Predicting the central 10 degrees visual field in glaucoma by applying a deep learning algorithm to optical coherence...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Glaucoma diagnosis and monitoring rely on visual field (VF) testing and optical coherence tomography (OCT) imaging.
• Accurate prediction of central visual field defects is crucial for early glaucoma detection and management.
• Current methods may have limitations in precisely predicting fine visual field changes.

Purpose of the Study:

• To develop and validate a convolutional neural network (CNN) model for predicting the Humphrey Field Analyzer (HFA) 10-2 visual field test results in glaucoma patients.
• To utilize optical coherence tomography (OCT) derived macular retinal layer thicknesses as input for the CNN model.
• To enhance prediction accuracy by incorporating adjustments from Humphrey Field Analyzer (HFA) 24-2 test data.

Main Methods:

• A CNN model was trained using OCT images and macular retinal layer thicknesses from 558 eyes of glaucoma patients and 90 eyes of normal subjects.
• The model predicted Total Deviation (TD) values for the HFA 10-2 test points.
• Predicted TD values were subsequently corrected using TD values from the innermost four points of the HFA 24-2 test, applied to a testing dataset of 105 glaucoma patient eyes.

Main Results:

• Initial CNN model predictions for HFA 10-2 TD values yielded a mean absolute error between 9.4 and 9.5 dB.
• Correction using HFA 24-2 test data significantly reduced the mean absolute error to an average of 5.5 dB.
• The study demonstrates the feasibility of predicting HFA 10-2 results with improved accuracy through data integration and model refinement.

Conclusions:

• A CNN model trained on OCT data can effectively predict HFA 10-2 visual field results in glaucoma.
• Adjusting predictions with HFA 24-2 test data substantially improves the accuracy of visual field defect prediction.
• This approach offers a promising method for enhancing glaucoma assessment and patient monitoring.