Glaucoma detection based on deep convolutional neural network.

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Glaucoma is a chronic, irreversible eye condition impacting vision and life quality.
• Early and accurate diagnosis is crucial for managing glaucoma and preventing vision loss.

Purpose of the Study:

• To develop and evaluate a deep learning (DL) model for automated glaucoma diagnosis.
• To assess the efficacy of a convolutional neural network (CNN) architecture in distinguishing glaucoma from non-glaucoma patterns.

Main Methods:

• A six-layer DL architecture comprising four convolutional and two fully-connected layers was designed.
• Dropout and data augmentation techniques were employed to enhance diagnostic performance.
• The model was trained and validated on the ORIGA and SCES datasets.

Main Results:

• The DL model achieved an area under the curve (AUC) of 0.831 on the ORIGA dataset.
• An AUC of 0.887 was recorded on the SCES dataset, outperforming existing state-of-the-art algorithms.
• The proposed method demonstrated significant potential for accurate glaucoma detection.

Conclusions:

• The developed DL architecture provides a robust and accurate method for automated glaucoma diagnosis.
• This AI-driven approach has the potential to aid clinicians in early glaucoma detection and management.
• Further research can explore integrating this model into clinical workflows for improved patient care.