Novel Deep Learning Model for Glaucoma Detection Using Fusion of Fundus and Optical Coherence Tomography Images

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma is a primary cause of irreversible blindness globally.
• Early detection is crucial for preventing vision loss.

Purpose of the Study:

• To develop and evaluate a novel deep learning model for automated glaucoma detection.
• To investigate the efficacy of fusing fundus photographs and optical coherence tomography (OCT) scans for improved diagnostic performance.

Main Methods:

• A deep learning approach using paired fundus and OCT images from the same eye.
• Development of separate convolutional neural network (CNN) models for each modality and a fusion model.
• Training and validation on a private clinical dataset from Bangladesh Eye Hospital and Institute Ltd.

Main Results:

• The fused model achieved 92% accuracy and an AUC of 0.95.
• This represents an absolute improvement of 6% and 8% over fundus-only and OCT-only models, respectively.
• Statistical analysis confirmed the significant performance gains of the fusion model (p < 0.003).

Conclusions:

• The novel deep learning fusion model demonstrates superior performance in glaucoma detection compared to single-modality approaches.
• This approach holds significant potential for early and accurate diagnosis of glaucoma, aiding in vision preservation.
• This study represents the first use of a deep learning model fusing paired fundus and OCT images for glaucoma detection.