Deep Kernel Learning for Stratifying Glaucoma Trajectories

2026-05-01

AI Summary

This study addresses the challenge of identifying glaucoma progression risk from sparse and irregular electronic health records by proposing a deep kernel learning architecture that integrates Transformer networks with clinical BERT embeddings. The model leverages Gaussian processes to capture multimodal longitudinal data, enabling precise characterization of individual disease trajectories. By disentangling current disease severity from future progression trends, the approach overcomes the limitations of conventional static assessments and successfully identifies a clinically critical subgroup of patients who exhibit relatively preserved visual function yet harbor high progression risk. Experimental results reveal three distinct clinical subgroups, with the high-risk cohort demonstrating sustained deterioration despite favorable baseline vision, thereby validating the model’s sensitivity to dynamic progression patterns and its potential clinical utility.

Abstract

Effectively stratifying patient risk in chronic diseases like glaucoma is a major clinical challenge. Clinicians need tools to identify patients at high risk of progression from sparse and irregularly-sampled electronic health records (EHRs). We propose a novel deep kernel learning (DKL) architecture that leverages a Gaussian Process (GP) backend. The GP's kernel is defined by a transformer-based feature extractor applied to clinical-BERT embeddings to model glaucoma patient trajectories from multimodal EHR data. Our method successfully identifies three clinically distinct patient subgroups. Crucially, the model learns to decouple disease progression from current severity, identifying a high-risk group with a worsening trajectory despite having better average visual acuity than a second, stably poor group. This reveals that the model learns to identify progression risk rather than just the current disease state. This ability to stratify patients based on their risk trajectory progression offers a powerful tool for clinical decision support, enabling targeted interventions for high-risk individuals and improving the management of glaucoma care.

Problem

• glaucoma
• risk stratification
• disease progression
• electronic health records
• patient trajectories

Innovation

• Deep Kernel Learning
• Gaussian Process
• Transformer-based Feature Extractor
• clinical-BERT
• Disease Trajectory Stratification