An Artificial Intelligence-Based Prognostic Model for Prediction of Functional Glaucoma Progression From Clinical and...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma diagnosis and progression prediction require integrating diverse data sources.
• Current methods for forecasting glaucoma progression have limitations.

Purpose of the Study:

• To develop and evaluate a deep learning (DL) model for predicting functional glaucoma progression.
• To compare the DL model's prognostic accuracy against that of clinical experts.

Main Methods:

• A retrospective cohort of 1,599 eyes with glaucoma was analyzed.
• A convolutional neural network was trained on clinical data, disc photographs, and OCT-derived measurements.
• The DL model's predictions were compared to those of two glaucoma specialists.

Main Results:

• The DL model achieved an AUC of 0.839 and accuracy of 76.0% in predicting glaucoma progression.
• The model significantly outperformed clinical graders (p<0.001).
• Performance was consistent on a validation cohort and for predicting rapid progression.

Conclusions:

• A novel DL model effectively integrates multimodal data for glaucoma progression prediction.
• The model provides clinically relevant prognostic information, surpassing human expert accuracy.
• This tool holds potential for improving glaucoma diagnostics and patient management.