaluation and compared the result with an on-site attending physician's dilated eye exam. The subjective assessment indicated that 97% of images captured with the mobile retinal camera were acceptable for telehealth analysis. Remote image assessment by achieved 92% sensitivity and 83% specificity in detecting optic disc cupping, compared to the gold-standard on-site dilated eye exam. In addition, the device was portable, affordable, and able to be used by those with relatively little ophthalmologic training. We have demonstrated the utility of this affordable mobile retinal camera for telehealth ophthalmologic evaluation during community screening events that reached an underserved population to detect disease and connect with long-term care.

Automated diagnosing primary open-angle glaucoma from fundus image by simulating human's grading with deep learning.

Primary open-angle glaucoma (POAG) is a leading cause of irreversible blindness worldwide. Although deep learning methods have been proposed to diagnose POAG, it remains challenging to develop a robust and explainable algorithm to automatically facilitate the downstream diagnostic tasks. In this study, we present an automated classification algorithm, GlaucomaNet, to identify POAG using variable fundus photographs from different populations and settings. GlaucomaNet consists of two convolutional neural networks to simulate the human grading process: learning the discriminative features and fusing the features for grading. We evaluated GlaucomaNet on two datasets: Ocular Hypertension Treatment Study (OHTS) participants and the Large-scale Attention-based Glaucoma (LAG) dataset. GlaucomaNet achieved the highest AUC of 0.904 and 0.997 for POAG diagnosis on OHTS and LAG datasets. An ensemble of network architectures further improved diagnostic accuracy. By simulating the human grading process, GlaucomaNet demonstrated high accuracy with increased transparency in POAG diagnosis (comprehensiveness scores of 97% and 36%). These methods also address two well-known challenges in the field: the need for increased image data diversity and relying heavily on perimetry for POAG diagnosis. These results highlight the potential of deep learning to assist and enhance clinical POAG diagnosis. GlaucomaNet is publicly available on https://github.com/bionlplab/GlaucomaNet .

Clinical Outcomes of Ahmed ClearPath Implantation in Glaucomatous Eyes: A Novel Valveless Glaucoma Drainage Device.

PRCIS: Designed with novel features to facilitate implantation and improve safety, the Ahmed ClearPath (ACP) glaucoma drainage device (GDD) provided intraocular pressure (IOP) reduction comparable with other GDDs in eyes with refractory glaucoma in a multicenter retrospective study. PURPOSE: To present clinical outcomes with a novel valveless GDD (ACP, New World Medical). The 250 and 350 mm2 models feature a contoured plate for optimal globe apposition, anteriorized suture points to facilitate suturing to the globe, and a prethreaded 4-0 polypropylene ripcord suture. METHODS: This was a multicenter retrospective analysis of eyes with medically and/or surgically uncontrolled glaucoma implanted with the 250 or 350 mm2 ACP either as a standalone procedure or in combination with other procedures. Pre-, intra-, and postoperative data through 6 months were collected. RESULTS: A total of 104 eyes (100 subjects) received the ACP by 10 US surgeons, 63.5% of which had primary open-angle glaucoma and 62.5% had severe glaucoma. Mean baseline IOP was 26.3 (9.0) mm Hg and mean medication use was 3.9 (1.3). Through 6 months' follow-up, mean IOP ranged from 13.6 to 16.7 mm Hg and mean medication use from 0.9 to 1.9 medications (P<0.0001 at all timepoints for each outcome measure). At 6 months, mean IOP was 13.7 mm Hg (-13.0 mm Hg, 43.0%, P<0.0001) and mean medication use was 1.9 medications per eye (-2.1, 47.7%, P<0.0001). Common adverse events included anterior chamber inflammation (16.3%), hyphema (15.4%), and hypotony (6.7%). CONCLUSION: The new ACP appears to be safe and efficacious as a standalone procedure or in combination with other procedures for uncontrolled glaucoma, and may be considered as a GDD option for patients in whom its unique design may facilitate the implantation process.

The 100 Most Mentioned Glaucoma Articles Online With Highest Altmetric Attention Scores.

PRCIS: Characteristics of the most mentioned glaucoma articles on the internet were analyzed, allowing a better understanding of the dissemination of glaucoma research to the general public. PURPOSE: The aim was to determine the 100 most mentioned articles on the internet in the field of glaucoma and analyze their characteristics. MATERIALS AND METHODS: We identified the top 100 glaucoma articles with the highest Altmetric Attention Score (AAS), an automatically calculated metric for monitoring social media. Each article was evaluated for several characteristics including year of publication, title, journal name, journal impact factor (IF), article topic, article type, affiliation, and online mentions (news, blog, policy, Twitter, Facebook, etc.). Correlation analysis was conducted for AAS with these characteristics. RESULTS: The selected 100 articles came from 44 journals with more than half (56%) published in ophthalmology-specific journals. There was no significant correlation between IF and number of articles in a specific journal or AAS (P>0.1), but the number of articles in the top 100 was higher for ophthalmology journals with a higher IF (P<0.05). Original study was the most common study type (87%), of which clinical observation study was the most common subgroup (40%). Epidemiology/risk factor and basic science were the most common article topics (each 24%), followed by medical treatment (13%). Article topics regarding medical treatment had a significantly greater AAS than other topics (P<0.05). Of the top 5 articles, more than half (60%) were related to "Lifestyle choice" topics. CONCLUSIONS: There was no association between journal IF and AAS, consistent with previous studies. 90% of journals that had articles in the top 100 had a Twitter page. "Lifestyle choice" activities and other modifiable risk factors attracted significant online attention regarding glaucoma studies, with two of the top three most mentioned articles related to dietary intake. The present study thus provides a better understanding of online engagement with glaucoma research and the dissemination of this research to the general public.

Multi-scale Multi-structure Siamese Network (MMSNet) for Primary Open-Angle Glaucoma Prediction.

Primary open-angle glaucoma (POAG) is one of the leading causes of irreversible blindness in the United States and worldwide. POAG prediction before onset plays an important role in early treatment. Although deep learning methods have been proposed to predict POAG, these methods mainly focus on current status prediction. In addition, all these methods used a single image as input. On the other hand, glaucoma specialists determine a glaucomatous eye by comparing the follow-up optic nerve image with the baseline along with supplementary clinical data. To simulate this process, we proposed a Multi-scale Multi-structure Siamese Network (MMSNet) to predict future POAG event from fundus photographs. The MMSNet consists of two side-outputs for deep supervision and 2D blocks to utilize two-dimensional features to assist classification. The MMSNet network was trained and evaluated on a large dataset: 37,339 fundus photographs from 1,636 Ocular Hypertension Treatment Study (OHTS) participants. Extensive experiments show that MMSNet outperforms the state-of-the-art on two "POAG prediction before onset" tasks. Our AUC are 0.9312 and 0.9507, which are 0.2204 and 0.1490 higher than the state-of-the-art, respectively. In addition, an ablation study is performed to check the contribution of different components. These results highlight the potential of deep learning to assist and enhance the prediction of future POAG event. The proposed network will be publicly available on https://github.com/bionlplab/MMSNet.

Glaucoma Prevalence and Glaucoma Surgical Considerations in Prosthetic Replacement of the Ocular Surface Ecosystem Device Use.

OBJECTIVES: The purpose of this study was to identify patients with glaucoma and corneal/ocular surface disease who have been fit with Prosthetic Replacement of the Ocular Surface Ecosystem (PROSE; BostonSight, Needham, MA) devices and to understand the PROSE device fit experience in patients with glaucoma at Weill Cornell Medicine (WCM). METHODS: This is a retrospective chart review of patients older than 18 years who underwent PROSE device fitting at WCM. Records were reviewed for demographic information and ophthalmic variables. Descriptive statistics were performed. RESULTS: Between 2011 and 2017, 281 patients underwent PROSE device fitting at WCM, of whom 24 patients (8.5%) had a glaucoma diagnosis and 17 patients (6.0%) were identified as glaucoma suspect. Ocular surface disease was the most common indication for PROSE device wear (58.3%). Five patients had a history of incisional glaucoma surgery in the eye undergoing PROSE device fitting: three valved glaucoma drainage implants and two trabeculectomies. Three of these eyes were successfully fit with PROSE devices. CONCLUSIONS: Patients fitted with Prosthetic Replacement of the Ocular Surface Ecosystem devices at WCM had a higher prevalence of glaucoma than in the general population. Among patients with incisional glaucoma surgery, PROSE device fitting was challenging, with a 40% failure rate.

Adverse events associated with femtosecond laser-assisted cataract surgery reported to the MAUDE database.

PURPOSE: To report real-world complications associated with femtosecond laser-assisted cataract surgery (FLACS). SETTING: Deidentified database. DESIGN: Retrospective review. METHODS: The U.S. Food and Drug Administration's Manufacturer and User Facility Device Experience (MAUDE) database was culled for adverse events involving ophthalmic femtosecond laser during cataract surgery from January 2010 to January 2020. Reports in which the narrative description was inadequate to determine the nature of the complication were excluded. Each report was reviewed for complications, which were subsequently assigned to a complication category as determined by these authors. The number of events per year was determined and compared. RESULTS: Between January 2010 and January 2020, the MAUDE database search yielded 2927 reports involving femtosecond laser, of which 1927 met inclusion criteria and were included for analysis. These reports revealed 2704 complications: 1115 capsule tears (41.2%), 329 of which required vitrectomy, 372 instances of suction loss, 321 other capsulotomy-related complications, and 432 corneal or incisional complications. Rare complications included 54 reports of miosis, 16 reports of endophthalmitis, and 4 reports of toxic anterior segment syndrome. The number of events in the MAUDE database increased year over year between 2011 and 2017, peaking in 2017 at 303 events. CONCLUSIONS: By drawing on real-world data in the MAUDE database, this study identified the complications of greatest importance in clinical practice. Surgeons should be aware of the spectrum of possible complications to best prepare for all potential outcomes and to effectively counsel patients.

Case Report: Glaucoma in an Infant With Retinopathy of Prematurity.

Retinopathy of prematurity (ROP) is a leading cause of childhood blindness that occurs due to incomplete development of retinal blood vessels in preterm infants. Glaucoma is an ocular comorbidity in some patients with ROP, and it may be associated with immature anterior chamber development, ROP itself, or the treatment for ROP. There have been a few reports of narrow-angle glaucoma after laser treatment for ROP. In this case report, we describe the course of a female infant born at 24 weeks and 5 days of gestational age with treatment-requiring ROP treated with laser photocoagulation who subsequently developed very elevated intraocular pressure and shallow anterior chambers without pupillary block. The patient required bilateral ab externo trabeculotomy for elevated intraocular pressure, which normalized after the procedure. The patient has remained stable at the last follow-up at 51 weeks postmenstrual age. Differing from previous glaucoma presentations in this setting, we illustrate a case of elevated intraocular pressure and anterior chamber narrowing after laser therapy without pupillary block or synechiae. The possible multifactorial etiology of glaucoma in this patient, including incomplete angle development, ischemia, and laser treatment, highlight the need for glaucoma screening in patients with ROP, both in the short and long term.

Overnight Safety Evaluation of a Multi-Pressure Dial in Eyes with Glaucoma: Prospective, Open-Label, Randomized Study.

PURPOSE: To investigate the safety and tolerability of the multi-pressure dial (MPD) worn overnight for seven consecutive days. DESIGN: Prospective, open-label, randomized, single-site study. SUBJECTS: Twenty eyes of 10 subjects with open-angle glaucoma were fitted with an MPD and randomized to negative pressure application of -10 mmHg in one eye (study eye) worn overnight for 7 consecutive days. METHODS: Safety measures included best-corrected visual acuity (BCVA), intraocular pressure (IOP) changes from baseline during and after negative pressure application, slit lamp and dilated fundus exam findings, and rate of adverse events. Subjective assessments were completed daily by the subjects during the 7-day study period and immediately following the study period. RESULTS: Prior to the 7-day testing period, application of 10 mmHg negative pressure reduced mean IOP from 18.2 ± 3.8 mmHg to 14.0 ± 2.1 mmHg (p<0.01), a 22% reduction. After 7 days of consecutive nightly wear, repeat IOP measurements with the application of negative pressure showed a decrease in mean IOP from 16.9 ± 4.3 mmHg to 13.5 ± 3.7 mmHg. The observed IOP reduction was in addition to the subjects' current treatment regimen. There were no statistically significant changes in IOP, BCVA from baseline following the 7-day period of nightly wear with the application of negative pressure. The patient-reported outcomes were favorable. CONCLUSION: The MPD can safely and comfortably be worn overnight. The decrease in IOP of >20% in addition to current therapy is both clinically and statistically significant. The MPD shows promise as a potential new treatment option for nocturnal IOP control.