Interpretation of Clinical Retinal Images Using an Artificial Intelligence Chatbot.

Purpose: To assess the performance of Chat Generative Pre-Trained Transformer-4 in providing accurate diagnoses to retina teaching cases from OCTCases. Design: Cross-sectional study. Subjects: Retina teaching cases from OCTCases. Methods: We prompted a custom chatbot with 69 retina cases containing multimodal ophthalmic images, asking it to provide the most likely diagnosis. In a sensitivity analysis, we inputted increasing amounts of clinical information pertaining to each case until the chatbot achieved a correct diagnosis. We performed multivariable logistic regressions on Stata v17.0 (StataCorp LLC) to investigate associations between the amount of text-based information inputted per prompt and the odds of the chatbot achieving a correct diagnosis, adjusting for the laterality of cases, number of ophthalmic images inputted, and imaging modalities. Main Outcome Measures: Our primary outcome was the proportion of cases for which the chatbot was able to provide a correct diagnosis. Our secondary outcome was the chatbot's performance in relation to the amount of text-based information accompanying ophthalmic images. Results: Across 69 retina cases collectively containing 139 ophthalmic images, the chatbot was able to provide a definitive, correct diagnosis for 35 (50.7%) cases. The chatbot needed variable amounts of clinical information to achieve a correct diagnosis, where the entire patient description as presented by OCTCases was required for a majority of correctly diagnosed cases (23 of 35 cases, 65.7%). Relative to when the chatbot was only prompted with a patient's age and sex, the chatbot achieved a higher odds of a correct diagnosis when prompted with an entire patient description (odds ratio = 10.1, 95% confidence interval = 3.3-30.3, P < 0.01). Despite providing an incorrect diagnosis for 34 (49.3%) cases, the chatbot listed the correct diagnosis within its differential diagnosis for 7 (20.6%) of these incorrectly answered cases. Conclusions: This custom chatbot was able to accurately diagnose approximately half of the retina cases requiring multimodal input, albeit relying heavily on text-based contextual information that accompanied ophthalmic images. The diagnostic ability of the chatbot in interpretation of multimodal imaging without text-based information is currently limited. The appropriate use of the chatbot in this setting is of utmost importance, given bioethical concerns. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Accuracy of an Artificial Intelligence Chatbot's Interpretation of Clinical Ophthalmic Images.

Importance: Ophthalmology is reliant on effective interpretation of multimodal imaging to ensure diagnostic accuracy. The new ability of ChatGPT-4 (OpenAI) to interpret ophthalmic images has not yet been explored. Objective: To evaluate the performance of the novel release of an artificial intelligence chatbot that is capable of processing imaging data. Design, Setting, and Participants: This cross-sectional study used a publicly available dataset of ophthalmic cases from OCTCases, a medical education platform based out of the Department of Ophthalmology and Vision Sciences at the University of Toronto, with accompanying clinical multimodal imaging and multiple-choice questions. Across 137 available cases, 136 contained multiple-choice questions (99%). Exposures: The chatbot answered questions requiring multimodal input from October 16 to October 23, 2023. Main Outcomes and Measures: The primary outcome was the accuracy of the chatbot in answering multiple-choice questions pertaining to image recognition in ophthalmic cases, measured as the proportion of correct responses. χ2 Tests were conducted to compare the proportion of correct responses across different ophthalmic subspecialties. Results: A total of 429 multiple-choice questions from 136 ophthalmic cases and 448 images were included in the analysis. The chatbot answered 299 of multiple-choice questions correctly across all cases (70%). The chatbot's performance was better on retina questions than neuro-ophthalmology questions (77% vs 58%; difference = 18%; 95% CI, 7.5%-29.4%; χ21 = 11.4; P < .001). The chatbot achieved a better performance on nonimage-based questions compared with image-based questions (82% vs 65%; difference = 17%; 95% CI, 7.8%-25.1%; χ21 = 12.2; P < .001).The chatbot performed best on questions in the retina category (77% correct) and poorest in the neuro-ophthalmology category (58% correct). The chatbot demonstrated intermediate performance on questions from the ocular oncology (72% correct), pediatric ophthalmology (68% correct), uveitis (67% correct), and glaucoma (61% correct) categories. Conclusions and Relevance: In this study, the recent version of the chatbot accurately responded to approximately two-thirds of multiple-choice questions pertaining to ophthalmic cases based on imaging interpretation. The multimodal chatbot performed better on questions that did not rely on the interpretation of imaging modalities. As the use of multimodal chatbots becomes increasingly widespread, it is imperative to stress their appropriate integration within medical contexts.

THE ABILITY OF ARTIFICIAL INTELLIGENCE CHATBOTS ChatGPT AND GOOGLE BARD TO ACCURATELY CONVEY PREOPERATIVE INFORMATION FOR PATIENTS UNDERGOING OPHTHALMIC SURGERIES.

INTRODUCTION: To determine whether the two popular artificial intelligence chatbots, ChatGPT and Bard, can provide high-quality information concerning procedure description, risks, benefits, and alternatives of various ophthalmic surgeries. METHODS: ChatGPT and Bard were prompted with questions pertaining to the description, potential risks, benefits, alternatives, and implications of not proceeding with various surgeries in different subspecialties of ophthalmology. Six common ophthalmic procedures were included in the authors' analysis. Two comprehensive ophthalmologists and one subspecialist graded each response independently using a 5-point Likert scale. RESULTS: Likert grading for accuracy was significantly higher for ChatGPT in comparison with Bard (4.5 ± 0.6 vs. 3.8 ± 0.8, P < 0.0001). Generally, ChatGPT performed better than Bard even when questions were stratified by the type of ophthalmic surgery. There was no significant difference between ChatGPT and Bard for response length (2,104.7 ± 271.4 characters vs. 2,441.0 ± 633.9 characters, P = 0.12). ChatGPT responded significantly slower than Bard (46.0 ± 3.0 vs. 6.6 ± 1.2 seconds, P < 0.0001). CONCLUSION: Both ChatGPT and Bard may offer accessible and high-quality information relevant to the informed consent process for various ophthalmic procedures. Nonetheless, both artificial intelligence chatbots overlooked the probability of adverse events, hence limiting their potential and introducing patients to information that may be difficult to interpret.

The Cataract Surgery Learning Curve: Quantitatively Tracking a Single Resident's Operative Actions Throughout Their Training.

PURPOSE: To track operative phases of cataract surgery over a resident's training to measure action times and frequencies as a surrogate for competency and skill progression. DESIGN: An n = 1 panel study. METHODS: Cataract surgery video recordings performed by a single resident between 2021-2022 were collected. Only full-length videos of adequate quality without supervisor intervention were included. The start and end times of 19 distinct operative phases of cataract surgery were manually labeled by a trained annotator. Timeseries analysis was employed to measure the direction and magnitude of trends in the resident's surgical timing for each action across their first year of training. RESULTS: The dataset comprised 100 videos spread across the resident's sixth to 760th cases. The median total time was 11.6 minutes (IQR 10.1-14.4 minutes), with overall speed increasing at a rate of 43.4 seconds for every 10 videos (95% CI 35.1, 52.7 seconds). Nine operative phases significantly decreased in time throughout training. The main incision, phacoemulsification, and hydrodissection had the greatest improvements in speed relative to their average procedural time. There was an average of 26.9 distinct operative actions (excluding idle periods) in each video (range 20-50). CONCLUSIONS: This is the first study to quantitatively track operative times and frequencies across all relevant actions in cataract surgery and derive learning curves for each. Consistent with previous works, it was found that a basic level of surgical competency was achieved after performing 80 cases. In addition, results from this study indicated that the next level in skill advancement towards surgical finesse occurs after 300 cases.

Anemia and Idiopathic Intracranial Hypertension: A Systematic Review and Meta-analysis.

BACKGROUND: Idiopathic intracranial hypertension (IIH) is a condition typically affecting young, obese women. Although anemia is recognized as a risk factor of IIH from case reports, their relationship remains controversial as several comparative studies showed no significant association. This study aimed to examine the relationship between anemia and IIH. METHODS: MEDLINE, Embase, Cochrane Library, and grey literature were searched to September 2020. Primary studies on patients with diagnoses of anemia of any kind and IIH were included. Primary outcomes included the total number of cases of anemia and IIH. A meta-analysis on the prevalence of anemia in IIH compared with control patients was conducted. Grading of Recommendations, Assessment, Development and Evaluations (GRADE) was used to rate the quality of evidence. RESULTS: Overall, 62 cases and 5 observational or case-control studies were included. Pooled incidence of anemia in patients with IIH was 195/1,073 (18.2%). Patients with IIH (n = 774) had a significantly higher prevalence of anemia compared with controls (n = 230,981) (RR 1.44 [95% confidence interval 1.08-1.92]). Patients were 67.7% females and had a mean age of 22.4 years. The mean opening pressure was 37.9 cmH2O. Anemia was most commonly caused by iron deficiency (51.6%) and aplastic anemia (19.4%). Most patients (59.7%) showed improvement or resolution with anemia treatment only without intracranial pressure-lowering therapy. Evidence was limited because of high risk of reporting bias from the large number of case reports and case-control studies. CONCLUSIONS: Anemia is significantly more common in IIH compared with control patients, and case reports suggest a direct relationship. Complete blood counts should be considered in all patients with papilledema, particularly in atypical presentations (male, nonobese, nonperipapillary retinal hemorrhages, prominent risk factor for anemia) or in treatment-refractory IIH.

Binasal Visual Field Defects Due to Sickle Cell Maculopathy.

ABSTRACT: A 26-year-old African American man with sickle cell disease noticed blurry vision in both eyes after a recent complicated hospital admission for sickle cell crisis. Anterior and posterior segment examination of each eye was normal, but visual field testing revealed binasal scotomas. Optical coherence tomography of the macula demonstrated severe thinning of the temporal inner retina, suggesting previous bilateral occlusions of terminal retinal arterioles involving the temporal macula, an uncommon complication of sickle cell disease. This case is a reminder that retinal pathology should always be considered as a potential cause of unexplained visual field defects and highlights the role of macular OCT in evaluation of these patients.

Computer-Controlled CO2 Laser Ablation System for Cone-beam Computed Tomography and Digital Image Guided Endodontic Access: A Pilot Study.

INTRODUCTION: Ideal endodontic access provides unobstructed entry to the pulp chamber and visualization of the canal orifices while preserving the maximum amount of tooth structure. The aim of this study was to implement the use of lasers to accurately and predictably access teeth to follow the principles of minimally invasive endodontics. METHODS: Traditional, conservative, ultraconservative, bridge, truss, and orifice-directed accesses were performed. A computer-controlled 9.3-µm CO2 laser ablation system was assembled and coupled with custom software capable of combining cone-beam computed tomographic (CBCT) volumetric data with spatially calibrated digital images of teeth to provide an augmented reality environment for designing and preparing endodontic accesses. Twenty (N = 20) sound posterior teeth with fully developed root canal systems were imaged with CBCT scans and accessed via laser ablation in vitro. RESULTS: All 20 (20/20) teeth were successfully accessed without iatrogenic errors. Volumetric renderings from post-access CBCT scans were used to verify the access and determine accuracy qualitatively. The volumetric measurements of hard tissue removed were as follows: traditional = 39.41 mm3, conservative = 9.76 mm3, ultraconservative = 7.1 mm3, bridge = 11.53 mm3, truss = 19.21 mm3, and orifice directed = 16.86 mm3. CONCLUSIONS: Digital image guidance based on feature recognition and registration with CBCT data is a viable method to address the challenge of dynamic navigation for accessing the pulp chamber. Modern lasers with high pulse repetition rates integrated with computer-controlled scanning systems are suitable for the efficient cutting of dental hard tissues.

Smartphone Compatible versus Conventional Ophthalmoscope: A Randomized Crossover Educational Trial.

Objective The aim of the study is to compare performance and ease-of-use (EOU) of optic disk assessment using a smartphone direct ophthalmoscope attachment (D-EYE) to the gold standard direct ophthalmoscope (DO). Design The type of study involved is prospective, randomized, crossover, and educational trial. Participants The participants involved were first year medical students inexperienced in ophthalmoscopy. Methods Optic disks of standardized and volunteer patients were examined using the D-EYE and a conventional DO. Optic disk identification, EOU ratings of the devices, self-reported confidence level in their examination with the devices, and estimation of vertical cup-to-disk ratio (VCDR) were compared. Analyses included Chi-square tests, independent samples t -tests, correlations, and multivariable linear regression. Results Forty-four medical students voluntarily participated in the study. Students using the DO required more attempts (3.57 vs. 2.69, p = 0.010) and time (197.00 vs. 168.02 seconds, p = 0.043) to match the patient's fundus to the correct photograph. Overall EOU between the devices (6.40 vs. 4.79, p < 0.001) and overall confidence in examination (5.65 vs. 4.49, p = 0.003) were greater when using the D-EYE. There were no statistically significant differences in accuracy of VCDR estimations between the two ophthalmoscopes. Conclusion Smartphone ophthalmoscopy could offer additional learning opportunities in medical education and may be considered in clinical practice by non-specialist physicians given its greater EOU and increased success in visualizing the optic disk.

Resolution of papilledema associated with cyclosporine use after change to tacrolimus.

Use of medications including vitamin A derivatives and tetracyclines have been associated with papilledema and raised intracranial pressure. A 46-year-old woman was referred to neuro-ophthalmology for bilateral optic disc oedema and had a 7-year history of cyclosporine use after renal transplantation. She had preserved visual function and moderate bilateral optic disc oedema. Magnetic resonance imaging and magnetic resonance venography of the brain were normal apart from signs of raised intracranial pressure. Lumbar puncture revealed an elevated opening pressure of 40 cm of water with normal cerebrospinal fluid contents. Nephrology was consulted and cyclosporine was switched to tacrolimus and she was treated with acetazolamide. The papilledema resolved within 1 month of her initial visit. It is important to recognise the role that cyclosporine plays in raising intracranial pressure, especially in patients requiring immunosuppression, such as transplant patients. Tacrolimus is a suitable alternative in these cases.

Use of a fundus photograph matching program in imparting proficiency in ophthalmoscopy.

OBJECTIVE: To determine whether practice using an online fundus photograph program results in a long-term increase in proficiency with direct ophthalmoscopy skills in medical students. DESIGN: This study was a prospective medical education trial. Students were enrolled to participate in a voluntary Objective Structured Clinical Examination (OSCE) using 5 patients with ocular findings. Students who matched a minimum of 6 discs 16 months before the study were assigned to the intervention group and were compared with students who did not participate in the exercise. PARTICIPANTS: Forty-six second-year medical students at Queen's University: 15 in the intervention group and 31 in the control group. METHODS: Students were evaluated using the Queen's University Ophthalmoscopy OSCE Checklist (QUOOC). Students were asked to calculate the cup-to-disc ratio, comment on disc margins, and whether there was any macular pathology. Students participated in a summative OSCE as part of the curriculum in which all students attempted to match fundus photographs. RESULTS: Students in the intervention group performed significantly better on the QUOOC, with a mean score of 78.3% (± 4.2), compared with the control, who had a mean score of 69.4% (± 4.2) (p = 0.005). The intervention group was significantly more accurate at matching optic nerve photographs, with 100% (15/15) of the students correctly identifying the correct optic nerve on first attempt compared with 53.3% (16/30) in the control group (p = 0.0014). CONCLUSIONS: The use of an online peer fundus photograph program leads to a long-term increase in examination technique, proficiency in ophthalmoscopy, and accuracy at matching optic nerve photographs.

Evaluation of an online peer fundus photograph matching program in teaching direct ophthalmoscopy to medical students.

OBJECTIVE: Direct ophthalmoscopy is an important clinical skill that is often poorly performed by medical professionals and students. This is attributable to a declining emphasis on ophthalmology in medical school. We present and evaluate a self-directed approach of teaching ophthalmoscopy to medical students that is suitable for the current medical curriculum. DESIGN: Prospective medical education trial. PARTICIPANTS: Ninety-five second-year medical students at Queen's University: 32 in the experimental group and 63 in the control group. METHODS: The experimental group consisted of medical students who practised ophthalmoscopy with one another using an online peer fundus photograph matching exercise created by the Department of Ophthalmology at Queen's University. To use the program, students first examined a peer with an ophthalmoscope and then selected an online photograph of a fundus corresponding to that of the examinee. The program notifies students if a correct selection is made. To encourage use of the program, students participated in a 2-week ophthalmoscopy competition during their ophthalmology rotation. The control group consisted of students who did not participate in the learning exercise. RESULTS: On assessment at the end of the ophthalmology rotation, the experimental group (n = 32) was more accurate in matching fundus photographs compared with the control group (n = 63) (p = 0.02). Participants were faster at performing ophthalmoscopy at the end of the learning exercise (p < 0.01). All students in the experimental group reported increased confidence levels in ophthalmoscopy after participation in the learning exercise. CONCLUSIONS: Matching online peer fundus photographs in a self-directed manner appeared to increase the skill and confidence of medical students in ophthalmoscopy.

Examining the impact of early longitudinal patient exposure on medical students' career choices.

BACKGROUND: Medical schools include career direction experiences to help students make informed career decisions. Most experiences are short, precluding students from attaining adequate exposure to long-term encounters within medicine. We investigated the impact of the First Patient Program (FPP), which fosters longitudinal patient exposure by pairing junior medical students with chronically ill patients through their healthcare journey, in instilling career direction. METHODS: Medical students who completed at least 6-months in the FPP participated in a cross-sectional survey. Students' answers were analyzed with respect to the number of FPP appointments attended. Thematic analysis was conducted to explore qualitative responses. RESULTS: One hundred and forty-eight students participated in the survey. Only 28 (19%) students stated that the FPP informed their career decisions. Thirty-nine percent of students who attended four or more appointments indicated that the FPP informed their career decisions, compared to 16% of students who attended less (p=0.021). Thematic analysis revealed two themes: 1) Students focused mainly on patient encounters within FPP; and 2) Students sought career directions from other experiences. CONCLUSION: The majority of students did not attain career guidance from the FPP, but rather used the program to understand the impact of chronic illness from the patient's perspective.