A study published in Translational Vision Science & Technology describes a novel dual-level pattern tree model for classifying visual field (VF) loss that researchers say outperforms traditional methods in predicting future vision decline and correlating with glaucoma genetic risk scores. The advance may improve how clinicians forecast disease progression and link visual function to genetic risk.

Researchers analyzed over 113,000 visual fields from multiple large datasets, creating a two-tier hierarchical model of VF loss patterns:

  • Trunk patterns: broad categories of visual field loss

  • Branch patterns: finer subtypes nested under each trunk pattern

Using archetypal analysis, the team identified 17 trunk patterns and 169 branch subpatterns, capturing a richer spectrum of glaucomatous vision loss. When forecasting functional progression over 5 years, the trunk–branch model consistently outperformed trunk-only classification across multiple familiar clinical indicators, including:

  • Mean Deviation (MD)

  • Fast MD slope

  • Total Deviation pointwise slope

  • Visual Field Index slope

For example, the area under the receiver operating characteristic curve (AUC) for MD-fast was 0.84 with the dual-level model compared to 0.78 for traditional approaches — indicating better discrimination between stable and progressing eyes.

Beyond prognosis, the dual-level patterns also showed stronger associations with glaucoma polygenic risk scores (PRSs)—composite measures of genetic susceptibility. In multivariable regression analyses, adding the branch pattern detail significantly improved model fit compared to trunk-only classification. This suggests that specific visual field subtypes may reflect underlying genetic influences on glaucoma development, the researchers said.

Reference

Song, L., Shen, L. Q., Pasquale, L. R., Boland, et al. (2025). Dual-Level Pattern Tree for Visual Field Improves Glaucoma Progression and Polygenic Risk Prediction. Translational Vision Science & Technology, 14(12), Article 11. https://doi.org/10.1167/tvst.14.12.11