Deep learning predicts glaucoma progression and spots fast vision loss
Deep learning of longitudinal visual fields predicts glaucoma progression rate and identifies fast progressors
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Glaucoma can cause blindness if it worsens too fast. The authors developed a deep learning model called GLAM that looks at sequences of vision tests and some clinical data to predict how quickly someone's glaucoma will get worse. Their method can identify people whose vision loss is rapid much more accurately than traditional ways. This could help doctors decide who needs urgent care to stop blindness.
What this means in practice
- •For eye care professionals: Predict glaucoma progression rates early from visual field test sequences to identify patients needing urgent intervention.
- •For medical ai developers: Develop AI tools that use only visual field data to monitor glaucoma progression without requiring additional patient information.$Commercial implications: Enables creation of standalone glaucoma progression prediction software tools for clinical use based on existing visual field tests.
Tested on one dataset.
Authors
Taiabur Rahman, Siddiqur Rahman, Muhammad Moniruzzaman, Ummay Kawsar, Sayedatunnessa Ratna, Shadman Siddique, Rafsan Siddique, Tausif Ahmad, Tahsin Ahmad, Golam Rabbani
Abstract
Glaucoma is the leading cause of irreversible blindness, and timely identification of fast progressors is essential to prevent disability. Current practice estimates progression by ordinary least-squares regression of mean deviation (MD) on time, requiring 6--10 visual field (VF) tests over several years to obtain a reliable slope. We present GLAM (Glaucoma Longitudinal Analysis Model), a deep learning framework that ingests longitudinal Humphrey 24-2 total deviation sequences with five clinical features and predicts MD and visual field index progression rates using attention-based fusion and aleatoric uncertainty. On the open-access University of Washington Humphrey Visual Field dataset (4,276 patient-eyes), GLAM achieved an MD-rate mean absolute error of 0.139 dB yr$^{-1}$ ($R^2 = 0.927$; 73.5% reduction over a ridge baseline) and an AUC of 0.990 for fast-progressor detection. VF-only deep learning can match multimodal pipelines for progression prognostication using routinely collected perimetry alone.