Glaucoma and Ocular Hypertension
Conditions
Brief summary
The study titled "Reliability of Artificial Intelligence in Glaucoma Severity Staging Using the Brusini GSS2 System: A Longitudinal Comparative Study against Expert Assessment" aims to evaluate the diagnostic reliability and clinical utility of an Artificial Intelligence (AI) system in staging glaucoma severity using the Brusini Glaucoma Staging System 2 (GSS2) compared to expert ophthalmologist assessment. The study protocol involves the longitudinal analysis of visual fields (VFs) collected at baseline (T0) and at a 1-year follow-up (T1). Statistical methodology will utilize quadratic weighted Cohen's kappa to assess diagnostic agreement between the AI system and expert assessment, Wilcoxon signed-rank tests to evaluate potential systematic bias, and Spearman's rank correlation to investigate the relationship between longitudinal progression and changes in the clinical Visual Field Index.
Interventions
DIAGNOSTIC_TESTArtificial Intelligence GSS2 Staging
An artificial intelligence system used to automatically stage glaucoma severity from standard automated visual field tests according to the Brusini Glaucoma Staging System 2 (GSS2). AI classifications were compared with those of an expert ophthalmologist at baseline and 1-year follow-up. The AI was used for assessment only and did not influence patient management or treatment.
Sponsors
University of Salerno
Study design
Observational model
COHORT
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Inclusion criteria
* Adults (≥18 years of age). * Diagnosis of glaucoma or suspected glaucoma undergoing routine clinical follow-up. * Availability of reliable standard automated visual field (VF) examinations at baseline and approximately 1-year follow-up. * Visual field tests suitable for staging using the Brusini Glaucoma Staging System 2 (GSS2). * Availability of all clinical data required for study analysis.
Exclusion criteria
* Unreliable visual field examinations (e.g., excessive fixation losses, false-positive or false-negative responses according to institutional reliability criteria). * Visual field defects attributable to ocular or neurological diseases other than glaucoma. * Missing baseline or follow-up visual field examinations. * Incomplete clinical data preventing GSS2 staging or statistical analysis. * Previous ocular conditions or interventions that could significantly affect visual field interpretation, if judged by the investigator.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Quadratic weighted Cohen's kappa coefficient for diagnostic agreement in glaucoma severity staging between AI and expert assessment | Baseline (T0) and 1-year follow-up (T1) | Glaucoma severity will be staged (from Stage 0 to Stage 5) using the Brusini Glaucoma Staging System 2 (GSS2) based on visual field parameters. Staging will be performed independently by the AI system and an expert ophthalmologist at each time point. Inter-rater agreement between the AI and expert staging will be quantified using the quadratic weighted Cohen's kappa ($\\kappa$), where values range from -1 (complete disagreement) to +1 (perfect agreement). |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Median difference in Brusini GSS2 staging scores between AI and expert assessment | Baseline (T0) and 1-year follow-up (T1) | Differences in Brusini Glaucoma Staging System 2 (GSS2) staging scores (ranging from Stage 0 to Stage 5) assigned by the AI system versus the expert ophthalmologist will be evaluated to detect systematic overestimation or underestimation of disease severity. Paired score differences (AI minus Expert) at each time point will be reported and statistically tested using the Wilcoxon signed-rank test. |
Countries
Italy
Outcome results
None listed