Clinical Characterization and Outcomes of Culture- and Polymerase Chain Reaction-Negative Cases of Infectious Keratitis.

PURPOSE: To examine the clinical presentation, management, and outcomes of culture and polymerase chain reaction (PCR) negative cases of infectious keratitis. METHODS: In this retrospective case series, we evaluated the laboratory and medical records of culture- and PCR-negative cases (2016-2020) reported to a tertiary care center, which were presumed to be infectious keratitis on the basis of clinical history and presentation. RESULTS: A total of 121 cases with culture-negative keratitis were included in this study. The mean age of the patients was 48.42 ± 1.89 years, and 53.72% were female. At presentation, the presumed etiology was viral in 38.01%, bacterial in 27.27%, fungal in 8.26%, Acanthamoeba in 6.61%, and unlisted in 28.92% of cases. The most common risk factors were a previous history of ocular surface diseases (96.69%) and contact lens use (37.19%). In total, 61.98% of the patients were already on antimicrobial medication at presentation. The initial management was altered in 79 cases (65.29%) during the treatment course. Average presenting and final (post-treatment) visual acuities (VA) were 0.98 ± 0.04 (LogMAR) and 0.42 ± 0.03 (LogMAR), respectively. A significantly higher frequency of patients with a final VA worse than 20/40 (Snellen) had worse VA at initial presentation (p < 0.0001). A history of ocular surface disease, cold sores, and recurrent infection (p < 0.05) were more commonly associated with a presumed diagnosis of viral keratitis. The patients with presumed bacterial etiology were younger and had a history of poor contact lens hygiene (p < 0.05). CONCLUSIONS: We observed a distinct difference in clinical features among patients with culture-negative and PCR-negative keratitis managed for presumed viral and bacterial infections. Although there was significant variability in presentation and management duration in this cohort, the visual outcomes were generally favorable.

AI-Human Hybrid Workflow Enhances Teleophthalmology for the Detection of Diabetic Retinopathy.

Objective: Detection of diabetic retinopathy (DR) outside of specialized eye care settings is an important means of access to vision-preserving health maintenance. Remote interpretation of fundus photographs acquired in a primary care or other nonophthalmic setting in a store-and-forward manner is a predominant paradigm of teleophthalmology screening programs. Artificial intelligence (AI)-based image interpretation offers an alternative means of DR detection. IDx-DR (Digital Diagnostics Inc) is a Food and Drug Administration-authorized autonomous testing device for DR. We evaluated the diagnostic performance of IDx-DR compared with human-based teleophthalmology over 2 and a half years. Additionally, we evaluated an AI-human hybrid workflow that combines AI-system evaluation with human expert-based assessment for referable cases. Design: Prospective cohort study and retrospective analysis. Participants: Diabetic patients ≥ 18 years old without a prior DR diagnosis or DR examination in the past year presenting for routine DR screening in a primary care clinic. Methods: Macula-centered and optic nerve-centered fundus photographs were evaluated by an AI algorithm followed by consensus-based overreading by retina specialists at the Stanford Ophthalmic Reading Center. Detection of more-than-mild diabetic retinopathy (MTMDR) was compared with in-person examination by a retina specialist. Main Outcome Measures: Sensitivity, specificity, accuracy, positive predictive value, and gradability achieved by the AI algorithm and retina specialists. Results: The AI algorithm had higher sensitivity (95.5% sensitivity; 95% confidence interval [CI], 86.7%-100%) but lower specificity (60.3% specificity; 95% CI, 47.7%-72.9%) for detection of MTMDR compared with remote image interpretation by retina specialists (69.5% sensitivity; 95% CI, 50.7%-88.3%; 96.9% specificity; 95% CI, 93.5%-100%). Gradability of encounters was also lower for the AI algorithm (62.5%) compared with retina specialists (93.1%). A 2-step AI-human hybrid workflow in which the AI algorithm initially rendered an assessment followed by overread by a retina specialist of MTMDR-positive encounters resulted in a sensitivity of 95.5% (95% CI, 86.7%-100%) and a specificity of 98.2% (95% CI, 94.6%-100%). Similarly, a 2-step overread by retina specialists of AI-ungradable encounters improved gradability from 63.5% to 95.6% of encounters. Conclusions: Implementation of an AI-human hybrid teleophthalmology workflow may both decrease reliance on human specialist effort and improve diagnostic accuracy. Financial Disclosures: Proprietary or commercial disclosure may be found after the references.

Comparison of machine learning-based algorithms using corneal asymmetry vs. single-metric parameters for keratoconus detection.

PURPOSE: To evaluate the diagnostic performance of three different parameter sets relevant to corneal asymmetry in comparison to conventional parameters including maximum anterior corneal curvature (Kmax) and thinnest corneal thickness for diagnosis of keratoconus. METHODS: In this retrospective case control study, 290 eyes with keratoconus and 847 eyes of normal patients were included in the analyses. Corneal tomography data were acquired from Scheimpflug tomography. The sklearn and FastAI libraries were used in a Python 3 environment to create all machine learning models. The original topography metrics and derived metrics together with the clinical diagnoses were used as the dataset for model training. The data were first split to assign 20% of the data to an isolated test set. The remaining data were then split 80/20 to a training and validation group for model training. Sensitivity and specificity outcomes with standard parameters (Kmax, central curvature, and thinnest pachymetry) and ratio of asymmetry across horizontal, apex centered, and flat axis-centered axis of reflection were studied via various machine learning models. RESULTS: Thinnest corneal pachymetry and Kmax were 549.8 ± 34.3 µm and 45.3 ± 1.7 D in normal eyes and 460.5 ± 62.6 µm and 59.3 ± 11.3 D in keratoconic eyes. Use of only corneal asymmetry ratios across all 4 meridians had mean sensitivity of 99.0% and mean specificity of 94.0%, better than utilizing Kmax alone or traditional measures combined (Kmax, thinnest cornea and inferior-superior asymmetry). CONCLUSIONS: By using the ratio of asymmetry between corneal axes alone, a machine learning model could identify patients with keratoconus in our dataset with satisfactory sensitivity and specificity. Further studies on pooled/larger datasets or more borderline population can help validate or refine these parameters.

Demonstration of Accuracy and Feasibility of Remotely Delivered Oximetry: A Blinded, Controlled, Real-World Study of Regional/Rural Children with Obstructive Sleep Apnoea.

Objectives: Evaluate diagnostic accuracy and feasibility of a mail-out home oximetry kit. Design: Patients were referred for both the tertiary/quaternary-centre hospital-delivered oximetry (HDO) and for the mail-out remotely-delivered oximetry (RDO). Quantitative and qualitative data were collected. The COVID-19 pandemic began during this study; therefore, necessary methodological adjustments were implemented. Setting: Patients were first evaluated in Swan Hill, Victoria. RDO kits were sent to home addresses. For the HDO, patients travelled to the Melbourne city area, received the kit, stayed overnight, and returned the kit the following morning. Participants: All consecutive paediatric patients (aged 2−18), diagnosed by a specialist in Swan Hill with obstructive sleep apnoea (OSA) on history/examination, and booked for tonsillectomy +/− adenoidectomy, were recruited. Main outcome measures: Diagnostic accuracy (i.e., comparison of RDO to HDO results) and test delivery time (i.e., days from consent signature to oximetry delivery) were recorded. Patient travel distances for HDO collection were calculated using home/delivery address postcodes and Google® Maps data. Qualitative data were collected with two digital follow-up surveys. Results: All 32 patients that had both the HDO and RDO had identical oximetry results. The HDO mean delivery time was 87.7 days, while the RDO mean delivery time was 23.6 days (p value: <0.001). Qualitatively, 3/28 preferred the HDO, while 25/28 preferred the RDO (n = 28). Conclusions: The remote option is as accurate as the hospital option, strongly preferred by patients, more rapidly completed, and also an ideal investigation delivery method during certain emergencies, such as the COVID-19 pandemic.

Non-visual complications of orbital cellulitis (NVCOC) in pediatric vs adult populations.

PURPOSE: There have been many publications evaluating the visual outcomes of patients treated for orbital cellulitis. The aim of this study was to evaluate non-visual complications of post-septal orbital cellulitis (NVCOC) present 30 days after discharge from index hospitalization. METHODS: This was an IRB-approved, retrospective chart review. RESULTS: Ninety patients (45 pediatric, 45 adults) were identified with OC. NVCOC were significantly more common in adult patients as compared to children (40.0% vs 15.6% respectively; p < 0.05). The most common NVCOC among children was persistent ptosis, while clinically significant scarring was most common in adults. NVCOC were less persistent in children as compared to adults with 71.4% of complications in children resolving spontaneously by 6 month follow up as compared to 11.1% adults. (p < 0.05). The only statistically significant risk factor identified for the development of NVCOC in children and adults was the presence of ICE (intracranial extension) during index hospitalization (p < 0.05) and the presence of an infected orbital implant (p < 0.05) respectively. CONCLUSIONS: Adult patients experience NVCOC complications more often than children. Furthermore, the nonvisual complications in children are more likely to resolve spontaneously than those in adults. The intracranial spread of infection is a significant risk factor for late complications in children. The presence of an orbital wall/floor implant is a significant risk factor for late complications in adults. There are many differences in the etiology, pathophysiology, and course of NVCOC in children and adults, so information on these two populations should be reported separately.

Fungal Keratitis: Clinical Features, Risk Factors, Treatment, and Outcomes.

Fungal keratitis (FK) can be challenging to diagnose and treat. In this retrospective case series, FK cases presenting at the University of Pittsburgh Medical Center, Pennsylvania, USA, from 2015 to 2021 were reviewed for ocular risk factors, clinical presentation, management, and outcomes. Twenty-eight cases of FK were included. The median presenting age was 58.5 (18.5) years, and the median symptom duration prior to presentation was 10 (35.8) days. Predisposing ocular risk factors included contact lens use (67.9%), recent ocular trauma/abrasion (42.9%), and history of ocular surgery (42.9%). The median presenting visual acuity (VA) was 1.35 (1.72) LogMAR. About half presented with a central ulcer (42.9%), large infiltrate (6.7 (6.3) mm2), corneal thinning (50.0%), and hypopyon (32.1%). The majority of isolated fungal species were filamentous (75.0%). Most common antifungal medications included topical voriconazole (71.4%) and natamycin (53.6%) drops and oral voriconazole (64.3%). Surgical management was necessary in 32.1% of cases and enucleation in one case. Defect resolution occurred in 42.5 (47.0) days, and median final VA was 0.5 (1.84) LogMAR. Features associated with poor final visual outcomes included poor initial VA (p < 0.001) and larger defect size (p = 0.002). In conclusion, unlike prior studies in the northeast region of the USA, FK was commonly caused by filamentous fungi, and antifungal management most often consisted of topical and oral voriconazole.

Predicting Systemic Health Features from Retinal Fundus Images Using Transfer-Learning-Based Artificial Intelligence Models.

While color fundus photos are used in routine clinical practice to diagnose ophthalmic conditions, evidence suggests that ocular imaging contains valuable information regarding the systemic health features of patients. These features can be identified through computer vision techniques including deep learning (DL) artificial intelligence (AI) models. We aim to construct a DL model that can predict systemic features from fundus images and to determine the optimal method of model construction for this task. Data were collected from a cohort of patients undergoing diabetic retinopathy screening between March 2020 and March 2021. Two models were created for each of 12 systemic health features based on the DenseNet201 architecture: one utilizing transfer learning with images from ImageNet and another from 35,126 fundus images. Here, 1277 fundus images were used to train the AI models. Area under the receiver operating characteristics curve (AUROC) scores were used to compare the model performance. Models utilizing the ImageNet transfer learning data were superior to those using retinal images for transfer learning (mean AUROC 0.78 vs. 0.65, p-value < 0.001). Models using ImageNet pretraining were able to predict systemic features including ethnicity (AUROC 0.93), age > 70 (AUROC 0.90), gender (AUROC 0.85), ACE inhibitor (AUROC 0.82), and ARB medication use (AUROC 0.78). We conclude that fundus images contain valuable information about the systemic characteristics of a patient. To optimize DL model performance, we recommend that even domain specific models consider using transfer learning from more generalized image sets to improve accuracy.

An Outperforming Artificial Intelligence Model to Identify Referable Blepharoptosis for General Practitioners.

The aim of this study is to develop an AI model that accurately identifies referable blepharoptosis automatically and to compare the AI model's performance to a group of non-ophthalmic physicians. In total, 1000 retrospective single-eye images from tertiary oculoplastic clinics were labeled by three oculoplastic surgeons as having either ptosis, including true and pseudoptosis, or a healthy eyelid. A convolutional neural network (CNN) was trained for binary classification. The same dataset was used in testing three non-ophthalmic physicians. The CNN model achieved a sensitivity of 92% and a specificity of 88%, compared with the non-ophthalmic physician group, which achieved a mean sensitivity of 72% and a mean specificity of 82.67%. The AI model showed better performance than the non-ophthalmic physician group in identifying referable blepharoptosis, including true and pseudoptosis, correctly. Therefore, artificial intelligence-aided tools have the potential to assist in the diagnosis and referral of blepharoptosis for general practitioners.

Microbiological evaluation of corneal and contact lens cultures in contact lens-associated bacterial keratitis.

PURPOSE: To determine the degree of microbiological agreement between corneal scrapings and contact lens cultures in cases of contact lens-associated keratitis. METHODS: Electronic medical records of all cases of contact lens-associated keratitis at a single institution from April 2006 to May 2019 were reviewed. Cases where both corneal scrapings and contact lens cultures were obtained were included in the study. Group 1 demonstrated agreement between corneal scrapings and contact lens cultures. Group 2 demonstrated growth on both cultures, but disagreement in isolated organism. Group 3 demonstrated negative corneal cultures but growth on contact lens cultures and the diagnostic yield of contact lens microbiological cultures and agreement between corneal and contact lens cultures. RESULTS: A total of 80 eyes of 72 patients were included in the study. 135 total incidences of microbiological results were included for data analysis. Group 1 contained nine incidences (6.7%), group 2 contained 60 incidences (44.4%) and group 3 contained 66 incidences (49%). In group 3, 50% of the cases were treated based on contact lens culture data. There was no statistically significant difference between all three groups in terms of baseline characteristics, presenting vision, vision at last follow-up, number of antibiotics used or complications from keratitis. Pseudomonas was the most common microorganism isolated, and in 89% of these cases, there was disagreement between corneal and contact lens cultures. CONCLUSIONS: Although there was a disagreement in the microbiological yield between contact lens and corneal cultures, contact lens cultures were useful in management of patients while achieving similar outcomes.

An 18-Year Overview of Serratia marcescens Ocular Infection.

PURPOSE: Serratia marcescens is a frequent ocular bacterial pathogen implicated in keratitis, endophthalmitis, and conjunctivitis. We evaluated the risk factors and treatment outcomes of ocular infections due to S. marcescens. METHODS: In this retrospective observational study, all S. marcescens-positive cases between February 2002 and February 2020 were reviewed for ocular risk factors that included log of minimal angle of resolution visual acuity (VA), medical management, and time to epithelial defect closure. RESULTS: Fifty-one patients were identified (72.5% females, 46.8±23.3 years). Forty-six patients had complete medical records, and 5 had microbiology data available. The most prevalent ocular risk factors were, contact lens (CL) use (68.6%), corneal disease (52.9%), and history of ocular surgery (41.2%). Mean presenting VA was 1.3±1.0. About half of the patients presented with a central ulcer (49%, 25), large infiltrate (20.4±31.8 mm2 mean), and hypopyon (43.1%, 22). All cases were reported to be susceptible to ciprofloxacin. Defect closure occurred in 52.3±117.1 days and final VA was 0.86±0.88. Adjunctive treatments were required in 14 cases (27.5%). One patient underwent surgical intervention. Features associated with poor VA outcomes included, history of glaucoma (P=0.038), older age at presentation (P<0.001), presence of hypopyon (0.045), poor VA at presentation (0.0086), time to epithelial defect closure (0.0196), and large infiltrate size (P=0.0345). CONCLUSIONS: S. marcescens keratitis and conjunctivitis is associated with CL use and history of ocular surface disease. Worse outcomes were associated with older age, infiltrate size, presence of hypopyon, worse initial VA, longer time to epithelial defect closure, and history of glaucoma.

Comparison of Clinical Features and Treatment Outcomes of Pseudomonas aeruginosa Keratitis in Contact Lens and Non-Contact Lens Wearers.

PURPOSE: To compare the outcomes of Pseudomonas aeruginosa keratitis (PAK) in contact lens wearers (CLWs) and non-contact lens wearers (non-CLWs) and identify risk factors for poor visual acuity (VA) outcomes in each group. DESIGN: Retrospective cohort study METHODS: Two hundred fourteen consecutive cases of PAK were included between January 2006 and December 2019. Clinical features, microbiologic results, and treatment course were compared between CLW and non-CLW groups. Analyses of clinical features predicting poor final VA were performed. RESULTS: This study identified 214 infected eyes in 207 patients with PAK, including 163 eyes (76.2%) in CLWs and 51 eyes (23.8%) in non-CLWs. The average age was 39.2 years in CLWs and 71.9 years in non-CLWs (P < .0001). The average logMAR visual acuity (VA) at presentation was 1.39 in CLWs and 2.17 in non-CLWs (P < .0001); average final VA was 0.76 in CLWs and 1.82 in non-CLWs (P < .0001). Stromal necrosis required a procedural or surgical intervention in 13.5% of CLWs and 49.0% of non-CLWs (P < .0001). A machine learning-based analysis yielded a list of clinical features that most strongly predict a poor VA outcome (worse than 20/40), including worse initial VA, older age, larger size of infiltrate or epithelial defect at presentation, and greater maximal depth of stromal necrosis. CONCLUSIONS: Non-CLWs have significantly worse VA outcomes and required a higher rate of surgical intervention, compared with CLWs. Our study elucidates risk factors for poor visual outcomes in non-CLWs with PAK.

A deep learning approach to identify blepharoptosis by convolutional neural networks.

PURPOSE: Blepharoptosis is a known cause of reversible vision loss. Accurate assessment can be difficult, especially amongst non-specialists. Existing automated techniques disrupt clinical workflow by requiring user input, or placement of reference markers. Neural networks are known to be effective in image classification tasks. We aim to develop an algorithm that can accurately identify blepharoptosis from a clinical photo. METHODS: A total of 500 clinical photographs from patients with and without blepharoptosis were sourced from a tertiary ophthalmic center in Taiwan. Images were labeled by two oculoplastic surgeons, with an independent third oculoplastic surgeon to adjudicate disagreements. These images were used to train a series of convolutional neural networks (CNNs) to ascertain the best CNN architecture for this particular task. RESULTS: Of the models that trained on the dataset, most were able to identify ptosis images with reasonable accuracy. We found the best performing model to use the DenseNet121 architecture without pre-training which achieved a sensitivity of 90.1 % with a specificity of 82.4 %, compared to the worst performing model which was used a Resnet34 architecture with pre-training, achieving a sensitivity of 74.1 %, and specificity of 63.6 %. Models with and without pre-training performed similarly (mean accuracy 82.6 % vs. 85.8 % respectively, p = 0.06), though models with pre-training took less time to train (1-minute vs. 16 min, p < 0.01). CONCLUSIONS: We report the use of AI to accurately diagnose blepharoptosis from a clinical photograph with no external reference markers or user input requirement. Most current-generation CNN architectures performed reasonably on this task, with the DenseNet121, and Resnet18 architectures without pre-training performing best in our dataset.

Automatic Identification of Referral-Warranted Diabetic Retinopathy Using Deep Learning on Mobile Phone Images.

Purpose: To evaluate the performance of a deep learning algorithm in the detection of referral-warranted diabetic retinopathy (RDR) on low-resolution fundus images acquired with a smartphone and indirect ophthalmoscope lens adapter. Methods: An automated deep learning algorithm trained on 92,364 traditional fundus camera images was tested on a dataset of smartphone fundus images from 103 eyes acquired from two previously published studies. Images were extracted from live video screenshots from fundus examinations using a commercially available lens adapter and exported as a screenshot from live video clips filmed at 1080p resolution. Each image was graded twice by a board-certified ophthalmologist and compared to the output of the algorithm, which classified each image as having RDR (moderate nonproliferative DR or worse) or no RDR. Results: In spite of the presence of multiple artifacts (lens glare, lens particulates/smudging, user hands over the objective lens) and low-resolution images achieved by users of various levels of medical training, the algorithm achieved a 0.89 (95% confidence interval [CI] 0.83-0.95) area under the curve with an 89% sensitivity (95% CI 81%-100%) and 83% specificity (95% CI 77%-89%) for detecting RDR on mobile phone acquired fundus photos. Conclusions: The fully data-driven artificial intelligence-based grading algorithm herein can be used to screen fundus photos taken from mobile devices and identify with high reliability which cases should be referred to an ophthalmologist for further evaluation and treatment. Translational Relevance: The implementation of this algorithm on a global basis could drastically reduce the rate of vision loss attributed to DR.

A fishy eyelid injury.

A 9-year-old girl was reviewed by a tertiary ophthalmology service after being hit in her right upper eyelid by a fish whilst swimming. Initial wound exploration demonstrated fish scales in the wound. She was first treated conservatively with washout of the 5-mm wound and was discharged with oral ciprofloxacin. Five days later, the patient re-presented with a worsened ptosis due to periorbital swelling. Ultrasound of the upper lid demonstrated a foreign body in the upper lid. The patient was taken to theatre and, with the assistance of intraoperative ultrasound, the jaws of a needle fish were removed from the upper lid. The case highlights the importance of ultrasound and its intraoperative utility in cases of trauma and a suspicion of retained foreign bodies as well as the potential danger of fish injuries off the West Australian coast.

A review of m-health in medical imaging.

OBJECTIVE: The increasing capabilities of camera-equipped mobile phones have led to a growing body of evidence regarding their use in medical imaging across a broad range of medical specialties. This article reviews the current evidence for the use of mobile health (m-health) in medical imaging. MATERIALS AND METHODS: We performed a structured review of the published literature regarding m-health in medical imaging using the Medline, PubMed, and Web of Science databases (January 2002-August 2013). The two authors independently extracted data regarding type of specialty, purpose, and study design of publications. RESULTS: In total, 235 articles were identified. The majority of studies were case reports or noncomparative product validation studies. The greatest volume of publications originated in the fields of radiology (21%), dermatology (15%), laboratory techniques (15%), and plastic surgery (12%). Among these studies, m-health was used as diagnostic aids, for patient monitoring, and to improve communication between health practitioners. With the growing use of mobile phones for medical imaging, considerations need to be given to informed consent, privacy, image storage and transfer, and guidelines for healthcare workers and patients. CONCLUSIONS: There are several novel uses of mobile devices for medical imaging that show promise across a variety of areas and subspecialties of healthcare. Currently, studies are mostly exploratory in nature. To validate these devices, studies with higher methodological rigor are required.

Lymphoma of the trigeminal nerve--the need for histological diagnosis.

CNS lymphoma involving the trigeminal nerve is a rare condition which presents as a cavernous sinus lesion. It may mimic the radiological appearance of other lesions, and biopsy is essential before considering empirical radiotherapy for lesions in this region. We report the radiological, histopathological and operative findings of a primary non Hodgkin B cell lymphoma involving the trigeminal nerve.