Fundus Photography vs. Ophthalmoscopy Outcomes in the Emergency Department (FOTO-ED) Phase III: Web-based, In-service Training of Emergency Providers

Application of a Deep Learning System to Detect Papilledema on Nonmydriatic Ocular Fundus Photographs in an Emergency Department

PURPOSE

The Fundus photography vs Ophthalmoscopy Trial Outcomes in the Emergency Department (FOTO-ED) studies showed that ED providers poorly recognized funduscopic findings in patients in the ED. We tested a modified version of the Brain and Optic Nerve Study Artificial Intelligence (BONSAI) deep learning system on nonmydriatic fundus photographs from the FOTO-ED studies to determine if the deep learning system could have improved the detection of papilledema had it been available to ED providers as a real-time diagnostic aid.

DESIGN

Retrospective secondary analysis of a cohort of patients included in the FOTO-ED studies.

METHODS

The testing data set included 1608 photographs obtained from 828 patients in the FOTO-ED studies. Photographs were reclassified according to the optic disc classification system used by the deep learning system ("normal optic discs," "papilledema," and "other optic disc abnormalities"). The system's performance was evaluated by calculating the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity using a 1-vs-rest strategy, with reference to expert neuro-ophthalmologists.

RESULTS

The BONSAI deep learning system successfully distinguished normal from abnormal optic discs (AUC 0.92 [95% confidence interval {CI} 0.90-0.93]; sensitivity 75.6% [73.7%-77.5%] and specificity 89.6% [86.3%-92.8%]), and papilledema from normal and others (AUC 0.97 [0.95-0.99]; sensitivity 84.0% [75.0%-92.6%] and specificity 98.9% [98.5%-99.4%]). Six patients with missed papilledema in 1 eye were correctly identified by the deep learning system as having papilledema in the other eye.

CONCLUSIONS

The BONSAI deep learning system was able to reliably identify papilledema and normal optic discs on nonmydriatic photographs obtained in the FOTO-ED studies. Our deep learning system has excellent potential as a diagnostic aid in EDs and non-ophthalmology clinics equipped with nonmydriatic fundus cameras. NOTE: Publication of this article is sponsored by the American Ophthalmological Society.

The BONSAI (Brain and Optic Nerve Study with Artificial Intelligence) deep learning system can accurately identify pediatric papilledema on standard ocular fundus photographs

BACKGROUND

Pediatric papilledema often reflects an underlying severe neurologic disorder and may be difficult to appreciate, especially in young children. Ocular fundus photographs are easy to obtain even in young children and in nonophthalmology settings. The aim of our study was to ascertain whether an improved deep-learning system (DLS), previously validated in adults, can accurately identify papilledema and other optic disk abnormalities in children.

METHODS

The DLS was tested on mydriatic fundus photographs obtained in a multiethnic pediatric population (<17 years) from three centers (Atlanta-USA; Bucharest-Romania; Singapore). The DLS's multiclass classification accuracy (ie, normal optic disk, papilledema, disks with other abnormality) was calculated, and the DLS's performance to specifically detect papilledema and normal disks was evaluated in a one-vs-rest strategy using the AUC, sensitivity and specificity, with reference to expert neuro-ophthalmologists.

RESULTS

External testing was performed on 898 fundus photographs: 447 patients; mean age, 10.33 (231 patients ≤10 years of age; 216, 11-16 years); 558 normal disks, 254 papilledema, 86 other disk abnormalities. Overall multiclass accuracy of the DLS was 89.6% (range, 87.8%-91.6%). The DLS successfully distinguished "normal" from "abnormal" optic disks (AUC 0.99 [0.98-0.99]; sensitivity, 87.3% [84.9%-89.8%]; specificity, 98.5% [97.6%-99.6%]), and "papilledema" from "normal and other" (AUC 0.99 [0.98-1.0]; sensitivity, 98.0% [96.8%-99.4%]; specificity, 94.1% (92.4%-95.9%)].

CONCLUSIONS

Our DLS reliably distinguished papilledema from normal optic disks and other disk abnormalities in children, suggesting it could be utilized as a diagnostic aid for the assessment of optic nerve head appearance in the pediatric age group.

The Current State of Artificial Intelligence in Neuro-Ophthalmology. A Review

This article presents a summary of recent advances in the development and use of complex systems using artificial intelligence (AI) in neuro-ophthalmology. The aim of the following article is to present the principles of AI and algorithms that are currently being used or are still in the stage of evaluation or validation within the neuro-ophthalmology environment. For the purpose of this text, a literature search was conducted using specific keywords in available scientific databases, cumulatively up to April 2023. The AI systems developed across neuro-ophthalmology mostly achieve high sensitivity, specificity and accuracy. Individual AI systems and algorithms are subsequently selected, simply described and compared in the article. The results of the individual studies differ significantly, depending on the chosen methodology, the set goals, the size of the test, evaluated set, and the evaluated parameters. It has been demonstrated that the evaluation of various diseases will be greatly speeded up with the help of AI and make the diagnosis more efficient in the future, thus showing a high potential to be a useful tool in clinical practice even with a significant increase in the number of patients.

Deep Learning System Outperforms Clinicians in Identifying Optic Disc Abnormalities

BACKGROUND

The examination of the optic nerve head (optic disc) is mandatory in patients with headache, hypertension, or any neurological symptoms, yet it is rarely or poorly performed in general clinics. We recently developed a brain and optic nerve study with artificial intelligence-deep learning system (BONSAI-DLS) capable of accurately detecting optic disc abnormalities including papilledema (swelling due to elevated intracranial pressure) on digital fundus photographs with a comparable classification performance to expert neuro-ophthalmologists, but its performance compared to first-line clinicians remains unknown.

METHODS

In this international, cross-sectional multicenter study, the DLS, trained on 14,341 fundus photographs, was tested on a retrospectively collected convenience sample of 800 photographs (400 normal optic discs, 201 papilledema and 199 other abnormalities) from 454 patients with a robust ground truth diagnosis provided by the referring expert neuro-ophthalmologists. The areas under the receiver-operating-characteristic curves were calculated for the BONSAI-DLS. Error rates, accuracy, sensitivity, and specificity of the algorithm were compared with those of 30 clinicians with or without ophthalmic training (6 general ophthalmologists, 6 optometrists, 6 neurologists, 6 internists, 6 emergency department [ED] physicians) who graded the same testing set of images.

RESULTS

With an error rate of 15.3%, the DLS outperformed all clinicians (average error rates 24.4%, 24.8%, 38.2%, 44.8%, 47.9% for general ophthalmologists, optometrists, neurologists, internists and ED physicians, respectively) in the overall classification of optic disc appearance. The DLS displayed significantly higher accuracies than 100%, 86.7% and 93.3% of clinicians (n = 30) for the classification of papilledema, normal, and other disc abnormalities, respectively.

CONCLUSIONS

The performance of the BONSAI-DLS to classify optic discs on fundus photographs was superior to that of clinicians with or without ophthalmic training. A trained DLS may offer valuable diagnostic aid to clinicians from various clinical settings for the screening of optic disc abnormalities harboring potentially sight- or life-threatening neurological conditions.

Diagnosis of Optic Disc Oedema: Fundus Features, Ocular Imaging Findings, and Artificial Intelligence

Optic disc swelling is a manifestation of a broad range of processes affecting the optic nerve head and/or the anterior segment of the optic nerve. Accurately diagnosing optic disc oedema, grading its severity, and recognising its cause, is crucial in order to treat patients in a timely manner and limit vision loss. Some ocular fundus features, in light of a patient's history and visual symptoms, may suggest a specific mechanism or aetiology of the visible disc oedema, but current criteria can at most enable an educated guess as to the most likely cause. In many cases only the clinical evolution and ancillary testing can inform the exact diagnosis. The development of ocular fundus imaging, including colour fundus photography, fluorescein angiography, optical coherence tomography, and multimodal imaging, has provided assistance in quantifying swelling, distinguishing true optic disc oedema from pseudo-optic disc oedema, and differentiating among the numerous causes of acute optic disc oedema. However, the diagnosis of disc oedema is often delayed or not made in busy emergency departments and outpatient neurology clinics. Indeed, most non-eye care providers are not able to accurately perform ocular fundus examination, increasing the risk of diagnostic errors in acute neurological settings. The implementation of non-mydriatic fundus photography and artificial intelligence technology in the diagnostic process addresses these important gaps in clinical practice.

Optimising fundoscopy practices across the medical spectrum: A focus group study

INTRODUCTION

Fundoscopy can be of great clinical value, yet remains underutilised. Educational attempts to improve fundoscopy utilisation have had limited success. We aimed to explore the barriers and facilitators underlying the uptake of clinical direct ophthalmoscopy across a spectrum of medical specialties and training levels.

METHODS

Ten focus groups were conducted with medical students (n = 42), emergency department doctors (n = 24), basic physician trainees (n = 7), hospital physicians (n = 6) and general practitioners (n = 7). Independent thematic analysis of transcripts was conducted by three investigators. A consensus thematic framework was developed, and transcripts were reanalysed using this framework.

RESULTS

Thematic analysis identified seven main themes: (1) technical barriers to performing fundoscopy examinations; (2) clinical culture and expectations regarding fundoscopy; (3) the influence of fundoscopy on clinical management; (4) motivation to perform the examination; (5) novel technology including smartphone fundoscopy, and the value of a digital fundus image; (6) training requirements, and; (7) use of limited resources.

CONCLUSION

Our results build a more nuanced picture of the factors which determine fundoscopy utilisation. As current barriers limit practice by clinicians and medical students, expertise and confidence performing and interpreting fundoscopy are lost. This shifts the balance of perceived clinical utility to futility in changing patient management, and reinforces a cycle of reducing fundoscopy utilisation. We identified important cultural barriers such as accepted incompetence, and misperceptions of senior discouragement. Emerging technologies reduce the technical barriers to fundoscopy. Therefore education should: focus on detecting pathology from digital images; clarify the role of fundoscopy in patient management, and; be targeted at key career progression points.

Screening for ocular candidiasis among patients with candidemia: Is it time to change practice?

Ocular candidiasis (OC) complicates ~10% of candidemia and carries potentially severe morbidity. There are conflicting recommendations about the need for routine funduscopic examinations of candidemic patients. Indirect funduscopy is accurate and safe in diagnosing OC, and positive findings change recommended treatment. However, conclusive evidence is lacking that treatment changes improve outcomes. Bringing perspectives as infectious diseases physicians and ophthalmologists, we review controversies about OC and endorse routine screening during candidemia. We acknowledge difficulties in obtaining inpatient ophthalmologic consults, and recommend studies evaluating digital fundus photography and tele-ophthalmology as an alternative to funduscopic examinations by ophthalmologists in asymptomatic patients.

Epidemiology, investigation, management, and outcome of headache in emergency departments (HEAD study)-A multinational observational study

OBJECTIVE

To describe the epidemiology of nontraumatic headache in adults presenting to emergency departments (EDs).

BACKGROUND

Headache is a common reason for presentation to EDs. Little is known about the epidemiology, investigation, and treatment of nontraumatic headache in patients attending EDs internationally.

METHODS

An international, multicenter, observational, cross-sectional study was conducted over one calendar month in 2019. Participants were adults (≥18 years) with nontraumatic headache as the main presenting complaint. Exclusion criteria were recent head trauma, missing records, interhospital transfers, re-presentation with same headache as a recent visit, and headache as an associated symptom. Data collected included demographics, clinical assessment, investigation, treatment, and outcome.

RESULTS

We enrolled 4536 patients (67 hospitals, 10 countries). "Thunderclap" onset was noted in 14.2% of cases (644/4536). Headache was rated as severe in 27.2% (1235/4536). New neurological examination findings were uncommon (3.2%; 147/4536). Head computed tomography (CT) was performed in 36.6% of patients (1661/4536), of which 9.9% showed clinically important pathology (165/1661). There was substantial variation in CT scan utilization between countries (15.9%-75.0%). More than 30 different diagnoses were made. Presumed nonmigraine benign headache accounted for 45.4% of cases (2058/4536) with another 24.3% classified as migraine (1101/4536). A small subgroup of patients have a serious secondary cause for their headache (7.1%; 323/4536) with subarachnoid hemorrhage (SAH), stroke, neoplasm, non-SAH intracranial hemorrhage/hematoma, and meningitis accounting for about 1% each. Most patients were treated with simple analgesics (paracetamol, aspirin, or nonsteroidal anti-inflammatory agents). Most patients were discharged home (83.8%; 3792/4526). In-hospital mortality was 0.3% (11/4526).

CONCLUSION

Diagnosis and management of headache in the ED is challenging. A small group of patients have a serious secondary cause for their symptoms. There is wide variation in the use of neuroimaging and treatments. Further work is needed to understand the variation in practice and to better inform international guidelines regarding emergent neuroimaging and treatment.

Telemedicine in neuro-ophthalmology

PURPOSE OF REVIEW

The COVID-19 public health emergency accelerated the adoption of telemedicine in neuro-ophthalmology. This review surveys the current telemedicine landscape and discusses associated limitations and opportunities.

RECENT FINDINGS

Logistic considerations and regulatory changes related to the COVID-19 pandemic have spurred a proliferation of tele-neuro-ophthalmic practice. One significant benefit of telemedicine is increased access to the limited number of neuro-ophthalmologists in the country. Certain elements of the neuro-ophthalmic examination can be ascertained during a video visit or through the use of mobile applications. However, data quality can be limited and more direct evaluation of the fundus currently requires the implementation of imaging techniques, such as fundus photography and/or optical coherence tomography. For cases that require it, a 'hybrid' model can be adopted in which patients physically present to designated locations for testing and subsequently participate in televisit with the physician to discuss the assessment and plan.

SUMMARY

Telemedicine provides an alternate pathway for patients to access the limited resource of neuro-ophthalmic care, and it will likely persist beyond the current COVID-19 pandemic. It has some limitations currently but continued progress in technologic, legal, and reimbursement strategies will hopefully facilitate further adoption.

Impact on patient management of non-mydriatic fundus photography compared to direct ophthalmoscopy in a regional Australian emergency department

OBJECTIVE

To investigate the management impact of non-mydriatic fundus photography (NMFP) implementation for appropriate ED patients; compare the diagnostic accuracy of direct ophthalmoscopy (DO) and NMFP, and determine the prevalence of fundus pathology in a regional Australian ED.

METHODS

This before/after crossover study prospectively enrolled patients presenting with headache, neurological deficit, visual disturbance and/or hypertensive urgency. Patients received DO examination, then separate NMFP examination. Emergency clinicians (ECs) were surveyed on their patient management plans following both DO examination and NMFP imaging. Telemedicine review of NMFP images was performed by an ophthalmologist within 48 h, and any additional management changes were documented.

RESULTS

The use of NMFP influenced changes in management in 52 (39%) of 133 enrolled patients (95% confidence interval 31-48%). Of these, 65% were escalations of management due to acute fundus pathology, while 35% were de-escalating changes following normal fundus findings. ECs diagnostic accuracy for acute fundus pathology improved from 0% to 29% sensitivity, and 59% to 84% specificity using DO and NMFP respectively, and telemedicine registrar review increased this to 50% sensitivity and 86% specificity. The period prevalence of acute fundus pathology was 10.5% (95% confidence interval 6-17%).

CONCLUSION

The addition of NMFP images can significantly impact the management of ED patients requiring fundus examination, facilitating expedited and optimised patient care. NMFP improves ECs diagnostic acumen for fundus pathology over DO examination and telehealth specialist review is important for diagnostic accuracy. There is a clinically important prevalence of fundus pathology in this regional ED setting.

eFOCUS 2: A randomised crossover trial of smartphone fundoscopy and direct ophthalmoscopy aiming to improve optic disc interpretation by medical students with e-learning support

BACKGROUND

Ophthalmoscopy and its interpretation are complex. We aimed to compare the diagnostic accuracy of smartphone fundoscopy with traditional direct ophthalmoscopy for optic disc interpretation, with e-learning support.

METHODS

We conducted a randomised, crossover study of 102 medical students. Students were offered e-learning for optic disc interpretation. A fundoscopy objective structured clinical examination was conducted after an introductory lecture and 10-min practical training session on smartphone fundoscopy and traditional ophthalmoscopy. Participants examined patients and simulator slides with a randomised crossover between smartphone [D-eye (Padova, Italy) or iExaminer (Welch Allyn, Macquarie Park, Australia)] and traditional ophthalmoscopy (Welch Allyn). Optic discs were graded independently by three masked ophthalmologists. The primary outcome was the ability to interpret an optic disc as normal or abnormal. Secondary outcomes included other optic disc aspects; student preferences; and e-learning performance.

RESULTS

Students' agreement with the gold standard for an abnormal or normal disc was significantly greater using a smartphone (74.4%) than with direct ophthalmoscopy (68.1%, p = 0.032). More students preferred smartphone (74%) over direct ophthalmoscopy (26%, p < 0.001). E-learning led to an improvement in optic disc interpretation scores (mean improvement = 4.5%, 95% CI = 3.7-5.2, p < 0.001).

CONCLUSIONS

Medical students are more accurate at recognising an abnormal optic disc using smartphone fundoscopy than traditional direct ophthalmoscopy, and have a strong preference for smartphone fundoscopy. E-learning may improve the interpretation of optic disc abnormalities. Smartphone fundoscopy may mitigate some technical challenges of fundoscopy and reinvigorate use of this valuable clinical examination.

Teaching Smartphone Funduscopy with 20 Diopter Lens in Undergraduate Medical Education

Purpose

To assess attitudes of pre-clinical undergraduate medical students toward learning smartphone funduscopy (SF) and its appropriateness as a teaching tool.

Patients and Methods

Second year medical students received instruction on direct ophthalmoscopy (DO) and SF; they were then paired with a peer and randomly assigned to perform DO or SF first. The SF technique involved freehand alignment of the axes of the smartphone camera with a condenser lens. Both techniques were done through a maximally dilated pupil. A questionnaire was completed to acquire data on baseline experience, performance of both examination techniques, attitudes, and appropriateness. Statistical significance testing and Bland-Altman analysis were used to determine differences between DO and SF, and a multivariable mixed regression model was fitted to identify any predictors for positive attitudes toward DO or SF.

Results

One hundred thirty-seven (137) individuals completed the study. A similar proportion of students could identify the optic nerve, macula, and vessels using DO and SF. However, self-reported quality scores were higher for DO for the optic nerve (p = 0.006) and macula (p = 0.08). The mean (standard deviation) attempts to identify these major structures were 2.7 (SD 2.3) for DO and 4.5 (SD 2.9) for SF (p < 0.001). Attitudes of students were consistently more positive toward DO across the five questions assessed. A small subset of students had equally positive attitudes toward DO and SF. Improved quality scores were predictive of positive attitudes for both DO and SF. Ultimately, 24% of students preferred SF over DO.

Conclusion

Among inexperienced examiners of the fundus through a dilated pupil, SF is a non-inferior technique to DO in identifying structures. Despite overall favorable attitudes towards the more familiar DO, those students who quickly learned the SF technique had similar satisfaction scores. Teaching SF should be considered in undergraduate medical education.

Diagnostic Error of Neuro-ophthalmologic Conditions: State of the Science

BACKGROUND

Diagnostic error is prevalent and costly, occurring in up to 15% of US medical encounters and affecting up to 5% of the US population. One-third of malpractice payments are related to diagnostic error. A complex and specialized diagnostic process makes neuro-ophthalmologic conditions particularly vulnerable to diagnostic error.

EVIDENCE ACQUISITION

English-language literature on diagnostic errors in neuro-ophthalmology and neurology was identified through electronic search of PubMed and Google Scholar and hand search.

RESULTS

Studies investigating diagnostic error of neuro-ophthalmologic conditions have revealed misdiagnosis rates as high as 60%-70% before evaluation by a neuro-ophthalmology specialist, resulting in unnecessary tests and treatments. Correct performance and interpretation of the physical examination, appropriate ordering and interpretation of neuroimaging tests, and generation of a differential diagnosis were identified as pitfalls in the diagnostic process. Most studies did not directly assess patient harms or financial costs of diagnostic error.

CONCLUSIONS

As an emerging field, diagnostic error in neuro-ophthalmology offers rich opportunities for further research and improvement of quality of care.

Using non-mydriatic fundus photography to detect fundus pathology in Australian metropolitan emergency departments: A prospective prevalence and diagnostic accuracy study

OBJECTIVE

To determine the prevalence of fundus pathology in metropolitan Australian EDs utilising a non-mydriatic fundus photography screening programme. Secondary objectives include diagnostic accuracy among emergency physicians compared to telehealth ophthalmologist review.

METHODS

Prospective cross-sectional study investigating non-mydriatic fundus photography as a new diagnostic test in two tertiary Australian EDs. Consecutive adult patients were enrolled if they presented with headache, focal neurological deficit, visual disturbance or diastolic BP >120 mmHg. Diagnostic agreement was determined using kappa statistics and sensitivity and specificity using a reference standard consensus ophthalmology review.

RESULTS

A total of 345 consecutive patients were enrolled among whom 56 (16%, 95% confidence interval [CI] 13-21) had urgent fundus pathology. Agreement between emergency physician and ophthalmic assessment of fundus photographs was 74% (kappa = 0.196, P = 0.001). Emergency physicians had 40% sensitivity (95% CI 27-54) and 82% specificity (95% CI 76-86) for detecting urgent pathology on photographs.

CONCLUSIONS

Fundus photography detects a clinically significant proportion of fundus pathology and urgent diagnoses. Telehealth specialist image review is important to detect some important, time-critical illnesses that can be missed in routine care. This offers an accurate alternative to direct ophthalmoscopy that warrants further research in Australian EDs.

Tele-Neuro-Ophthalmology: Vision for 20/20 and Beyond

BACKGROUND

Telehealth provides health care to a patient from a provider at a distant location. Before the COVID-19 pandemic, adoption of telehealth modalities was increasing slowly but steadily. During the public health emergency, rapid widespread telehealth implementation has been encouraged to promote patient and provider safety and preserve access to health care.

EVIDENCE ACQUISITION

Evidence was acquired from English language Internet searches of the medical and business literature and following breaking news on the COVID-19 pandemic and responses from health care stakeholders, including policymakers, payers, physicians, health care organizations, and patients. We also had extensive discussions with colleagues who are developing telehealth techniques relevant to neuro-ophthalmology.

RESULTS

Regulatory, legal, reimbursement, and cultural barriers impeded the widespread adoption of telehealth before the COVID-19 pandemic. With the increased use of telehealth in response to the public health emergency, we are rapidly accumulating experience and an evidence base identifying opportunities and challenges related to the widespread adoption of tele-neuro-ophthalmology. One of the major challenges is the current inability to adequately perform funduscopy remotely.

CONCLUSIONS

Telehealth is an increasingly recognized means of health care delivery. Tele-Neuro-Ophthalmology adoption is necessary for the sake of our patients, the survival of our subspecialty, and the education of our trainees and students. Telehealth does not supplant but supplements and complements in-person neuro-ophthalmologic care. Innovations in digital optical fundus photography, mobile vision testing applications, artificial intelligence, and principles of channel management will facilitate further adoption of tele-neuro-ophthalmology and bring the specialty to the leading edge of health care delivery.

Optic Disc Classification by Deep Learning versus Expert Neuro-Ophthalmologists

OBJECTIVE

To compare the diagnostic performance of an artificial intelligence deep learning system with that of expert neuro-ophthalmologists in classifying optic disc appearance.

METHODS

The deep learning system was previously trained and validated on 14,341 ocular fundus photographs from 19 international centers. The performance of the system was evaluated on 800 new fundus photographs (400 normal optic discs, 201 papilledema [disc edema from elevated intracranial pressure], 199 other optic disc abnormalities) and compared with that of 2 expert neuro-ophthalmologists who independently reviewed the same randomly presented images without clinical information. Area under the receiver operating characteristic curve, accuracy, sensitivity, and specificity were calculated.

RESULTS

The system correctly classified 678 of 800 (84.7%) photographs, compared with 675 of 800 (84.4%) for Expert 1 and 641 of 800 (80.1%) for Expert 2. The system yielded areas under the receiver operating characteristic curve of 0.97 (95% confidence interval [CI] = 0.96-0.98), 0.96 (95% CI = 0.94-0.97), and 0.89 (95% CI = 0.87-0.92) for the detection of normal discs, papilledema, and other disc abnormalities, respectively. The accuracy, sensitivity, and specificity of the system's classification of optic discs were similar to or better than the 2 experts. Intergrader agreement at the eye level was 0.71 (95% CI = 0.67-0.76) between Expert 1 and Expert 2, 0.72 (95% CI = 0.68-0.76) between the system and Expert 1, and 0.65 (95% CI = 0.61-0.70) between the system and Expert 2.

INTERPRETATION

The performance of this deep learning system at classifying optic disc abnormalities was at least as good as 2 expert neuro-ophthalmologists. Future prospective studies are needed to validate this system as a diagnostic aid in relevant clinical settings. ANN NEUROL 2020;88:785-795.

Artificial Intelligence to Detect Papilledema from Ocular Fundus Photographs

BACKGROUND

Nonophthalmologist physicians do not confidently perform direct ophthalmoscopy. The use of artificial intelligence to detect papilledema and other optic-disk abnormalities from fundus photographs has not been well studied.

METHODS

We trained, validated, and externally tested a deep-learning system to classify optic disks as being normal or having papilledema or other abnormalities from 15,846 retrospectively collected ocular fundus photographs that had been obtained with pharmacologic pupillary dilation and various digital cameras in persons from multiple ethnic populations. Of these photographs, 14,341 from 19 sites in 11 countries were used for training and validation, and 1505 photographs from 5 other sites were used for external testing. Performance at classifying the optic-disk appearance was evaluated by calculating the area under the receiver-operating-characteristic curve (AUC), sensitivity, and specificity, as compared with a reference standard of clinical diagnoses by neuro-ophthalmologists.

RESULTS

The training and validation data sets from 6779 patients included 14,341 photographs: 9156 of normal disks, 2148 of disks with papilledema, and 3037 of disks with other abnormalities. The percentage classified as being normal ranged across sites from 9.8 to 100%; the percentage classified as having papilledema ranged across sites from zero to 59.5%. In the validation set, the system discriminated disks with papilledema from normal disks and disks with nonpapilledema abnormalities with an AUC of 0.99 (95% confidence interval [CI], 0.98 to 0.99) and normal from abnormal disks with an AUC of 0.99 (95% CI, 0.99 to 0.99). In the external-testing data set of 1505 photographs, the system had an AUC for the detection of papilledema of 0.96 (95% CI, 0.95 to 0.97), a sensitivity of 96.4% (95% CI, 93.9 to 98.3), and a specificity of 84.7% (95% CI, 82.3 to 87.1).

CONCLUSIONS

A deep-learning system using fundus photographs with pharmacologically dilated pupils differentiated among optic disks with papilledema, normal disks, and disks with nonpapilledema abnormalities. (Funded by the Singapore National Medical Research Council and the SingHealth Duke-NUS Ophthalmology and Visual Sciences Academic Clinical Program.).

Nonmydriatic Fundus Photography in Patients with Acute Vision Loss

Background: Acute visual loss is a common chief complaint in emergency department (ED) patients, but the scarcity of ophthalmologists in most EDs limits its evaluation. Introduction: Our objective was to evaluate whether nonmydriatic fundus photography (NMFP) in the ED helps triage patients with acute visual loss. Materials and Methods: We included 213 patients with acute visual loss evaluated in the ED with NMFP as part of the Fundus Photography versus Ophthalmoscopy Trial Outcomes in the ED studies. Demographics, referral patterns, results of NMFP, and final diagnoses were recorded. Results: A final ophthalmological diagnosis was made in 109/213 (51%) patients. NMFP allowed a definite diagnosis in 51/109 (47%) patients: 14 nonglaucomatous optic neuropathies, 10 papilledema, 13 acute retinal ischemia, 2 retinal detachments, 2 choroidal metastases, 4 maculopathies, and 6 glaucoma. In 58/109 (53%) patients, NMFP was not diagnostic even when interpreted remotely by ophthalmologists due to disorders undiagnosable with NMFP. Ophthalmology consultation was requested in 109/213 (51%) patients, 41/54 (76%) patients with abnormal NMFP versus 68/159 (43%) patients with normal NMPF (p < 0.001). Discussion: Although NMFP allowed rapid diagnosis in 51/213 (24%) patients presenting to the ED with acute visual loss, NMFP alone was not sufficient to detect all ocular diseases; ophthalmology consultation was more often requested when NMFP was abnormal. Conclusions: Our study emphasizes the limitations of teleophthalmology with NMFP in remotely detecting ocular diseases related to acute visual loss in the ED. NMFP helped triage and referral decisions and can be used to complement ophthalmology consultations in the ED.

Optic nerve head edema among patients presenting to the emergency department

OBJECTIVE

To determine the frequency of and predictive factors for optic nerve head edema (ONHE) among patients with headache, neurologic deficit, visual loss, or elevated blood pressure in the emergency department (ED).

METHODS

Cross-sectional analysis was done of patients with ONHE in the prospective Fundus Photography vs Ophthalmoscopy Trial Outcomes in the Emergency Department (FOTO-ED) study. Demographics, neuroimaging results, management, and patient disposition were collected. Patients in the ONHE and non-ONHE groups were compared with bivariate and logistic regression analyses.

RESULTS

Of 1,408 patients included, 37 (2.6%, 95% confidence interval 1.9-3.6) had ONHE (median age 31 [interquartile range 26-40] years, women 27 [73%], black 28 [76%]). ONHE was bilateral in 27 of 37 (73%). Presenting complaints were headache (18 of 37), visual loss (10 of 37), acute neurologic deficit (4 of 37), elevated blood pressure (2 of 37), and multiple (3 of 37). The most common final diagnoses were idiopathic intracranial hypertension (19 of 37), CSF shunt malfunction/infection (3 of 37), and optic neuritis (3 of 37). Multivariable logistic regression found that body mass index ≥35 kg/m² (odds ratio [OR] 1.9, p = 0.0002), younger age (OR 0.5 per 10-year increase, p < 0.0001), and visual loss (OR 5, p = 0.0002) were associated with ONHE. Patients with ONHE were more likely to be admitted (62% vs 19%), to be referred to other specialists (100% vs 54%), and to receive neuroimaging (89% vs 63%) than patients without ONHE (p < 0.001). Fundus photographs in the ED allowed initial diagnosis of ONHE for 21 of 37 (57%) patients. Detection of ONHE on ED fundus photography changed the final diagnosis for 10 patients.

CONCLUSIONS

One in 38 patients (2.6%) presenting to the ED with a chief complaint of headache, neurologic deficit, visual loss, or elevated blood pressure had ONHE. Identification of ONHE altered patient disposition and contributed to the final diagnosis, confirming the importance of funduscopic examination in the ED.

Ophthalmoscopy in the 21st century: The 2017 H. Houston Merritt Lecture

Although the usefulness of viewing the ocular fundus is well-recognized, ophthalmoscopy is infrequently and poorly performed by most nonophthalmologist physicians, including neurologists. Barriers to the practice of ophthalmoscopy by nonophthalmologists include not only the technical difficulty related to direct ophthalmoscopy, but also lack of adequate training and discouragement by preceptors. Recent studies have shown that digital retinal fundus photographs with electronic transmission and remote interpretation of images by an ophthalmologist are an efficient and reliable way to allow examination of the ocular fundus in patients with systemic disorders such as diabetes mellitus. Ocular fundus photographs obtained without pharmacologic dilation of the pupil using nonmydriatic fundus cameras could be of great value in emergency departments (EDs) and neurologic settings. The Fundus Photography vs Ophthalmoscopy Trial Outcomes in the Emergency Department (FOTO-ED) study showed that ED providers consistently failed to correctly identify relevant ocular funduscopic findings using the direct ophthalmoscope, and that nonmydriatic fundus photography was an effective alternate way of providing access to the ocular fundus in the ED. Extrapolating these results to headache clinics, outpatient neurology clinics, and adult and pediatric primary care settings seems self-evident. As technology advances, nonmydriatic ocular fundus imaging systems will be of higher quality and more portable and affordable, thereby circumventing the need to master the use of the ophthalmoscope. Visualizing the ocular fundus is more important than the method used. Ocular fundus photography facilitates nonophthalmologists' performance of this essential part of the physical examination, thus helping to reestablish the value of doing so.