
Deleted Journal, Journal Year: 2024, Volume and Issue: unknown, P. 100092 - 100092
Published: Dec. 1, 2024
Language: Английский
Deleted Journal, Journal Year: 2024, Volume and Issue: unknown, P. 100092 - 100092
Published: Dec. 1, 2024
Language: Английский
Survey of Ophthalmology, Journal Year: 2024, Volume and Issue: unknown
Published: Sept. 1, 2024
Language: Английский
Citations
6Ophthalmology Glaucoma, Journal Year: 2025, Volume and Issue: unknown
Published: April 1, 2025
Language: Английский
Citations
0Vision, Journal Year: 2025, Volume and Issue: 9(2), P. 33 - 33
Published: April 11, 2025
This study assesses the accuracy and consistency of a commercially available large language model (LLM) in extracting interpreting sensitivity reliability data from entire visual field (VF) test reports for evaluation glaucomatous defects. Single-page anonymised VF 60 eyes subjects were analysed by an LLM (ChatGPT 4o) across four domains-test reliability, defect type, severity overall diagnosis. The main outcome measures extraction, interpretation defects diagnostic classification. displayed 100% extraction global metrics classifying reliability. It also demonstrated high (96.7%) diagnosing whether was consistent with healthy, suspect or eye. correctly defining type moderate (73.3%), which only partially improved when provided more defined region interest. causes incorrect mostly attributed to wrong location, particularly confusing superior inferior hemifields. Numerical/text-based notably image-based demonstrates potential limitations multimodal LLMs processing medical investigation such as reports.
Language: Английский
Citations
0Ophthalmic and Physiological Optics, Journal Year: 2024, Volume and Issue: 44(2), P. 426 - 441
Published: Jan. 16, 2024
Abstract Purpose To examine the diagnostic accuracy of performing two (frontloaded) versus one (clinical standard) visual field (VF) test per visit for detecting progression early glaucoma in data derived from clinical populations. Methods A computer simulation model was used to follow VFs 10,000 patients (derived cohorts: Heijl et al., Swedish cohort; and Chauhan Canadian Glaucoma Study [CGS]) over a 10‐year period identify whose mean deviation (MD) detected. Core (baseline MD rates) were extracted studies cohorts glaucoma, which modulated using SITA‐Faster variability characteristics previous work. Additional variables included follow‐up intervals (six‐monthly or yearly) rates perimetric loss any reason (0%, 15% 30%). The main outcome measures proportions progressors Results When cohort reviewed six‐monthly, frontloaded strategy detected more compared non‐frontloaded method up years 8, 9 10 0%, 30% conditions. time required detect 50% cases 1.0–1.5 less frontloading non‐frontloading. At 4 years, increased detection by 26.7%, 28.7% 32.4% conditions, respectively. Where both techniques progression, earlier (78.5%–81.5% 1.0–1.3 when six‐monthly; 81%–82.9% 1.2–2.1 yearly). Accordingly, these had severe scores review: 0.63–1.67 dB ‘saved’; yearly 1.10–2.87 dB). differences with higher loss. Similar tendencies noted applied CGS cohort. Conclusions Frontloaded distributions led progression.
Language: Английский
Citations
3Ophthalmology, Journal Year: 2023, Volume and Issue: 131(6), P. 658 - 666
Published: Dec. 16, 2023
Language: Английский
Citations
5Deleted Journal, Journal Year: 2024, Volume and Issue: unknown, P. 100092 - 100092
Published: Dec. 1, 2024
Language: Английский
Citations
1