High-Accuracy Digitization of Humphrey Visual Field Reports Using Convolutional Neural Networks

Shian-Sen Shie ¹, Wei-Wen Su ²

⁰Department of Internal Medicine, Division of Infectious Diseases, Chang Gung Memorial Hostpital, Linkou Branch, Taoyuan Taiwan; and College of Medicine, Chang Gung University, Taoyuan, Taiwan.

Translational Vision Science & Technology +

|

May 5, 2025

View abstract on PubMed

Summary

This study introduces an AI framework using convolutional neural networks (CNNs) to digitize visual field (VF) reports for glaucoma patients. The AI model achieves high accuracy, improving data accessibility and clinical workflows for better patient outcomes.

Area Of Science

Ophthalmology
Artificial Intelligence
Medical Informatics

Background

Glaucoma is a primary cause of irreversible blindness globally.
Accurate visual field (VF) assessments are crucial for glaucoma diagnosis and management.
Digitizing VF reports is essential for data utilization in clinical practice.

Purpose Of The Study

To develop an efficient and accurate method for digitizing visual field reports.
To address challenges in data accessibility for clinical evaluations.
To enhance the utility of data from visual field assessments.

Main Methods

A lightweight convolutional neural network (CNN) framework was developed.
A dataset of 15,000 visual field reports spanning a decade was utilized.
Portable document format (PDF) files were preprocessed, and data was standardized into 48x48 pixel images, incorporating diverse font types for generalization.

Main Results

The CNN model achieved 100% accuracy for numerical data extraction and over 98.6% for metadata recognition.
Post-processing with keyword mapping improved metadata reliability by correcting visually similar character errors.
The AI method significantly reduced processing time compared to manual data entry while maintaining high accuracy.

Conclusions

The AI-driven digitization method effectively interprets visual field images, offering a reliable solution for complex report digitization.
This framework enhances clinical workflows and facilitates better interpretation of visual field data.
The study demonstrates the potential of AI in advancing glaucoma care through streamlined processes and improved patient outcome analysis.