A Multimodal Classification Architecture for the Severity Diagnosis of Glaucoma Based on Deep Learning

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma is an optic neuropathy causing irreversible visual field defects.
• Early and accurate diagnosis of glaucoma severity is critical for effective management and prevention of vision loss.
• Current diagnostic methods may benefit from enhanced computational approaches.

Purpose of the Study:

• To propose a novel multimodal deep learning classification architecture for glaucoma severity diagnosis.
• To integrate high-resolution visual field images and fundus images for improved diagnostic performance.
• To overcome limitations of traditional convolutional neural networks in glaucoma classification.

Main Methods:

• A multimodal classification architecture utilizing deep learning was developed.
• Preprocessing involved reconstructing grayscale visual field images to higher resolution.
• Multimodal fusion integrated processed visual field images with fundus images.
• A novel classifier replaced the standard convolutional neural network classifier.

Main Results:

• The proposed architecture demonstrated superior performance in glaucoma severity diagnosis.
• Enhanced feature extraction from high-resolution visual field images contributed to improved accuracy.
• Multimodal data integration provided a more comprehensive basis for classification.
• The model was validated on clinical datasets from the First Affiliated Hospital of Kunming Medical University.

Conclusions:

• The developed deep learning architecture offers a promising tool for accurate glaucoma severity diagnosis.
• Multimodal data fusion and enhanced image preprocessing are key to improving diagnostic capabilities.
• This approach has the potential to aid clinicians in preventing further vision loss in glaucoma patients.