AI Supports Glaucoma Surgical Planning

Predictive models identify high-risk patients and support earlier treatment decisions.

By
Kathryn Wighton
April 16, 2026
3 min

Conexiant

Overview

Artificial intelligence (AI) models demonstrate high diagnostic accuracy in glaucoma, achieving a pooled area under the curve (AUC) of 0.93. These models also predict visual field deterioration years in advance and assist in surgical decision-making, supporting individualized glaucoma care.

Background

Glaucoma is a progressive optic neuropathy that can lead to irreversible vision loss if not managed effectively. Early diagnosis and timely surgical intervention are critical to preserving visual function. Traditional diagnostic and prognostic methods have limitations in predicting disease progression and guiding treatment. AI applications, including deep learning and machine learning, have emerged as promising tools to enhance diagnostic accuracy and support clinical decision-making in glaucoma management.

Data Highlights

Metric	Value
Number of studies included	42
Studies central to surgical planning	13
Patient population	>10,000 patients
Image dataset size	>75,000 images
Diagnostic AUC range	0.91 to 0.95
Pooled diagnostic AUC (meta-analysis)	0.93
Highest AUC with multimodal models	0.95
Visual field deterioration prediction horizon	Up to 4 years

Key Findings

AI diagnostic models, mainly convolutional neural networks and deep learning systems, achieved AUC values between 0.91 and 0.95 for glaucoma detection.
Multimodal AI models integrating fundus photography, OCT, and visual field data outperformed single-modality systems, reaching an AUC of 0.95.
Meta-analysis of four diagnostic studies with over 64,000 images showed a pooled AUC of 0.93, confirming high diagnostic accuracy.
Machine learning models predicted visual field deterioration up to 4 years in advance, enabling early identification of fast progressors.
Risk stratification models using electronic health records and multimodal data predicted the need for surgical intervention and differentiated patients suitable for medical versus surgical management.
Deep-learning-based structural imaging facilitated segmentation of optic nerve head and lamina cribrosa, aiding presurgical assessment.

Clinical Implications

AI models can enhance glaucoma diagnosis and prognosis by providing accurate, early predictions of disease progression, which supports timely surgical planning. Integration of AI decision support systems into clinical workflows is feasible and may optimize referral pathways and surgical triaging. These advances promote personalized, data-driven glaucoma care to improve patient outcomes.

Conclusion

AI applications demonstrate transformative potential in glaucoma management by improving diagnostic accuracy and supporting individualized surgical decision-making. Continued validation and integration into clinical practice could mark a paradigm shift toward optimized, data-driven glaucoma care.