Comparing generative and retrieval-based chatbots in answering patient questions regarding age-related macular degeneration and diabetic retinopathy.

BACKGROUND/AIMS: To compare the performance of generative versus retrieval-based chatbots in answering patient inquiries regarding age-related macular degeneration (AMD) and diabetic retinopathy (DR). METHODS: We evaluated four chatbots: generative models (ChatGPT-4, ChatGPT-3.5 and Google Bard) and a retrieval-based model (OcularBERT) in a cross-sectional study. Their response accuracy to 45 questions (15 AMD, 15 DR and 15 others) was evaluated and compared. Three masked retinal specialists graded the responses using a three-point Likert scale: either 2 (good, error-free), 1 (borderline) or 0 (poor with significant inaccuracies). The scores were aggregated, ranging from 0 to 6. Based on majority consensus among the graders, the responses were also classified as 'Good', 'Borderline' or 'Poor' quality. RESULTS: Overall, ChatGPT-4 and ChatGPT-3.5 outperformed the other chatbots, both achieving median scores (IQR) of 6 (1), compared with 4.5 (2) in Google Bard, and 2 (1) in OcularBERT (all p ≤8.4×10-3). Based on the consensus approach, 83.3% of ChatGPT-4's responses and 86.7% of ChatGPT-3.5's were rated as 'Good', surpassing Google Bard (50%) and OcularBERT (10%) (all p ≤1.4×10-2). ChatGPT-4 and ChatGPT-3.5 had no 'Poor' rated responses. Google Bard produced 6.7% Poor responses, and OcularBERT produced 20%. Across question types, ChatGPT-4 outperformed Google Bard only for AMD, and ChatGPT-3.5 outperformed Google Bard for DR and others. CONCLUSION: ChatGPT-4 and ChatGPT-3.5 demonstrated superior performance, followed by Google Bard and OcularBERT. Generative chatbots are potentially capable of answering domain-specific questions outside their original training. Further validation studies are still required prior to real-world implementation.

Management of corneal neovascularization: Current and emerging therapeutic approaches.

Corneal neovascularization (CoNV) is a sight-threatening condition affecting an estimated 1.4 million people per year, and the incidence is expected to rise. It is a complication of corneal pathological diseases such as infective keratitis, chemical burn, corneal limbal stem cell deficiency, mechanical trauma, and immunological rejection after keratoplasties. CoNV occurs due to a disequilibrium in proangiogenic and antiangiogenic mediators, involving a complex system of molecular interactions. Treatment of CoNV is challenging, and no therapy thus far has been curative. Anti-inflammatory agents such as corticosteroids are the mainstay of treatment due to their accessibility and well-studied safety profile. However, they have limited effectiveness and are unable to regress more mature neovascularization. With the advent of advanced imaging modalities and an expanding understanding of its pathogenesis, contemporary treatments targeting a wide array of molecular mechanisms and surgical options are gaining traction. This review aims to summarize evidence regarding conventional and emerging therapeutic options for CoNV.

Popular large language model chatbots' accuracy, comprehensiveness, and self-awareness in answering ocular symptom queries.

In light of growing interest in using emerging large language models (LLMs) for self-diagnosis, we systematically assessed the performance of ChatGPT-3.5, ChatGPT-4.0, and Google Bard in delivering proficient responses to 37 common inquiries regarding ocular symptoms. Responses were masked, randomly shuffled, and then graded by three consultant-level ophthalmologists for accuracy (poor, borderline, good) and comprehensiveness. Additionally, we evaluated the self-awareness capabilities (ability to self-check and self-correct) of the LLM-Chatbots. 89.2% of ChatGPT-4.0 responses were 'good'-rated, outperforming ChatGPT-3.5 (59.5%) and Google Bard (40.5%) significantly (all p < 0.001). All three LLM-Chatbots showed optimal mean comprehensiveness scores as well (ranging from 4.6 to 4.7 out of 5). However, they exhibited subpar to moderate self-awareness capabilities. Our study underscores the potential of ChatGPT-4.0 in delivering accurate and comprehensive responses to ocular symptom inquiries. Future rigorous validation of their performance is crucial to ensure their reliability and appropriateness for actual clinical use.

Analysis of the pachychoroid phenotype in an Asian population: methodology and baseline study population characteristics.

AIM: To describe the clinical characteristics in a cohort of patients with the pachychoroid phenotype and to evaluate the association of ocular and systemic factors with type of complications observed. METHODS: We report baseline findings from a prospective observational study which recruited subjects with subfoveal choroidal thickness (SFCT) of ≥300 µm on spectral-domain optical coherence tomography (OCT). Multimodal imaging was used to classify eyes as uncomplicated pachychoroid (UP) or pachychoroid disease with pachychoroid pigment epitheliopathy (PPE), central serous chorioretinopathy (CSC) or pachychoroid neovasculopathy (PNV) subtypes. RESULTS: Among 181 eyes of 109 participants (mean age 60.6 years, 33 (30.3%) female, 95 (7.2%) Chinese), 38 eyes (21.0%) had UP. Of 143 eyes (79.0%) with pachychoroid disease, 82 (45.3%), 41 (22.7%) and 20 (11.0%) had PPE, CSC and PNV, respectively. Addition of autofluorescence and OCT angiography to structural OCT led to reclassification of 31 eyes to a more severe category. Systemic and ocular factors evaluated, including SFCT, were not associated with disease severity. Comparison of PPE, CSC and PNV eyes showed no significant difference in OCT features of retinal pigment epithelial (RPE) dysfunction, but disruption of the ellipsoid zone (PPE 30.5% vs CSC 70.7% vs PNV 60%, p<0.001) and thinning of inner nuclear/inner plexiform layers (PPE 7.3% vs CSC 36.6% vs PNV 35%, p<0.001) were more frequent in CSC and PNV eyes. CONCLUSIONS: These cross-sectional associations suggest pachychoroid disease manifestations may reflect progressive decompensation from the choroid to the RPE then retinal layers. Planned follow-up of this cohort will be beneficial in clarifying the natural history of the pachychoroid phenotype.

Treat and extend regimen for diabetic macular oedema-a systematic review and meta-analysis.

PURPOSE: Various treatment regimens are currently practiced in the treatment of CI-DMO (centre-involving diabetic macular oedema). In recent years, there has been a growing body of evidence supporting a treat and extend (T&E) regimen for DMO which offers the promise of comparable visual and anatomical outcomes while reducing injection burden. This meta-analysis was hence performed to evaluate the aforementioned outcomes in the treatment of DMO. Ten studies met the inclusion criteria. METHODS: A search of PubMed, MEDLINE, Current Contents, and Cochrane Central Register of Controlled Trials (CENTRAL) databases was performed. We employed the terms 'treat AND extend AND (diabetic AND macular AND edema OR oedema)' to ensure a comprehensive search. The search workflow adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses. RESULTS: The pooled analysis of the mean number of injections in 1 year for T&E-aflibercept (AFL), T&E-ranibizumab (RBZ) and collectively was 9.1 (95% CI: 7.63-10.63), 10.0 (95% CI: 9.55-10.47) and 9.6 (95% CI: 8.62-10.49), respectively. Improvements in vision at 1 year for T&E-AFL, T&E-RBZ and collectively were 6.26 (95% CI: 3.24-9.29), 7.14 (95% CI: 4.76-9.52) and 7.08 (95% CI: 5.32-8.84) letters, respectively. The improvements in central subfield thickness at 1 year for T&E-AFL, T&E-RBZ and collectively were 131.94 (95% CI: 100.29-163.60), 108.64 (95% CI: 82.82-134.46) and 121.32 (95% CI: 102.89-139.75) microns, respectively. CONCLUSION: The meta-analysis of T&E for DMO did not show a clear advantage in reducing the number of injections compared to landmark clinical trials with pro-re-nata (PRN) treatment regimens in the first year of treatment with limited gains in visual and anatomical outcomes. However, the T&E approach offers the potential for fewer patient visits, thereby reducing treatment burden. Longer term studies on T&E with a standardised protocol would be required to assess the longevity of the vision gain in the first year despite a likely reduced treatment burden compared to the PRN trials.

Systemic Factors Associated with Treatment Response in Diabetic Macular Edema.

PURPOSE: To identify systemic factors that may influence the response to anti-VEGF therapy in patients with diabetic macular edema (DME). METHODS: 35 patients undergoing anti-VEGF injections for centre-involving DME were studied in this prospective observational study. The primary outcome was change in macular thickness one month after treatment, measured using spectral-domain optical coherence tomography (OCT). At baseline, information on various systemic factors was collected including glycosylated hemoglobin (HbA1c), serum VEGF levels, lipid profile and markers of renal function, and blood pressure. Thirty-three of the 35 patients were included in this study. Nonparametric statistical tests were used for the analysis of the data in view of the nonnormal distribution of the outcome variables. Multivariate analysis was performed using logistic regression. Stata 12.1 software was used for the analysis. Main Outcome Measures. Reduction in macular central subfield thickness (on spectral-domain OCT) and change in logMAR visual acuity at one month after injection. RESULTS: Lower HbA1c levels (7% or less) were significantly associated with greater reduction in central macular subfield thickness at one month after injection of bevacizumab or ranibizumab on both univariate analysis (p=0.012) and multivariate analysis (p=0.012) and multivariate analysis (. CONCLUSIONS: Better glycemic control is associated with a greater reduction in central macular thickness after the first injection of bevacizumab or ranibizumab in diabetic macular edema. Patients with high levels of HbA1c and poor response to anti-VEGF may benefit from strict control of their blood glucose.

Ophthalmic involvement in nasopharyngeal carcinoma.

This article desribes the ophthalmic involvement in patients with nasopharyngeal carcinoma (NPC) pre- and post-treatment. We retrospectively reviewed data for 354 consecutive patients diagnosed with or treated for NPC at a single tertiary centre between April 2007 and July 2015. We identified 27 (7.6%) patients with ophthalmic involvement due to NPC or its treatment. Symptomatic orbital invasion by tumor occurred in 13 of 27 patients (48.1%). The mean age of diagnosis in these patients was 54 years, and 8 (61.5%) had no prior diagnosis of NPC. Ocular signs, but no orbital invasion, was present in 8 patients (29.6%). Incidental orbital involvement was detected on imaging in 2 patients (7.4%). Radiotherapy-related ocular complications affected 4 patients (14.8%). Ophthalmic symptoms from orbital invasion can be the initial presentation of NPC. This disease tends to affect a younger group of patients and early recognition is important to minimize morbidity and mortality.

Thyroid cytology-nuclear versus architectural atypia within the "Atypia of undetermined significance/follicular lesion of undetermined significance" Bethesda category have significantly different rates of malignancy.

BACKGROUND: The Bethesda System for Reporting Thyroid Cytopathology is the most widely used classification system for the reporting of thyroid fine-needle aspiration cytology (FNAC) specimens. However, the "atypical" category ("atypia of undetermined significance" [AUS] or "follicular lesion of undetermined significance" [FLUS]) continues to cause diagnostic and therapeutic dilemmas. The objectives of this study were to describe the differential malignancy rates of FNACs diagnosed as AUS/FLUS based on nuclear or architectural atypia and to assess the significance of demographic and ultrasonographic features in predicting malignancy in this category. METHODS: A retrospective review was performed of all thyroid FNACs between 2008 and 2014 that were diagnosed as AUS/FLUS at a tertiary referral center in Singapore. Patient demographics, preoperative ultrasonographic features, and follow-up data were collected and correlated with the final histopathologic diagnosis in resected cases. RESULTS: In total, 309 thyroid nodules were diagnosed as AUS/FLUS, and 137 (44%) were surgically excised. Final histology yielded 37 (27%) malignancies. The malignancy rate for nodules that featured nuclear atypia was significantly higher at 36.8% than the rate for nodules that had only architectural atypia at 14.7% (P < .01). After up to 3 repeat FNACs, 67.1% of cases had a more definitive diagnosis. The only predictive sonographic finding for malignancy was irregular margins (P < .01). CONCLUSIONS: The disparity between malignancy risks within the Bethesda "atypical" category suggests that cytologic (nuclear) atypia is significantly more predictive of malignancy than architectural atypia. This supports the substratification of patients according to risk and a corresponding management approach within this category. A sonographic finding of irregular margins is also predictive for malignancy. Cancer Cytopathol 2017;125:245-256. © 2016 American Cancer Society.