Search Results

Abstract

Objetivo: Avaliar indicações, evolução e complicações do uso do adesivo de cianoacrilato no tratamento dos afinamentos e perfurações corneanas. Material e Métodos: Vinte e seis olhos de 26 pacientes portadores de afinamento ou perfuração corneana (£ 2,5 mm de diâmetro) receberam adesivo tecidual à base de metil 2-cianoacrilato e, a seguir, lente de contato terapêutica hidrofílica sobre a área colada. Resultados: Entre as indicações para aplicação do adesivo de cianoacrilato, detectaram-se 12 (46%) casos de perfuração corneana causada por úlcera de córnea infecciosa. Dos 26 olhos que receberam adesivo, 15 (57,7%) necessitaram de uma única aplicação e 11 (42,3%) de duas aplicações. Houve evolução para leucoma corneano após desprendimento da cola em 13 (50%) olhos; necessidade de intervenção cirúrgica em 4 (15%) e permanência da cola em 9 (35%). A acuidade visual manteve-se inalterada em 15 (57,7%) olhos, melhorou em 7 (27%) e diminuiu em 4 (15,3%). Conclusões: O adesivo de cianoacrilato pode ser usado com sucesso no tratamento imediato dos afinamentos e perfurações corneanas, prorrogando ou até mesmo dispensando um futuro procedimento cirúrgico.

Keywords: Córnea; perfurações; adesivo

Abstract

Os autores descrevem um novo método para produzir um instrumento para colheita de material de córnea. Tal instrumento pode ser manufaturado com uma cânula ou agulha fina. O material é moldado de forma plana com o auxílio de um laminador, ou de um martelo e placa de metal. Lixa fina é utilizada para arredondar as bordas e a ponta é ligeiramente dobrada. Um cabo pode ser adicionado de modo a facilitar a manipulação do instrumento, que pode ser reutilizado após lavagem e esterilização local na ponta do metal, ou por óxido de etileno.

Keywords: Espátula; Microbiologia; Diagnóstico; Bioengenharia; Córnea

Abstract

Objetivo: Comparar as medidas de acuidade visual, espessura macular central e área de neovasos ativos na angiofluoresceinografia submetidos a panfotocoagulação retiniana padrão ETDRS associado a injeção intravítrea de ranibizumabe versus panfotocoagulação padrão PASCAL associado a injeção intravítrea de ranibizumabe versus somente injeção intravítrea de ranibizumabe em pacientes com retinopatia diabética proliferativa.
Métodos: Pacientes com retinopatia diabética proliferativa e virgens de tratamento, randomicamente divididos nas três diferentes terapias retinianas. Panfotocoagulação no grupo ETDRS em 2 sessões (semanas 0 e 2) e no grupo PASCAL, na semana 0. Injeção intravítrea de ranibizumabe realizado ao fim da primeira sessão de laser em ambos os grupos: ETDRS e PASCAL, e na semana 0 no grupo injeção intravítrea de ranibizumabe. Avaliações oftalmológicas, tomografia de coerência óptica e angiofluoesceinografia realizados na visita basal e a cada 4 semanas por 48 semanas.
Resultados: Trinta pacientes (n=40 olhos) completaram as 48 semanas de seguimento. Após o tratamento, a acuidade visual melhorou significantemente em todas a visitas no grupo injeção intravítrea de ranibizumabe (p<0,05); em todas exceto na semana 4 no grupo ETDRS, em todas exceto nas semanas 4 e 8 no grupo PASCAL. Redução significativa na espessura do subcampo central foi evidenciada no grupo PASCAL nas semanas 4, 8 e 48; somente na semana 48 no grupo injeção intravítrea de ranibizumabe, e em nenhuma visita no grupo ETDRS. Redução também na área de neovasos ativos em todas as visitas em todos os grupos. Não houve diferença significante entre os três grupos com relação a mudança media na medidas de acuidade visual, espessura macular central ou área de neovasos ativos da visita inicial para a semana 48.
Conclusões: Somente IVB ou este associado a panfotocoagulação ETDRS ou PASCAL, apresentaram efeitos semelhantes em relação a medidas de acuidade visual, espessura do subcampo central e área de neovasos ativos no decorrer de 48 semanas de seguimento.

Keywords: Retinopatia diabetica; Retina; Diabetes; Fator A de crescimento do endotélio vascular; Inibidorres da angiogenese/uso terapêutico; Ranibizumab/uso terapêutico; Panfotocoagulação; Acuidade visual

Abstract

Objetivo: Investigar se as aberrações da córnea e as amplitudes de acomodação alteram mais em pacientes com esclerose múltipla do que em populações normais.
Métodos: Vinte pacientes previamente diagnosticados com esclerose múltipla com envolvimento do nervo óptico (grupo com eslerose múltipla) e 20 indivíduos saudáveis pareados por sexo e idade (grupo controle) foram incluídos no estudo. Pacientes com menos de 40 anos de idade foram incluídos em ambos os grupos devido à deterioração significativa de acomodação em pacientes com mais de 40 anos de idade. Para cada participante, a amplitude de acomodação foi medida em dioptrias pelo teste de lentes negativas e as aberrações de alta ordem foram avaliadas com o aberrômetro iDesign. Em seguida, a amplitude de acomodação e a média da raiz quadrada de aberrações de alta ordem foram comparadas entre os grupos.
Resultados: As médias da idade dos grupos com esclerose múltipla e controle foram 35,25 ± 4,52 anos e 32,28 ± 6,83 anos, respectivamente (p=0,170). A amplitude de acomodação foi de 4,05 ± 1,25 D no grupo com esclerose múltipla e 6,00 ± 1,03 D no grupo controle. A diferença entre os com esclerose múltipla e o grupo controle foi estatisticamente significativa (p<0, 001). A média da raiz quadrada das aberrações de alta ordem não foi significativamente diferente entre os grupos (com esclerose múltipla, 0,44 ± 0,22; controle, 0,43 ± 0,10, p<0,824). Não houve diferenças estatisticamente significativas entre os grupos em termos de alterações de aberrações entre a linha de base e o estímulo 5 D.
Conclusões: Este estudo mostra que a amplitude de acomodação diminuiu em pacientes com esclerose múltipla. Portanto, esses resultados podem causar possíveis razões de deficiências visuais transitórias em pacientes com esclerose múltipla. No entanto, esta diferença de amplitude de acomodação não fez uma diferença significativa entre os grupos quanto à alteração das aberrações de alta ordem durante a acomodação.

Keywords: Acomodação ocular; Esclerose múltipla; Nervo óptico

Abstract

Objetivo: Desenvolver um aplicativo (TopEye) na plataforma iOS para dispositivos móveis que possibilite a captação e interpretação do mapa de cores gerados por qualquer topógrafo corneano através da inteligência artificial (IA).
Métodos: A execução, acompanhamento e avaliação do projeto foi utilizada a metodologia Scrum, processo de desenvolvimento interativo e incremental para gerenciamento de projetos e desenvolvimento ágil de software. O banco de padrões de diagnóstico gerado consiste em 1172 exemplos, divididos em: 275 padrões esféricos, 302 regulares simétricos, 295 regulares assimétricos e 300 irregulares (ceratocone). Para o desenvolvimento da inteligência artificial do aplicativo, foi estabelecido o treinamento da rede com 240 imagens de cada tipo de padrão, totalizando 960 (81,91%) padrões. O restante das imagens, 212 (18,09%), foram utilizadas para testar o aplicativo e usadas para gerar os resultados. O processo é semiautomático, assim a captação da imagem topográfica é realizada com smartphone, o examinador realiza o contorno do relevo corneano manualmente para em seguida a rede neural realizar o diagnóstico.
Resultados: O aplicativo diagnosticou 201 (94,81%) imagens corretamente. De um total de 212 imagens, o algoritmo errou a classificação de apenas 11 (5,19%). A principal ocorrência de erro foi na distinção das classes simétrica e assimétrica. No rastreio do ceratocone o aplicativo alcançou 95,00% de sensibilidade e 98,68% especificidade.
Conclusão: O trabalho resultou na obtenção de um aplicativo eficiente na captura da imagem topográfica pela câmera do smartphone e na interpretação da mesma através da inteligência artificial aplicada.

Keywords: Dispositivos móveis; Inteligência artificial; Topografia corneana; Astigmatismo

Abstract

Objetivo: Avaliar os efeitos da aplicação pré-operatória de lágrimas artificiais combinadas com o fator de crescimento de fibroblastos básicos bovinos recombinantes na função da superfície ocular e níveis de fator inflamatório após cirurgia em pacientes com catarata complicada com olhos secos.
Métodos: Um total de 118 pacientes com catarata complicada com olhos secos (118 olhos), tratados entre fevereiro de 2019 e fevereiro de 2020, foram divididos em grupos de controle e de observação (n=59, 59 olhos) usando uma tabela de números aleatórios. Uma semana antes da cirurgia, o grupo controle recebeu colírio de hialuronato de sódio a 0,1% (lágrimas artificiais), enquanto o grupo de observação recebeu colírio Beifushu (fator de crescimento de fibroblastos básicos bovinos recombinantes), ambos, seis vezes ao dia, por uma semana. Antes do tratamento e um mês após a cirurgia, os escores de sintomas clínicos, índices de função da superfície ocular, níveis de fatores inflamatórios nas lágrimas e índices de estresse oxidativo foram comparados. A função da superfície ocular foi avaliada pelo questionário do índice de doença da superfície ocular, ensaio de tempo de ruptura do filme lacrimal, teste I de Schirmer e teste de coloração por fluoresceína da córnea. Os níveis de fatores inflamatórios nas lágrimas foram medidos.
Resultados: Não houve diferenças significativas nos dados gerais e no escore de sintomas clínicos, índice de doença da superfície ocular, tempo de ruptura do filme lacrimal, escore do teste I de Schirmer, pontuação do teste de coloração por fluoresceína da córnea, interleucina-6, fator de necrose tumoral alfa, malondialdeído, superóxido dismutase, peróxido lipídico e capacidade antioxidante total antes do tratamento entre os dois grupos (p>0,05). Após o tratamento, o escore de sintomas clínicos, índice de doença da superfície ocular, escore do teste de coloração por fluoresceína da córnea, fator de necrose tumoral alfa, interleucina-6, malondialdeído e peróxido lipídico diminuíram significativamente, e o tempo de ruptura do filme lacrimal, escore do teste I de Schirmer, superóxido dismutase e a capacidade antioxidante total aumentou em ambos os grupos. As melhorias no escore de sintomas clínicos, bem como os índices de função da superfície ocular, fatores inflamatórios e estresse oxidativo foram mais proeminentes no grupo de observação do que no grupo controle (p<0,05).
Conclusões: Lágrimas artificiais combinadas com fator de crescimento de fibroblastos básicos recombinantes antes da cirurgia melhoram notavelmente a função da superfície ocular, diminuem os níveis de fatores inflamatórios nas lágrimas e aliviam os sintomas de olho seco após a cirurgia em pacientes com catarata complicada com olhos secos.

Keywords: Lágrima artificial; Fator básico de crescimento de fibroblasto; Superfície ocular; Catarata; Olho seco

Abstract

PURPOSE: To compare the refractive prediction error of Hill-radial basis function 3.0 with those of 3 conventional formulas and 11 combination methods in eyes with short axial lengths.
METHODS: The refractive prediction error was calculated using 4 formulas (Hoffer Q, SRK-T, Haigis, and Hill-RBF) and 11 combination methods (average of two or more methods). The absolute error was determined, and the proportion of eyes within 0.25-diopter (D) increments of absolute error was analyzed. Furthermore, the intraclass correlation coefficients of each method were computed to evaluate the agreement between target refractive error and postoperative spherical equivalent.
RESULTS: This study included 87 eyes. Based on the refractive prediction error findings, Hoffer Q formula exhibited the highest myopic errors, followed by SRK-T, Hill-RBF, and Haigis. Among all the methods, the Haigis and Hill-RBF combination yielded a mean refractive prediction error closest to zero. The SRK-T and Hill-RBF combination showed the lowest mean absolute error, whereas the Hoffer Q, SRK-T, and Haigis combination had the lowest median absolute error. Hill-radial basis function exhibited the highest intraclass correlation coefficient, whereas SRK-T showed the lowest. Haigis and Hill-RBF, as well as the combination of both, demonstrated the lowest proportion of refractive surprises (absolute error >1.00 D). Among the individual formulas, Hill-RBF had the highest success rate (absolute error ≤0.50 D). Moreover, among all the methods, the SRK-T and Hill-RBF combination exhibited the highest success rate.
CONCLUSIONS: Hill-radial basis function showed accuracy comparable to or surpassing that of conventional formulas in eyes with short axial lengths. The use and integration of various formulas in cataract surgery for eyes with short axial lengths may help reduce the incidence of refractive surprises.

Keywords: Cataract; Lenses, intraocular; Axial length, eye; Refractive errors; Artificial intelligence

Abstract

PURPOSE: Myopia, or nearsightedness, is one of the most common eye conditions worldwide. However, a comparison of the effectiveness of different laser-assisted interventions is lacking. Thus, we aimed to compare the efficacy and safety of LASIK and IntraLASIK in addressing myopia.
METHODS: The study was conducted in two ophthalmology clinics in Beijing, China, in 2022. A total of 84 patients (152 eyes) with different degrees of myopia were examined and underwent LASIK (n=46, 80 eyes) or IntraLASIK (n=38, 72 eyes). Keratometry, corneal topography, pachymetry, visual acuity evaluation, and corneal biomechanical analysis were performed before and after the intervention.
RESULTS: IntraLASIK produced more precise flaps than LASIK, with deviations of <8 mm and 0.1 mm from the intended thickness and diameter, respectively. LASIK resulted in nonuniform flaps, with thickness deviations of 5-86 mm. IntraLASIK demonstrated a superior efficacy for patients with severe myopia and thin corneas, with a mean spherical equivalent of 0.9 D at 6 months compared to the 1.4 D for LASIK. Approximately 91% and 83% of the patients with mild to moderate and severe myopia, respectively, achieved results within ± 0.49 D from the refractive target with IntraLASIK.
CONCLUSIONS: Corneal hysteresis and corneal resistance factor decreased with an increase in laser intensity, and they decreased faster with thinner corneas. Thus, IntraLASIK is more useful than LASIK in patients with thin corneas and severe myopia.

Keywords: Myopia; Lasers; Cornea; Keratomileusis; Laser in situ

Abstract

PURPOSE: We developed an artificial intelligence program for calculating intraocular lenses and analyzed its accuracy rate via ultrasonic biometry. This endeavor is aimed at enhancing precision and efficacy in the selection of intraocular lenses, particularly in cases where optical biometry is unavailable.
METHODS: Data was collected from the Hospital de Clínicas de Porto Alegre, which included cases of phacoemulsification with intraocular lens implantation, in which the lens selection was based on ultrasonic biometry. The program, implemented in Python, Java, and PHP, employs the ridge regression method. Two design options were developed: a basic model, which uses only keratometry variables (K1 and K2), axial size and final target refraction in the spherical equivalent, and an advanced model, which incorporates preoperative refraction and the patient's age. The Universal Barrett II formula was used to compare both models.
RESULTS: The sample consisted of 486 eyes from 313 patients, with 350 eyes used for program training and 136 for program validation. The spherical equivalent hit rates, with a variation of ±0.5 D, were 86% and 87.5% for the basic and advanced models, respectively, with no statistically significant difference between them. With the Barret Universal II formula, the success rate was 69%, which was significantly different from the values of the two aforementioned models (p<0.0001). The system was better for medium and long eyes but worse for short eyes (<=22.00 mm).
CONCLUSION: The developed artificial intelligence program was superior to the Barrett formula in terms of performance, in the general context and within the subgroup of patients with longer eyes. This innovation can considerably contribute to the selection of intraocular lenses, particularly in cases where optical biometry is unavailable.

Keywords: Biometry; Intraocular lens; Cataract; Artificial intelligence

Abstract

PURPOSE: Standard automated perimetry has been the standard method for measuring visual field changes for several years. It can measure an individual’s ability to detect a light stimulus from a uniformly illuminated background. In the management of glaucoma, the primary objective of perimetry is the identification and quantification of visual field abnormalities. It also serves as a longitudinal evaluation for the detection of disease progression. The development of artificial intelligence-based models capable of interpreting tests could combine technological development with improved access to healthcare.
METHODS: In this observational, cross-sectional, descriptive study, we used an artificial intelligence-based model [Inception V3] to interpret gray-scale crops from standard automated perimetry that were performed in an ophthalmology clinic in the Brazilian Amazon rainforest between January 2018 and December 2022.
RESULTS: The study included 1,519 standard automated perimetry test results that were performed using Humphrey HFA-II-i-750 (Zeiss Meditech). The Subsequently, 70%, 10%, and 20% of the dataset were used for training, validation, and testing, respectively. The model achieved 80% (68.23%–88.9%) sensitivity and 94.64% (88.8%–98%) specificity for detecting altered perimetry results. Furthermore, the area under the receiver operating characteristic curve was 0.93.
CONCLUSIONS: The integration of artificial intelligence in the diagnosis, screening, and monitoring of pathologies represents a paradigm shift in ophthalmology, enabling significant improvements in safety, efficiency, availability, and accessibility of treatment.

Keywords: Glaucoma; Disease progression; Perimetry; Visual Fields; Visual field tests; Artificial intelligence; Neural networks, computers; Machine learning

Abstract

PURPOSE: This study aimed to evaluate the practices employed by oculoplastic surgeons in the assessment and management of anophthalmic sockets and external ocular prostheses.
METHODS: Oculoplastic surgeons from two countries, who specialized in the management of anophthalmic sockets, participated in a web-based survey. Data collected included demographics, types of surgery, implant use, external ocular prostheses management (including fabrication and cleaning), complications encountered, and follow-up times. The frequencies and distributions of the responses were statistically analyzed.
RESULTS: A total of 177 oculoplastic surgeons participated, 113 (63.8%) from Brazil, the remainder from Spain. Evisceration was the preferred surgical procedure of 149 (84.2%) surgeons. The most commonly reported indication for enucleation was a painful blind eye (n=103, 58.1%; both Brazil and Spain, p<0.001). Brazilian surgeons preferred polymethyl methacrylate implants (n=65, 57.5%), while Spanish surgeons favored porous polyethylene implants (n=53, 82.8%; p<0.001). Discharge was the most frequently observed clinical feature during socket evaluation (n=164, 92.6%; p<0.001). Brazilian surgeons recommended daily (n=53, 46.9%) or weekly (n=41, 36.2%) cleaning of external ocular prostheses, while Spanish surgeons more commonly recommended monthly cleaning (n=31, 48.4%; p<0.001). The majority of Brazilian surgeons (n=83, 73.4%) advised patients to remove their external ocular prostheses at night. Only a small number of Spanish surgeons (n=3, 4.6%) suggested this practice (p<0.001). Overall, the follow-up recommendations varied, with 70 (39.5%) surgeons recommending follow-up based on individual case needs, and 59 (33.3%) suggesting annual visits (p<0.001). The primary indications for external ocular prostheses replacement were edge damage (n=75, 42.3%) and loss of volume (n=68, 38.4%). The replacement intervals given typically ranged from 1 to 5 years (n=92, 51.9%; p<0.001).
CONCLUSION: Oculoplastic surgeons in Brazil and Spain demonstrated similar practices in the management of anophthalmic sockets. However, notable differences were observed in the choice of implant materials, cleaning protocols, and recommendations regarding external ocular prostheses removal during sleep.

Keywords: Anophthalmos; Eye, artificial; Polymethyl methacrylate; Polyethylene; Surgeons; Surveys and questionnaires; Brazil; Spain.

Abstract

PURPOSE: This pilot study evaluated the diagnostic accuracy of a deep learning model for detecting pterygium in anterior segment photographs taken using smartphones in the Brazilian Amazon. The model’s performance was benchmarked against assessments made by experienced ophthalmologists, considered the clinical gold standard.
METHODS: In this cross-sectional study, 38 participants (76 eyes) from Barcelos, Brazil, were enrolled. Trained nonmedical health workers captured high-resolution anterior segment images using smartphones. These images were analyzed using a deep learning model based on the MobileNet-V2 convolutional neural network. Diagnostic metrics–including sensitivity, specificity, accuracy, positive predictive value, negative predictive value, and area under the receiver operating characteristic curve–were calculated and compared with the ophthalmologists’ evaluations.
RESULTS: The deep learning model achieved a sensitivity of 91.43%, specificity of 90.24%, positive predictive value of 88.46%, negative predictive value of 92.79%, and an area under the curve of 0.91. Logistic regression revealed no statistically significant association between pterygium and demographic variables such as age or gender.
CONCLUSIONS: The deep learning model demonstrated high diagnostic performance in identifying pterygium in a remote Amazonian population. These preliminary findings support the potential use of artificial intelligence–based tools to facilitate early detection and screening in underserved regions, thereby enhancing access to ophthalmic care.

Keywords: Pterygium/diagnostic imaging; Smartphone; Diagnostic techniques, ophthalmological; Deep learning; Telemedicine; Artificial intelligence; Cross-sectional studies; Brazil/epidemiology

Abstract

PURPOSE: This clinical study aimed to assess the effectiveness of microemulsion artificial tears containing povidone and propylene glycol in the management of dry eye disease. Secondary objectives included evaluating improvements in tear-film stability, measured by tear break-up time and corneal staining scores, along with the tolerability and safety of the formulation.
METHODS: This was a prospective, single-arm interventional study involving 30 participants (52 eyes) diagnosed with dry eye disease. Participants self-administered the investigational eye drops twice daily for 28 consecutive days. Primary and secondary outcomes included changes in the Ocular Surface Disease Index, tear break-up time, and corneal staining scores. Adverse events were documented throughout the study period.
RESULTS: Significant improvements in Ocular Surface Disease Index scores were observed, reflecting a reduction in dry eye disease symptoms. Tear break-up time increased notably between baseline and follow-up assessments, with the proportion of eyes exhibiting tear break-up time ≥10 srising from 25.0% to 63.5%. Additionally, the percentage of eyes with a corneal staining score of zero improved from 23.1% to 69.2%. Conjunctival staining also decreased, with the proportion of eyes with scores of 2 and 3 dropping from 11.5% to 3.8% and 5.8% to 0%, respectively.
CONCLUSIONS: The findings suggest that povidone and propylene glycol-based artificial tears significantly enhance tear-film stability and alleviate symptoms in patients with mild to moderate dry eye disease, with minimal adverse effects. This formulation represents a safe and effective short-term treatment option for dry eye disease management.

Keywords: Artificial tears; Dry eye disease; Tear-film stability; Propylene glycol; Povidone; Visual acuity; Surveys and questionnaires

Abstract

PURPOSE: Diabetic retinopathy screening in low- and middle-income countries is limited by restricted access to specialized care. Portable retinal cameras offer a practical alternative; however, image quality – affected by mydriasis – directly influences the performance of artificial intelligence models. This study evaluated the effect of mydriasis on image gradability and AI-based diabetic retinopathy detection in real-world, resource-limited settings.
METHODS: The proportions of gradable images were compared between mydriatic and non-mydriatic groups. Generalized estimating equations were used to identify factors associated with image gradability, including age, sex, race, diabetes duration, and systemic hypertension. A ResNet-200d model was trained on the mobile Brazilian Ophthalmological dataset and externally validated on both mydriatic and non-mydriatic images. Model performance was evaluated using accuracy, F1 score, area under the curve, and confusion matrix metrics. Sensitivity differences were assessed using the McNemar test, and area under the curves were compared using DeLong's test. The Youden index was used to determine optimal classification thresholds. Agreement between macula- and disc-centered images was analyzed using Cohen's κ.
RESULTS: The mydriatic group demonstrated a higher proportion of gradable images compared with the non-mydriatic group (82.1% vs. 55.6%; p<0.001). In non-mydriatic images, lower gradability was associated with systemic hypertension, older age, male sex, and longer diabetes duration. The AI model achieved better performance in mydriatic images (accuracy, 85.15%; area under the curve, 0.94) than in non-mydriatic images (accuracy, 79.68%; area under the curve, 0.93). The McNemar test showed a significant difference in sensitivity (p=0.0001), whereas DeLong's test revealed no significant difference in area under the curve (p=0.4666). The Youden index indicated that optimal classification thresholds differed based on mydriasis status. Agreement between image fields was moderate to substantial and improved with mydriasis.
CONCLUSION: Mydriasis significantly improves image gradability and enhances AI performance in diabetic retinopathy screening. Nonetheless, in low- and middle-income countries where pharmacologic dilation may be impractical, optimizing model calibration and thresholding for non-mydriatic images is essential to ensure effective AI implementation in real-world clinical environments.

Keywords: Artificial intelligence; Bias; Diabetic retinopathy; Portable camera; Retina

Abstract

PURPOSE: Natural language models and chatbots, particularly OpenAI’s Generative Pre-Trained Transformer architecture, have transformed human interaction with digital interfaces. The latest versions, including ChatGPT-4o, offer enhanced functionalities compared to their predecessors. This study evaluates the accuracy of ChatGPT-4, ChatGPT-4o, and Claude 3.5 Sonnet in answering questions from the Brazilian Retina and Vitreous Society certification exam.
METHODS: We compiled 200 multiple-choice questions from the Brazilian Retina and Vitreous Society 2018 and 2019 exams. Questions were categorized into three domains: Anatomy and Physiology of the Retina, Retinal Pathology, and Diagnosis and Treatment. Using a standardized prompt developed according to prompt design guidelines, we tested ChatGPT-4, ChatGPT-4o, and Claude 3.5 Sonnet, recording their first responses as final. Three retina specialists performed a qualitative analysis of the answers. Accuracy was determined by comparing responses to the official correct answers. Statistical analysis was conducted using chi-square tests and Cohen’s Kappa.
RESULTS: Claude 3.5 Sonnet achieved the highest overall accuracy (72.5%), followed by ChatGPT-4o (66.0%) and ChatGPT-4 (55.5%). Claude 3.5 Sonnet and ChatGPT-4o significantly outperformed ChatGPT-4 (p<0.01 and p=0.03, respectively), while no significant difference was observed between Claude 3.5 Sonnet and ChatGPT-4o (p=0.16). Model responses agreed 74.5% of the time, with a Cohen’s κ of 0.47. Retinal Pathology was the best-performing domain for all models, whereas Anatomy and Physiology of the Retina and Diagnosis and Treatment were the weakest domains for Claude 3.5 Sonnet and ChatGPT-4, respectively.
CONCLUSIONS: This study is the first to assess Claude 3.5 Sonnet, ChatGPT-4, and ChatGPT-4o in retina specialist certification exams. Claude 3.5 Sonnet and ChatGPT-4o significantly outperformed ChatGPT-4, highlighting their potential as effective tools for studying retina specialist board exams. These findings suggest that the enhanced functionalities of Claude 3.5 Sonnet and ChatGPT-4o offer substantial improvements in medical education contexts.

Keywords: Artificial intelligence; ChatGPT; Retina; Medical education; Ophthalmology, Large language model; Natural language processing