Performance evaluation of various deep learning based models for effective glaucoma evaluation using optical...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma is a leading cause of irreversible blindness globally.
• Early detection and timely intervention are crucial for preventing vision loss.
• Optical coherence tomography (OCT) is a key imaging modality for diagnosing glaucomatous damage.

Purpose of the Study:

• To evaluate the efficacy of various ImageNet deep learning models for glaucoma prediction using OCT images.
• To identify the optimal deep learning model and optimizer for accurate glaucoma detection.
• To assess the potential of automated systems for glaucoma screening.

Main Methods:

• Eight ImageNet models were trained and fine-tuned on a private OCT dataset (4220 images).
• Model performance was evaluated using multiple efficiency metrics across three optimizers: Adam, RMSProp, and SGD.
• Top-performing models were further validated on the public Mendeley dataset.

Main Results:

• VGG16 model combined with the Root Mean Squared Propagation optimizer achieved 95.68% accuracy.
• Four best-performing models were shortlisted after initial evaluation on the private dataset.
• The study demonstrated the effectiveness of deep learning in differentiating normal and glaucomatous OCT scans.

Conclusions:

• ImageNet models offer a viable alternative for automated, computer-based glaucoma screening.
• The proposed deep learning system can accurately distinguish between normal and glaucomatous OCT images.
• This automated approach can aid in early diagnosis, reduce expert workload, and prevent vision loss.