LARGE UVEAL MELANOMA (≥10 MM THICKNESS): Clinical Features and Millimeter-by-Millimeter Risk of Metastasis in 1311 Cases. The 2018 Albert E. Finley Lecture.

PURPOSE: To analyze the clinical features and rate of metastatic disease in eyes with large (≥10 mm thickness) uveal melanoma. DESIGN: Retrospective noncomparative case series. PARTICIPANTS: There were 1,311 consecutive patients. METHODS: Retrospective medical chart review. MAIN OUTCOME MEASURES: Clinical features and rate of metastatic melanoma. RESULTS: Of 1,311 patients with large melanoma, the mean age was 59 years (median 60, range 6-98 years) and 95% were white. Mean tumor basal dimension was 17 mm (median 17, range 7-25 mm), and mean tumor thickness was 12 mm (median 12, range 10-24 mm). Mean distance to the foveola was 6 mm (median 6, range 0-19 mm) and to optic nerve was 6 mm (median 5, range 0-19 mm). Of all eyes, using Kaplan-Meier analysis, metastasis occurred in 11, 30, 45, and 52% at 1, 3, 5, and 7 years, respectively. According to tumor thickness (10.0-11.0, 11.1-12.0, 12.1-13.0, 13.1-14.0, 14.1-15.0, 15.1-16.0, and >16.0 mm), metastasis at 1 year was found in 7, 12, 13, 15, 18, 22, and 20%; metastasis at 3 years was 24, 27, 37, 35, 51, 69, and 57%; metastasis at 5 years was 38%, 42%, 56%, 48%, 61%, not available, and 66%; and metastasis at 7 years was 47%, 47%, 61%, 57%, 61%, not available, and 66%. Clinical features associated with fewer metastatic events included Bruch membrane rupture (7-year metastasis at 48%, P = 0.018) and macular location (7-year metastasis at 32%, P = 0.014), whereas those with worse outcome included extraocular extension (7-year metastasis at 79%, P < 0.001). There was no significant difference in rate of melanoma-related metastasis for patients treated with plaque radiotherapy versus enucleation. CONCLUSION: Large uveal melanoma demonstrates 7-year rate of metastasis at 52%, with generalized increasing risk per 1-mm or 2-mm thickness increments. Extraocular extension was associated with greater metastatic rate, whereas Bruch membrane rupture and macular location demonstrated lower rate.

Assessment of a Large Language Model's Responses to Questions and Cases About Glaucoma and Retina Management.

Importance: Large language models (LLMs) are revolutionizing medical diagnosis and treatment, offering unprecedented accuracy and ease surpassing conventional search engines. Their integration into medical assistance programs will become pivotal for ophthalmologists as an adjunct for practicing evidence-based medicine. Therefore, the diagnostic and treatment accuracy of LLM-generated responses compared with fellowship-trained ophthalmologists can help assess their accuracy and validate their potential utility in ophthalmic subspecialties. Objective: To compare the diagnostic accuracy and comprehensiveness of responses from an LLM chatbot with those of fellowship-trained glaucoma and retina specialists on ophthalmological questions and real patient case management. Design, Setting, and Participants: This comparative cross-sectional study recruited 15 participants aged 31 to 67 years, including 12 attending physicians and 3 senior trainees, from eye clinics affiliated with the Department of Ophthalmology at Icahn School of Medicine at Mount Sinai, New York, New York. Glaucoma and retina questions (10 of each type) were randomly selected from the American Academy of Ophthalmology's commonly asked questions Ask an Ophthalmologist. Deidentified glaucoma and retinal cases (10 of each type) were randomly selected from ophthalmology patients seen at Icahn School of Medicine at Mount Sinai-affiliated clinics. The LLM used was GPT-4 (version dated May 12, 2023). Data were collected from June to August 2023. Main Outcomes and Measures: Responses were assessed via a Likert scale for medical accuracy and completeness. Statistical analysis involved the Mann-Whitney U test and the Kruskal-Wallis test, followed by pairwise comparison. Results: The combined question-case mean rank for accuracy was 506.2 for the LLM chatbot and 403.4 for glaucoma specialists (n = 831; Mann-Whitney U = 27976.5; P < .001), and the mean rank for completeness was 528.3 and 398.7, respectively (n = 828; Mann-Whitney U = 25218.5; P < .001). The mean rank for accuracy was 235.3 for the LLM chatbot and 216.1 for retina specialists (n = 440; Mann-Whitney U = 15518.0; P = .17), and the mean rank for completeness was 258.3 and 208.7, respectively (n = 439; Mann-Whitney U = 13123.5; P = .005). The Dunn test revealed a significant difference between all pairwise comparisons, except specialist vs trainee in rating chatbot completeness. The overall pairwise comparisons showed that both trainees and specialists rated the chatbot's accuracy and completeness more favorably than those of their specialist counterparts, with specialists noting a significant difference in the chatbot's accuracy (z = 3.23; P = .007) and completeness (z = 5.86; P < .001). Conclusions and Relevance: This study accentuates the comparative proficiency of LLM chatbots in diagnostic accuracy and completeness compared with fellowship-trained ophthalmologists in various clinical scenarios. The LLM chatbot outperformed glaucoma specialists and matched retina specialists in diagnostic and treatment accuracy, substantiating its role as a promising diagnostic adjunct in ophthalmology.

Deep Learning for Localized Detection of Optic Disc Hemorrhages.

PURPOSE: To develop an automated deep learning system for detecting the presence and location of disc hemorrhages in optic disc photographs. DESIGN: Development and testing of a deep learning algorithm. METHODS: Optic disc photos (597 images with at least 1 disc hemorrhage and 1075 images without any disc hemorrhage from 1562 eyes) from 5 institutions were classified by expert graders based on the presence or absence of disc hemorrhage. The images were split into training (n = 1340), validation (n = 167), and test (n = 165) datasets. Two state-of-the-art deep learning algorithms based on either object-level detection or image-level classification were trained on the dataset. These models were compared to one another and against 2 independent glaucoma specialists. We evaluated model performance by the area under the receiver operating characteristic curve (AUC). AUCs were compared with the Hanley-McNeil method. RESULTS: The object detection model achieved an AUC of 0.936 (95% CI = 0.857-0.964) across all held-out images (n = 165 photographs), which was significantly superior to the image classification model (AUC = 0.845, 95% CI = 0.740-0.912; P = .006). At an operating point selected for high specificity, the model achieved a specificity of 94.3% and a sensitivity of 70.0%, which was statistically indistinguishable from an expert clinician (P = .7). At an operating point selected for high sensitivity, the model achieves a sensitivity of 96.7% and a specificity of 73.3%. CONCLUSIONS: An autonomous object detection model is superior to an image classification model for detecting disc hemorrhages, and performed comparably to 2 clinicians.

Visual Outcome and Millimeter Incremental Risk of Metastasis in 1780 Patients With Small Choroidal Melanoma Managed by Plaque Radiotherapy.

Importance: Early detection of choroidal melanoma at a small tumor size is emphasized in the literature. However, there is little published information on the specific risks of plaque-irradiated small choroidal melanoma on visual acuity and metastasis. Objective: To analyze outcomes of plaque radiotherapy for small choroidal melanoma 3 mm in thickness or less. Design, Setting, and Participants: This retrospective noncomparative series at a tertiary referral center included 1780 consecutive patients who had received plaque radiotherapy treatment for small choroidal melanoma. Main Outcomes and Measures: Visual acuity outcomes and melanoma-associated metastasis, assessed by Kaplan-Meier analyses. Results: The mean (SD) patient age at melanoma diagnosis was 58 (14) years. Of 1780 patients, 908 were female (51.0%), and 1752 were white (98.4%). Visual acuity was 20/40 OU or better in 1276 of the patients (71.7%), and the mean (SD) visual acuity was 20/40 (20/50) OU (median, 20/30; range, 20/20 to counting fingers). The mean (SD) tumor basal dimension was 8.8 (2.9) mm (median, 8.0 mm; range, 2.0-20.0 mm) and mean (SD) tumor thickness was 2.6 (0.5) mm (median, 2.7; range, 0.2-3.4 mm). Mean (SD) distance to the foveola was 3.4 (3.9) mm and to the optic disc was 3.7 (3.7) mm. The Kaplan-Meier rate of visual acuity loss (≥3 Snellen lines) was 9.5% (95% CI, 8.2%-11.0%) at 1 year, 39.2% (95% CI, 36.5%-42.0%) at 5 years, and 48.9% (95% CI, 45.6%-52.3%) at 10 years, whereas poor visual acuity (≤20/200) was 7.1% (95% CI, 5.9%-8.4%) at 1 year, 38.2% (95% CI, 35.5%-41.1%) at 5 years, and 53.5% (95% CI, 50.1%-57.1%) at 10 years. Regarding melanoma-associated metastasis, the rate was 0.2% (95% CI, 0.09%-0.6%) at 1 year, 4.5% (95% CI, 3.4%-5.9%) at 5 years, and 8.8% (95% CI, 6.9%-11.1%) at 10 years. Using 1.0-mm thickness increments, the 10-year risk for metastasis was 25.0% (95% CI, 3.9%-87.2%) at 0-mm to 1.0-mm thickness, 5.9% (95% CI, 2.5%-13.5%) at 1.1-mm to 2.0-mm thickness, 8.1% (95% CI, 5.9%-11.0%) at 2.1-mm to 3.0-mm thickness, and 13.4% (95% CI, 8.7%-20.4%) at thicknesses greater than 3.0 mm. The greater relative risk (RR) for metastasis in thinnest tumors was 1.83 (95% CI, 1.09-3.07), which likely represented more aggressive diffuse (flat) melanoma. By multivariable analysis, clinical features predictive of melanoma-associated metastasis included increasing patient age (RR, 1.32 [95% CI, 1.07-1.63] per decade; P = .01), tumor diameter (RR, 1.15 [95% CI, 1.06-1.24] per mm; P < .001), tumor thickness (RR, 2.22 [95% CI, 1.22-4.05] per mm; P = .01), photopsia symptoms (RR, 2.45 [95% CI, 1.35-4.43]; P = .003), and prior treatment before plaque radiotherapy (RR, 3.31 [95% CI, 1.31-8.33]; P = .01). Conclusions and Relevance: This retrospective study suggests that small choroidal melanoma treated with plaque radiotherapy has a 10-year risk for visual acuity loss of 48.9% (95% CI, 45.6%-52.3%) and a 10-risk of systemic metastasis of 8.8% (95% CI, 6.9%-11.1%). In this analysis, each millimeter of increasing thickness and diameter contributed risk for metastatic disease.