Predicting intraocular pressure using neural networks: Incorporating eye fundus images and clinical data from PAPILA dataset

This study explores the potential of deep learning models to predict intraocular pressure (IOP) using a combination of retinal fundus images and clinical ophthalmology data. Utilizing the PAPILA dataset, which includes images categorized as normal or glaucoma, we trained a neural network model on 70% of the data, reserving 15% each for validation and testing. Our results indicate that the model achieved a Mean Absolute Error (MAE) of 2.52, suggesting an average deviation of 2.52 units from the actual IOP values. The model's R-squared value was 0.10, reflecting that approximately 10.24% of the variance in IOP was accounted for by the predictors used. These outcomes underscore the challenges and nuances of predicting IOP solely from ocular images and emphasize the importance of incorporating clinical data for more accurate predictions. This approach could be particularly beneficial in regions with limited access to ophthalmic healthcare, providing a cost-effective tool for early screening and management of glaucoma.