Photo Credit: aykuterd

The following is a summary of “Machine-learning algorithms for the identification of visual field loss associated with the antiseizure medication vigabatrin—a proof of concept,” published in the March 2025 issue of British Journal of Ophthalmology by Wild et al. 

Researchers conducted a retrospective study to develop and evaluate machine-learning pattern recognition algorithms for identifying vigabatrin-associated visual field loss (VAVFL) objectively.  

They followed the European Medicines Agency-approved protocol for detecting VAVFL using the 3 Zone Age Corrected Full Field 135 Screening Test (FF135) and the Central C30-2 Threshold Test (C30-2T) with the Humphrey Field Analyzer. Each algorithm assessed the similarity of measured fields in both eyes to modelled VAVFL reference patterns, adjusted for severity, derived from a prior case series of 123 adults. Symmetrisation, a signal-to-noise enhancement method based on between-eye mirror image symmetry, was optionally incorporated. The algorithms’ effectiveness in identifying VAVFL was tested on 89 individuals across 6 diagnostic categories, including homonymous and glaucomatous losses.  

The results showed that the algorithms demonstrated strong agreement with the ‘gold standard’ clinical interpretation, with sensitivity and specificity values of 95.7% (22/23) and 100% (30/30) for the FF135, and 94.4% (17/18) and 94.1% (48/51) for the C30-2T. Symmetrisation proved beneficial in detecting VAVFL, particularly when perimetric learning or fatigue affected one eye’s outcome and in cases with concurrent homonymous loss.  

Investigators concluded that a directly interpretable machine-learning outcome accurately identified VAVFL and could aid patient management in community (neuro-)ophthalmology. 

Source: bjo.bmj.com/content/early/2025/03/18/bjo-2024-325804