Narrative review of artificial intelligence (AI) in neuro-ophthalmic education: approaches for AI literacy

Introduction

Background

There is a national shortage of neuro-ophthalmologists and a supply-demand mismatch in accessibility to neuro-ophthalmic care (1,2). In fact, a survey conducted by the North American Neuro-Ophthalmology Society showed that there were no practicing neuro-ophthalmologists in six states, and 180 more neuro-ophthalmologists are required in the U.S. to reach an adequate ratio of 1 neuro-ophthalmologist:1.2 million individuals (1,3). Prior literature also indicates a decline in the duration of ophthalmology training in medical schools worldwide over the past 20 years, and students have reported low skill and knowledge levels (50–56%), on average (4). Among neurology residents, 19% cited the absence of adequate exposure to neuro-ophthalmology as their reason for not pursuing a neuro-ophthalmology fellowship (5,6). Additionally, projected data predict a decline in the number of working ophthalmologists, accompanied by a concurrent increase in demand for ophthalmologists (7). The difficulty of neuro-ophthalmology training (1), including establishing clinical reasoning for the neuro-visual pathway and mastering differential diagnostic and therapeutic capabilities, as well as the necessity of ophthalmology specialists during the training process may shed further light onto supply-demand mismatch of neuro-ophthalmologists in the setting of supply-demand mismatch of ophthalmologists. Combined, these findings suggest that the future availability of ophthalmologists, including neuro-ophthalmologists, may be inadequate to educate and train learners, such as medical students, residents, and fellows.

Rationale and knowledge gap

Considering the current shortage of neuro-ophthalmologists (1,2), artificial intelligence (AI) has the potential to help bridge the educator-student gap and provider-patient gap upon achieving consistent accuracy of medical knowledge and diagnosis identification (2,8). The advantages of large language models (LLMs) include their accessibility and rapid speed (9,10), which may enable them to assist clinicians with diagnostics and assist trainees and learners in learning to identify neuro-ophthalmology conditions by comparing their own independent evaluations, such as differentials or diagnostics from imaging, with those of AI models. However, further information regarding the efficacy, including sensitivity and accuracy, of AI-based tools for neuro-ophthalmic education, as well as long-term teaching outcomes, such as learners’ retention of information taught with the support of AI-based tools, is necessary.

Objective

In this narrative review, we review the current literature and discuss the role of AI in neuro-ophthalmology education and in the responsible use of AI for neuro-ophthalmology learning. Compared to prior reviews in the field of AI and neuro-ophthalmology (11,12), this review focuses on the application of current evidence on AI models’ performance for medical education purposes, including utilizing education-focused search terms. Based on the data from the current literature on the performance of AI tools in neuro-ophthalmology, our goal is to provide actionable ideas for ways ophthalmology educators and trainees can harness the power of AI tools, through Socratic questioning and AI-generated simulated cases, to establish a framework for active learning in neuro-ophthalmic academic and clinical settings. We present this article in accordance with the Narrative Review reporting checklist (available at https://aes.amegroups.com/article/view/10.21037/aes-25-41/rc).

Methods

To conduct this narrative review, we searched two databases, PubMed and Google Scholar, using the following keywords separately and in combinations: Neuro-ophthalmology, Ophthalmology, Artificial Intelligence, Large Language Models, Deep Learning Models, Education, Medical Education, Teaching, Curriculum, Training, Diagnostics, and Diagnosis. Inclusion criteria included peer-reviewed sources that included information regarding neuro-ophthalmology and AI in the English language that were published through 2025. Exclusion criteria included non-peer-reviewed sources; sources that did not include any content related to neuro-ophthalmology and AI, even in the larger setting of general ophthalmology research; and non-English sources. The search was not limited by the year of publication. The search was performed between May and July 2025 and was further updated in October 2025. This review aims to focus on the role of AI for neuro-ophthalmic education. Hence, we included more education-related search terms. However, we included diagnostic-related terms in order to access sources regarding the accuracy of AI-based tools in guiding neuro-ophthalmic diagnostics to subsequently determine their potential effectiveness for future trainee education on imaging and diagnostics. The search strategy is summarized in Table 1.

Review of AI models for education on neuro-ophthalmic imaging and diagnostics

A review by Muro-Fuentes and Stunkel [2022] notes that errors in diagnosing and managing neuro-ophthalmic conditions may be linked to the lack of accurate examinations, inaccurate readings of imaging, and a lack of adequate neuro-ophthalmology experience (13). On the other hand, AI tools, such as deep learning-based tools, have been shown to evaluate images of optic discs with equal or greater accuracy than neuro-ophthalmologists (14,15). In fact, a recent study showed an AI model’s diagnostic sensitivity between pediatric papilledema versus pseudopapilledema based on 851 fundus photographs from 235 patients was higher than that of four human pediatric neuro-ophthalmologists (16). Such data sheds light on how AI may enhance diagnostic accuracy and the need for AI implementation. However, limitations of this study include that it was a retrospective study, and the fundus photographs were selected by pediatric neuro-ophthalmologists from academic institutions and may thus have reduced external validity (16).

Caution must be exercised when evaluating the accuracy of LLMs’ clinical diagnoses in neuro-ophthalmology, such as through subsequent confirmation with evidence-based sources, which should be explicitly communicated to learners when incorporating AI into neuro-ophthalmic education. There is great variability in the diagnostic accuracy of AI tools, likely related to the robustness of the training dataset, complexity of task, and model infrastructure (10,17,18). For instance, prior literature shows that the diagnostic accuracy of LLMs based on published case reports ranges between 59% [generative pre-training transformer (GPT)-3.5] to 82% (GPT-4) (8). A limitation of this study is the small sample size (22 case reports) and that case prompts were from a public online database that ChatGPT may have had access to (8). On the other hand, the cases were primarily from prior to the end of the ChatGPT training period, and the authors utilized the memorization effects Levenshtein detector (MELD) technique to show evidence that the utilized cases were most probably not a part of the ChatGPT training dataset (8). Furthermore, when neuro-ophthalmology study questions were inserted into publicly available AI chatbots, accuracy rates were 52.5% for ChatGPT by OpenAI, 55% for Bing by Microsoft, and 65% for Bard by Google AI; however, this study was similarly limited by its low sample size of questions (40 questions) (19). A study of image interpretation revealed that the ChatGPT-4 vision feature achieved 40% diagnostic accuracy for Hess screen charts and 60% accuracy for automated visual field images (20). Limitations of this study include that the clinical context of the images was not provided, possibly alluding to the low diagnostic accuracy, and the small sample size of inserted images (five Hess screen charts and five visual field images) (20). Notably, the accuracy of LLMs, such as ChatGPT-4, in answering board-style exam questions is higher for ophthalmology overall compared to neuro-ophthalmology-specific content, likely due to the limited coverage of neuro-ophthalmology in the LLM training datasets (11).

AI tools may be harnessed to provide a teaching reference for students to identify neuro-ophthalmic conditions from imaging and diagnostic data. For example, according to the Brain and Optic Nerve Study with Artificial Intelligence (BONSAI) study data, the deep learning system algorithm detected papilledema from fundus photography images with 87.5% accuracy (11,21). Notable strengths of this study, in contrast to the aforementioned studies, include its large sample size of training images (14,341 images from 6,779 patients) and external testing (1,505 images), statistical power, and diversity of images in the training set (11 counties) (21). However, limitations of this study include its retrospective nature, potential convenience sampling bias, potential concerns with generalizability since fundus photographs with pathology were obtained after dilation and the low threshold for classifying papilledema (21). Furthermore, deep learning systems have successfully identified healthy control images, glaucomatous optic neuropathy (ON), and non-glaucomatous ON with 99.1% accuracy, which is higher than the 75% accuracy rate reported among neuro-ophthalmologists and glaucoma subspecialists (11,22,23). Strengths of this study include the large sample size (3,815 fundus images), statistical power, and use of unstandardized images for generalizability (22). In fact, the CascadeNet-5 model deep learning model also yielded a 98% accuracy in detecting visual field defects within the Humphrey visual field (HVF) analyzer at the University of Washington dataset and has the potential to predict future visual field defects from a single image (24,25). Wen et al. [2019] noted that the model was developed with information from only one academic institution and only included pairs of HVF data that were within 5.5 years apart (24). However, the study team aimed to enhance the generalizability of the tool by including information across 20 years across providers and subspecialties (24). Additionally, this study includes a robust dataset with over 1.7 million perimetry points from 32,443 24-2 HVFs (24). Additionally, Szanto et al. developed a novel deep learning system with an extensive sample size (15,088 fundus photos from 5,866 eyes) that distinguished the etiology of optic disc swelling based on fundus images from patients with idiopathic intracranial hypertension (IIH), non-arteritic anterior ischemic optic neuropathy (NAION), and healthy eyes with 93.6% external validity (26). Limitations of this study include the use of images from clinical trials and public databases; however, strengths of this study include supplementation from cases across different institutions, robust sample size and statistical power, and use of an external dataset to minimize sampling bias (26).

Each AI model requires individualized testing, ideally with a standardized rubric that evaluates models across relevant pedagogical categories (i.e., clinical reasoning, comprehensiveness, clarity, evidence-based nature, and effectiveness) since different models yield varying accuracy levels across different populations. For example, machine learning technology has also achieved 80% accuracy in detecting multiple sclerosis from optical coherence tomography (OCT) imaging among children (11,27). This study had a lower sample size than the prior studies with 187 children with demyelinating diseases and 69 healthy controls, and this study excluded patients with acute optic neuritis (27); however, this small sample size may be reflective of the smaller population size of pediatric patients with multiple sclerosis than adult patients (28). Bénard-Séguin et al. developed an AI algorithm that detected non-multiple sclerosis ON from multiple sclerosis ON specifically based on fundus photographs with 76.2% accuracy after using a sample size of 262 patients with multiple sclerosis and 59 patients with non-multiple sclerosis ON (17). Additionally, Seok et al. developed a deep learning model that accurately identified multiple sclerosis and neuromyelitis optica spectrum disorder from magnetic resonance imaging (MRI) scans with 76.1% accuracy in a sample of 86 patients with multiple sclerosis and 70 patients with neuromyelitis optica spectrum disorder (29). While such accuracy rates may be lower than the previously described studies (e.g., BONSAI study) (21), they reflect the specific patient populations of interest (e.g., subgroups among patients with nonetheless) and nonetheless have the potential for facilitating nuanced student learning with further model training and expert oversight. Similarly, Motamedi et al. noted a deep learning network that identified optic neuritis from OCT imaging with 85% accuracy (30). While this study had a lower sample size (1,033 OCT images) than prior AI-diagnostic studies in neuro-ophthalmology, it addressed the gap in literature regarding the diagnostic accuracy in distinguishing images of previous optic neuritis versus healthy controls (30).

AI tools with high evidence-based diagnostic accuracy levels may be utilized for learners’ education. For instance, students may independently review fundus photographs, practice identifying papilledema, and then compare their responses with the diagnostic answer determined by the deep learning system of the BONSAI study (11,21). Similar self-evaluations may be performed by students with learning to identify ON and visual field defects from corresponding images with accuracy-verified AI tools serving as a potential “gold standard”. In other words, the results of AI-based imaging analyses may be compared with students’ assessments of such diagnostic images, enabling learners to develop a mental schema of the imaging characteristics that are associated with various neuro-ophthalmic disorders. Similarly, AI-based diagnoses from imaging results may be utilized as a focal point for discussions during physician-led clinical rounds among medical students, residents, and fellows.

Hybrid AI models in neuro-ophthalmology education

AI combined with human expertise is emerging as the gold standard for integrating AI into medical education (31-33), including neuro-ophthalmology education (34). These hybrid models use the strengths of AI, including efficiency, scalability, and pattern recognition, while ensuring safety, empathy, and contextual accuracy through human oversight (35). Studies show that when neuro-ophthalmology experts review and revise AI-generated answers, such as those from LLMs, the resulting educational content scores are highest in both quality and empathy in comparison to content created by AI or humans alone (36). This approach ensures that learners receive information that is not only factually accurate but also contextually appropriate and sensitive to patient needs (36). Human oversight is essential for validating AI outputs, correcting errors, and providing clinical judgment that AI may lack, especially in complex or ambiguous cases (11,37). For example, in a study of an AI tool’s accuracy in diagnosing pediatric papilledema versus pseudopapilledema, the authors note that the AI tool mischaracterized a patient’s eyes as having pseudopapilledema, whereas the patient has papilledema with almost resolved IIH (16). Furthermore, for complex diagnoses, such as detecting the subtype of optic neuritis, Bénard-Séguin et al. noted that a developed deep learning tool had 76.2% accuracy (17). The authors note that the lack of adequate data prevented them from being able to train the model to be more accurate (17). Also, the heterogeneity between the time of symptoms and the time of imaging, as well as the quality of the imaging, added to the complexity of AI tool’s task (17).

In educational settings, instructors can use AI-generated case scenarios, quizzes, or explanations as a starting point and subsequently tailor and elaborate these materials to fit learners’ needs (34). For example, virtual reality (VR) platforms, such as the MetaNODES head-mounted display, integrate intelligent AI-driven patient simulations with real-time guidance from expert educators, creating an engaging and hands-on learning environment in neuro-ophthalmology (34). For instance, this system helps learners visualize the symptoms of diverse neuro-ophthalmic disorders and understand the pathophysiology of neuro-ophthalmic conditions (e.g., cranial nerve palsies and extraocular muscle weakness) (34). It also allows learners to practice diagnostic reasoning and clinical skills with immediate feedback from both AI and human mentors (34). However, further refinement is necessary due to students’ feedback of motion sickness and unfamiliarity with operating the tool (34). Additionally, there may be a risk of users’ development of detachment from reality after use of the tool (34). In essence, hybrid models facilitate ongoing improvement of educational materials, and AI can rapidly generate or update content, while human experts ensure alignment with best practices and evolving standards (11,36).

Active learning is a powerful pedagogy that engages learners directly in the learning process, requiring them to participate in meaningful activities and think critically rather than passively receive information (38-40). Extensive evidence demonstrates that active learning consistently leads to better learning outcomes compared to traditional lecture-based methods (38). The application of hybrid generative AI models enhances active learning by introducing new methods for engagement, personalization, and critical thinking, making it more interactive, scalable, and creative, while also supporting instructors in delivering deeper learning and educational experiences (37,41). Furthermore, this supports adaptive learning, where educational resources evolve based on learner performance and feedback in the academic and clinical settings (37,42-44).

Practical approaches for implementing hybrid models

Modifying LLMs in the form of custom GPTs to retrieve information from pre-identified sources ensures that the output is based on reliable, evidence-based resources (45), such as the Neuro-Ophthalmology Virtual Education Library (NOVEL), an open-access bank of multitudinous educational content with about 20,000 learning items (46). These tools are primarily built upon comprehensive ophthalmology knowledge, and there is a need for neuro-ophthalmic-specialized AI tools to be developed for the education of trainees.

For curriculum design, neuro-ophthalmic educators may (I) incorporate modules that explicitly teach the principles and limitations of both AI and hybrid systems, emphasizing the importance of human oversight, especially in ambiguous cases such as pseudopapilledema versus papilledema with almost resolved IIH (16,37), (II) use case-based learning where students analyze both AI-only and hybrid AI-human solutions, fostering critical thinking about when and how to trust or question AI outputs, (III) facilitate hands-on sessions where learners interact with AI tools and then review and discuss outputs with human experts, (IV) encourage learners to edit or critique AI-generated responses, mirroring the expert-edited model shown to yield the highest educational value (36), (V) teach the ethical imperatives of human oversight in AI use, highlighting real-world examples where unchecked AI could lead to errors or bias, (VI) discuss regulatory and medico-legal responsibilities associated with hybrid decision-making in clinical education and practice (11), and (VII) foster collaboration between neuro-ophthalmologists, data scientists, and educators to co-develop hybrid educational tools that are clinically relevant and pedagogically sound. It is additionally salient for medical educators to ensure that hybrid AI models are competency-based, meaning they aim to help learners develop specific salient skills (competencies). Competencies may include, but are not limited to, problem-based learning, history-taking skills, clinical reasoning skills, and communication (47,48). The adoption of competency-based curricula (47,48) and the integration of technology (49) have further enhanced the effectiveness and accessibility of ophthalmic education, a trend that accelerated following the COVID-19 pandemic (50,51).

In short, hybrid AI models with human oversight represent a potential effective approach for AI literacy in neuro-ophthalmology education (37). By combining the efficiency and scalability of AI with the contextual judgment and empathy of human experts, these models ensure that learners receive high-quality, safe, and relevant education (37). Educational strategies should focus on developing skills to appraise AI outputs critically, understand the role of human oversight, and engage in continuous, collaborative learning (11,34,36).

Discussion

Generative AI and AI-based tools are continuously increasing, yet there is no current unifying consensus regarding the role of AI in medical education (52). Meanwhile, survey data from medical students display a need for knowledge regarding AI use to be incorporated into medical education (53). Due to the strong interest in using AI-based tools among medical trainees, there is a need to develop accessible, accurate, and efficacious chatbots for students’ use. One of the most prevalent AI-based tools in medical education is ChatGPT (54), which is an LLM that can be harnessed to improve the learning experience of medical trainees, including in the field of neuro-ophthalmology.

LLMs may play a vast role in supporting medical education, such as answering direct queries from students, responding to exam-style questions for test preparation, and teaching learners about formulating a differential diagnosis and management plan for simulated patient cases and conditions (11). Additionally, AI tools can conduct image-based diagnoses and accordingly allow learners to compare their own image-based diagnostic skills with the AI output as a learning reference (11).

LLMs may also assist with expanding training for primary care providers to identify life-threatening neuro-ophthalmic conditions or with enhancing comprehensive ophthalmologists’ knowledge of intricate neuro-ophthalmic conditions in geographic areas where access to a neuro-ophthalmologist is limited (1,2,8). The purpose of utilizing LLMs by ophthalmologists and trainees may range from clinical diagnostic support to education and research use (11). In fact, LLMs have the potential to shape learners’ education in neuro-ophthalmology, especially in regions where access to neuro-ophthalmologist mentors may be difficult (1,11). There is an increasing number of custom GPTs for ophthalmology medical education, such as “EyeGPT”, “Eye Teacher”, “Eye Assistant”, and “The GPT for Geographic Atrophy (GA)” (45,55). Furthermore, Mayorga et al. have developed and are validating a novel custom GPT for ophthalmology learners across training levels called Learning Utility for Clinical Interpretation & Analysis (“LUCIA”), which is currently in the beta testing stage (unpublished data). These tools utilize prompts entered by learners as well as domain-specific databases to generate educational output (45).

A notable aspect of medical education is developing diagnostic skills and schemas for various diseases (56). With its speed and very broad database of knowledge, AI-based tools, such as ChatGPT-based applications, may be used to educate trainees on creating differential diagnoses and expanding diagnostic schemas for a variety of neuro-ophthalmic conditions upon iterative standardized testing and continuous enhancement (57).

Generative AI, including LLMs, has been utilized in a diverse number of purposes within the field of ophthalmology, including diagnostic support, monitoring of disease progression or therapy efficacy, research tasks, and physician and patient education (58). The benefits of LLMs towards trainees’ education include the ability to personalize learning discussions for each student’s requests and gaps in knowledge, tailor responses based on students’ learning level, simulate standardized patient experiences, and further enhance students’ problem-based learning skills (58). Furthermore, AI tools have also been developing an increasing capacity to individualize treatment courses for each patient and to predict each patient’s prognosis, such as with the use of neuroimaging for neurodegenerative and neuro-ophthalmic disorders (59), and OCT for prediction of postoperative visual acuity in patients with full-thickness macular holes (60). Such technology has the potential to facilitate education on personalized medicine in neuro-ophthalmology, though further development of such AI tools with greater quantities of robust datasets (59) is necessary before application for education purposes.

Limitations of AI in neuro-ophthalmology education

When ChatGPT was requested to answer question banks that are used in residents’ preparation for the Ophthalmic Knowledge Assessment Program (OKAP), the accuracy scores ranged from 46% (OphthoQuestions practice question bank) to 59.4 % [Basic and Clinical Science Course (BCSC) Self-Assessment Program], which is lower than the average human test taker’s scores (58). Ophthalmic subspecialty topics with the highest accuracy include cornea, and the lowest accuracy includes neuro-ophthalmology, oculoplastics, and optics (58), thus underscoring the need for exercising caution and training learners in the responsible use of LLMs for neuro-ophthalmology, both clinically and educationally. Increased utilization of broader neuro-ophthalmology training data from institutionally backed learning resources may further enhance the accuracy of LLMs for teaching neuro-ophthalmology. Similarly, the accuracy of AI tools for diagnostic purposes depends on the quality, quantity, and robustness of the training datasets that are fed to the AI tool for development (17). The development of AI tools with heterogenous quality fundoscopy images, image positioning, duration of symptoms at the time of imaging, and definitions of diagnoses (i.e., such as including suspected multiple sclerosis within the multiple sclerosis category) plays a role in the accuracy level (17).

While AI has been shown to enhance learners’ experiences with individualized feedback (42-44), there is a need to ensure that students do not acquire incorrect, hallucinated, or oversimplified content from the AI-generated output, as well as a need to ensure the integrity and ethics of student-submitted work in the setting of AI accessibility (11,61,62). These limitations may be overcome through hybrid learning models when combined with human expertise and oversight, as AI’s full potential can then be harnessed, enhancing accuracy and interpretability and ensuring ethical delivery and practice (37). Furthermore, there is a notable distinction in the extent of the LLMs’ knowledge base and accuracy when it comes to general medical specialties (i.e., fundamentals of general ophthalmology) versus sub-specializations (i.e., neuro-ophthalmology) (58) based on the availability and nature of the LLMs’ training data, which warrants further exploration. Lastly, when teaching learners about AI literacy, it is salient to underscore the need for entered or engineered prompts as well as patient case simulations to be Health Insurance Portability and Accountability Act (HIPAA)-compliant and to not include protected health information (37). Finally, the use of AI has created new and potentially problematic issues surrounding scholarly effort, author attribution, and AI content generation (63). The use of AI “checkers”, however, is equally fraught with poor accuracy and other ethical issues that can generate both false positive and false negative results in the detection of AI use (63).

Limitations

This review is not without limitations. First, this narrative review does not include an exhaustive representation of all literature regarding AI and neuro-ophthalmic education, especially since it is based on an evaluation of two databases. Next, this review may not be generalizable to all settings globally since the inclusion criteria only included articles that were published in English. Lastly, the conclusions of this narrative review are based on published literature only; with the rapid advancement of AI and ongoing studies regarding the use of AI, this review represents a current snapshot in time. However, this review provides an up-to-date representation of the role of AI for neuro-ophthalmic education for students, educators, and clinicians to gain perspective from, alike. This review also provides evidence-based information and guidance for neuro-ophthalmic clinicians and educators to incorporate AI for teaching purposes. In light of the limitations of AI in neuro-ophthalmic education and of this review, future development of neuro-ophthalmic-specific AI models is necessary with large, diverse datasets, and future prospective, longitudinal research is needed regarding the role of AI, including hybrid AI with adaptive learning functionality, in neuro-ophthalmic educational settings with large, heterogenous samples.

Conclusions

In this narrative review, we describe the available literature and nuances regarding the use of generative AI tools in ophthalmic medical education, the applications of such technology for active learning purposes, and the available GPTs for ensuring a learner-centered approach that fosters deep engagement and critical thinking across trainee levels in ophthalmology and neuro-ophthalmology. A synthesis of the current literature highlights the increasing role of AI-based tools in ophthalmology and neuro-ophthalmology medical education across various trainee levels, which has the potential to transform the landscape of medical education and enhance global access to neuro-ophthalmic knowledge. Despite the numerous benefits of AI educational tools and great accuracy with AI technology in neuro-ophthalmology diagnostics, there are current limitations associated with AI educational tools in neuro-ophthalmology, including limited training data that is involved in generating AI output and limited neuro-ophthalmologist-developed algorithms. Hence, there is a strong need for learners (medical students, residents, and fellows) to be educated on AI literacy, to utilize hybrid AI for active learning with human oversight, and to conduct fact-checking of AI-generated output with credible, evidence-based gold-standard resources.

Acknowledgments

During the preparation of this work, the authors used ChatGPT to consider a potential manuscript structure and for gap determination after development of the full draft. After using this tool/service, the authors reviewed and edited the content as needed, and they take full responsibility for the content of the publication.

Footnote

Provenance and Peer Review: This article was commissioned by the Guest Editor (Andrew G. Lee) for the series “Special Consideration for Teaching and Learning in Neuro-Ophthalmology. The article has undergone external peer review.

Reporting Checklist: The authors have completed the Narrative Review reporting checklist. Available at https://aes.amegroups.com/article/view/10.21037/aes-25-41/rc

Peer Review File: Available at https://aes.amegroups.com/article/view/10.21037/aes-25-41/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://aes.amegroups.com/article/view/10.21037/aes-25-41/coif). The series “Special Consideration for Teaching and Learning in Neuro-Ophthalmology” was commissioned by the editorial office without any funding or sponsorship. The authors have no other conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. DeBusk A, Subramanian PS, Scannell Bryan M, et al. Mismatch in Supply and Demand for Neuro-Ophthalmic Care. J Neuroophthalmol 2022;42:62-7. [Crossref] [PubMed]
2. Pakravan P, Lai J, Cavuoto KM. Demographics, Practice Analysis, and Geographic Distribution of Neuro-Ophthalmologists in the United States in 2023. Ophthalmology 2024;131:333-40. [Crossref] [PubMed]
3. Lai KE, Ko MW. Can Tele-Neuro-Ophthalmology Be Useful Beyond the Pandemic? Curr Neurol Neurosci Rep 2023;23:15-23. [Crossref] [PubMed]
4. Spencer SKR, Ireland PA, Braden J, et al. A Systematic Review of Ophthalmology Education in Medical Schools: The Global Decline. Ophthalmology 2024;131:855-63. [Crossref] [PubMed]
5. Frohman LP. How can we assure that neuro-ophthalmology will survive? Ophthalmology 2005;112:741-3. [Crossref] [PubMed]
6. Loebel EM, Stein LK, Fara M, et al. Curriculum Innovations: "Eye"-ing Enhanced Educational Methods for Neurology Trainees: Pilot of an Online, Case-based Neuro-ophthalmology Interactive Workshop. Neurol Educ 2023;2:e200052. [Crossref] [PubMed]
7. Berkowitz ST, Finn AP, Parikh R, et al. Ophthalmology Workforce Projections in the United States, 2020 to 2035. Ophthalmology 2024;131:133-9. [Crossref] [PubMed]
8. Madadi Y, Delsoz M, Lao PA, et al. ChatGPT Assisting Diagnosis of Neuro-Ophthalmology Diseases Based on Case Reports. J Neuroophthalmol 2024;45:301-6. [Crossref] [PubMed]
9. Zarei M, Eftekhari Mamaghani H, Abbasi A, et al. Application of artificial intelligence in medical education: A review of benefits, challenges, and solutions. Medicina Clínica Práctica 2024;7:100422.
10. Leong YY, Vasseneix C, Finkelstein MT, et al. Artificial Intelligence Meets Neuro-Ophthalmology. Asia Pac J Ophthalmol (Phila) 2022;11:111-25. [Crossref] [PubMed]
11. Kenney RC, Requarth TW, Jack AI, et al. AI in Neuro-Ophthalmology: Current Practice and Future Opportunities. J Neuroophthalmol 2024;44:308-18. [Crossref] [PubMed]
12. Sathianvichitr K, Lamoureux O, Nakada S, et al. Through the eyes into the brain, using artificial intelligence. Ann Acad Med Singap 2023;52:88-95. [Crossref] [PubMed]
13. Muro-Fuentes EA, Stunkel L. Diagnostic Error in Neuro-ophthalmology: Avenues to Improve. Curr Neurol Neurosci Rep 2022;22:243-56. [Crossref] [PubMed]
14. Vasseneix C, Najjar RP, Xu X, et al. Accuracy of a Deep Learning System for Classification of Papilledema Severity on Ocular Fundus Photographs. Neurology 2021;97:e369-77. [Crossref] [PubMed]
15. Biousse V, Newman NJ, Najjar RP, et al. Optic Disc Classification by Deep Learning versus Expert Neuro-Ophthalmologists. Ann Neurol 2020;88:785-95. [Crossref] [PubMed]
16. Chang MY, Heidary G, Beres S, et al. Artificial Intelligence to Differentiate Pediatric Pseudopapilledema and True Papilledema on Fundus Photographs. Ophthalmol Sci 2024;4:100496. [Crossref] [PubMed]
17. Bénard-Séguin É, Nielsen C, Sarhan A, et al. The Role of Artificial Intelligence in Predicting Optic Neuritis Subtypes From Ocular Fundus Photographs. J Neuroophthalmol 2024;44:462-8. [Crossref] [PubMed]
18. Shao Y, Lou Y, Gao Y, et al. Expert consensus on the application of artificial intelligence in neuro-ophthalmic diseases. Visual Neuroscience 2025;42:e019.
19. Sensoy E, Citirik M. Examining the competence of artificial intelligence programs in neuro-ophthalmological disorders and analyzing their comparative superiority. Oman J Ophthalmol 2024;17:348-51. [Crossref] [PubMed]
20. Nikdel M, Ghadimi H, Suh DW, et al. Accuracy of the Image Interpretation Capability of ChatGPT-4 Vision in Analysis of Hess Screen and Visual Field Abnormalities. J Neuroophthalmol 2024;45:307-15. [Crossref] [PubMed]
21. Milea D, Najjar RP, Zhubo J, et al. Artificial Intelligence to Detect Papilledema from Ocular Fundus Photographs. N Engl J Med 2020;382:1687-95. [Crossref] [PubMed]
22. Yang HK, Kim YJ, Sung JY, et al. Efficacy for Differentiating Nonglaucomatous Versus Glaucomatous Optic Neuropathy Using Deep Learning Systems. Am J Ophthalmol 2020;216:140-6. [Crossref] [PubMed]
23. O'Neill EC, Danesh-Meyer HV, Kong GX, et al. Optic disc evaluation in optic neuropathies: the optic disc assessment project. Ophthalmology 2011;118:964-70. [Crossref] [PubMed]
24. Wen JC, Lee CS, Keane PA, et al. Forecasting future Humphrey Visual Fields using deep learning. PLoS One 2019;14:e0214875. [Crossref] [PubMed]
25. Abdalgadir T, Amin S, Helmy T, et al. Artificial intelligence techniques in visual field assessment using Humphrey field analysis - a survey. Int J Intell Computing Inf Sci 2024;24:82-98.
26. Szanto D, Erekat A, Woods B, et al. Deep Learning Approach Readily Differentiates Papilledema, Non-Arteritic Anterior Ischemic Optic Neuropathy, and Healthy Eyes. Am J Ophthalmol 2025;276:99-108. [Crossref] [PubMed]
27. Ciftci Kavaklioglu B, Erdman L, Goldenberg A, et al. Machine learning classification of multiple sclerosis in children using optical coherence tomography. Mult Scler 2022;28:2253-62. [Crossref] [PubMed]
28. Boiko A, Vorobeychik G, Paty D, et al. Early onset multiple sclerosis: a longitudinal study. Neurology 2002;59:1006-10. [Crossref] [PubMed]
29. Seok JM, Cho W, Chung YH, et al. Differentiation between multiple sclerosis and neuromyelitis optica spectrum disorder using a deep learning model. Sci Rep 2023;13:11625. [Crossref] [PubMed]
30. Motamedi S, Yadav SK, Kenney RC, et al. Prior optic neuritis detection on peripapillary ring scans using deep learning. Ann Clin Transl Neurol 2022;9:1682-91. [Crossref] [PubMed]
31. Ali M. The Role of AI in Reshaping Medical Education: Opportunities and Challenges. Clin Teach 2025;22:e70040. [Crossref] [PubMed]
32. Turner L, Knopp MI, Mendonca EA, et al. Bridging Artificial Intelligence and Medical Education: Navigating the Alignment Paradox. ATS Sch 2025;6:135-48. [Crossref] [PubMed]
33. Narayanan S, Ramakrishnan R, Durairaj E, et al. Artificial Intelligence Revolutionizing the Field of Medical Education. Cureus 2023;15:e49604. [Crossref] [PubMed]
34. Yang T, Tam V, Lam EY, et al. MetaNODES: A Virtual Reality-Based Approach to Intelligent Neuro-Ophthalmology Education. 2024 IEEE Digital Education and MOOCS Conference (DEMOcon); 2024:1-6.
35. Chiu TKF, Rospigliosi Pa. Encouraging human-AI collaboration in interactive learning environments. Interactive Learning Environments 2025;33:921-924.
36. Tailor PD, Dalvin LA, Starr MR, et al. A Comparative Study of Large Language Models, Human Experts, and Expert-Edited Large Language Models to Neuro-Ophthalmology Questions. J Neuroophthalmol 2025;45:71-7.
37. Sriram A, Ramachandran K, Krishnamoorthy S. Artificial Intelligence in Medical Education: Transforming Learning and Practice. Cureus 2025;17:e80852. [Crossref] [PubMed]
38. Freeman S, Eddy SL, McDonough M, et al. Active learning increases student performance in science, engineering, and mathematics. Proc Natl Acad Sci U S A 2014;111:8410-5. [Crossref] [PubMed]
39. Dubinsky JM, Hamid AA. The neuroscience of active learning and direct instruction. Neurosci Biobehav Rev 2024;163:105737. [Crossref] [PubMed]
40. Graffam B. Active learning in medical education: strategies for beginning implementation. Med Teach 2007;29:38-42. [Crossref] [PubMed]
41. Pahi K, Hawlader S, Hicks E, et al. Enhancing active learning through collaboration between human teachers and generative AI. Computers and Education Open 2024;6:100183.
42. Strielkowski W, Grebennikova V, Lisovskiy A, et al. AI-driven adaptive learning for sustainable educational transformation. Sustainable Development 2025;33:1921-1947.
43. Katona J, Gyonyoru KIK. AI-based Adaptive Programming Education for Socially Disadvantaged Students: Bridging the Digital Divide. TechTrends 2025;69:925-942.
44. du Plooy E, Casteleijn D, Franzsen D. Personalized adaptive learning in higher education: A scoping review of key characteristics and impact on academic performance and engagement. Heliyon 2024;10:e39630. [Crossref] [PubMed]
45. Mayorga E. LUCIA: AI Tools for Assisting Ophthalmologists. Available online: https://www.linkedin.com/pulse/lucia-ai-tool-assisting-ophthalmologists-eduardo-mayorga-md-13cpf
46. Seay MD, Kedar S, Hull BE, et al. The Neuro-Ophthalmology Virtual Education Library (NOVEL). Neuroophthalmology 2024;48:444-7. [Crossref] [PubMed]
47. Lee AG, Beaver HA, Greenlee E, et al. Teaching and assessing systems-based competency in ophthalmology residency training programs. Surv Ophthalmol 2007;52:680-9. [Crossref] [PubMed]
48. Succar T, McCluskey P, Grigg J. Enhancing Medical Student Education by Implementing a Competency-Based Ophthalmology Curriculum. Asia Pac J Ophthalmol (Phila) 2017;6:59-63. [Crossref] [PubMed]
49. Succar T, Zebington G, Billson F, et al. The impact of the Virtual Ophthalmology Clinic on medical students' learning: a randomised controlled trial. Eye (Lond) 2013;27:1151-7. [Crossref] [PubMed]
50. Wendt S, Abdullah Z, Barrett S, et al. A virtual COVID-19 ophthalmology rotation. Surv Ophthalmol 2021;66:354-61. [Crossref] [PubMed]
51. Succar T, Beaver HA, Lee AG. Impact of COVID-19 pandemic on ophthalmology medical student teaching: educational innovations, challenges, and future directions. Surv Ophthalmol 2022;67:217-25. [Crossref] [PubMed]
52. Lee J, Wu AS, Li D, et al. Artificial Intelligence in Undergraduate Medical Education: A Scoping Review. Acad Med 2021;96:S62-70. [Crossref] [PubMed]
53. Civaner MM, Uncu Y, Bulut F, et al. Artificial intelligence in medical education: a cross-sectional needs assessment. BMC Med Educ 2022;22:772. [Crossref] [PubMed]
54. Maaß L, Grab-Kroll C, Koerner J, et al. Artificial Intelligence and ChatGPT in Medical Education: A Cross-Sectional Questionnaire on students' Competence. J CME 2025;14:2437293. [Crossref] [PubMed]
55. Chen X, Zhao Z, Zhang W, et al. EyeGPT for Patient Inquiries and Medical Education: Development and Validation of an Ophthalmology Large Language Model. J Med Internet Res 2024;26:e60063. [Crossref] [PubMed]
56. LeClair RJ, Cleveland JL, Eden K, et al. An integrated pre-clerkship curriculum to build cognitive medical schema: It’s not just about the content. Front Physiol 2023;14:1148916. [Crossref] [PubMed]
57. Hadi A, Tran E, Nagarajan B, et al. Evaluation of ChatGPT as a diagnostic tool for medical learners and clinicians. PLoS One 2024;19:e0307383. [Crossref] [PubMed]
58. Heinke A, Radgoudarzi N, Huang BB, et al. A review of ophthalmology education in the era of generative artificial intelligence. Asia Pac J Ophthalmol (Phila) 2024;13:100089. [Crossref] [PubMed]
59. Kumar R, Waisberg E, Ong J, et al. Artificial Intelligence-Based Methodologies for Early Diagnostic Precision and Personalized Therapeutic Strategies in Neuro-Ophthalmic and Neurodegenerative Pathologies. Brain Sci 2024;14:1266. [Crossref] [PubMed]
60. Rizzo S, Savastano A, Lenkowicz J, et al. Artificial Intelligence and OCT Angiography in Full Thickness Macular Hole. New Developments for Personalized Medicine. Diagnostics (Basel) 2021;11:2319. [Crossref] [PubMed]
61. Zhou J, Zhang J, Wan R, et al. Integrating AI into clinical education: evaluating general practice trainees’ proficiency in distinguishing AI-generated hallucinations and impacting factors. BMC Med Educ 2025;25:406. [Crossref] [PubMed]
62. Gruenhagen JH, Sinclair PM, Carroll JA, et al. The rapid rise of generative AI and its implications for academic integrity: Students’ perceptions and use of chatbots for assistance with assessments. Computers and Education: Artificial Intelligence 2024;7:100273.
63. Popkov AA, Barrett TS. AI vs academia: Experimental study on AI text detectors' accuracy in behavioral health academic writing. Account Res 2025;32:1072-88. [Crossref] [PubMed]