Deep Relation Transformer for Diagnosing Glaucoma With Optical Coherence Tomography and Visual Field Function.

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma is a leading cause of irreversible blindness, necessitating early detection and treatment.
• Optical Coherence Tomography (OCT) and Visual Field (VF) tests are complementary diagnostic tools, assessing optic nerve head structure and visual function, respectively.

Purpose of the Study:

• To propose a Deep Relation Transformer (DRT) model for glaucoma diagnosis by integrating OCT and VF data.
• To develop a novel deep reasoning mechanism to capture pairwise relations between OCT and VF information.
• To enhance diagnostic accuracy by leveraging complementary information from both imaging and functional tests.

Main Methods:

• A Deep Relation Transformer (DRT) model was developed, incorporating a novel deep reasoning mechanism.
• Three modules were designed: global relation, guided regional relation, and interaction transformer modules.
• A large dataset (ZOC-OCT&VF) with 1395 OCT-VF pairs was created for training and evaluation.

Main Results:

• The proposed DRT method achieved 88.3% accuracy in glaucoma diagnosis.
• The integrated approach significantly outperformed single-modal diagnostic methods.
• Experimental validation confirmed the effectiveness of the DRT model.

Conclusions:

• Combining OCT and VF data using the DRT model offers a powerful approach for accurate glaucoma diagnosis.
• The DRT model demonstrates superior performance compared to existing single-modality methods.
• This integrated approach holds promise for improving early detection and management of glaucoma to prevent vision loss.