Comprehensive Survey of OCT-Based Disorders Diagnosis: From Feature Extraction Methods to Robust Security Frameworks.

Area of Science:

• Ophthalmology and Medical Imaging
• Artificial Intelligence in Healthcare
• Computer Vision for Diagnostics

Background:

• Optical Coherence Tomography (OCT) is crucial for diagnosing retinal diseases like macular degeneration and diabetic retinopathy.
• Early detection and monitoring of optic nerve head changes are vital for conditions such as glaucoma.
• A comparative analysis of OCT-based feature extraction techniques for disease prediction has been lacking.

Purpose of the Study:

• To survey and comparatively analyze hand-crafted and deep learning feature extraction methods for ocular disease prediction using OCT.
• To evaluate the performance of different methods based on accuracy, sensitivity, specificity, and computational cost.
• To explore adversarial attacks on OCT diagnostic systems and propose robust defense strategies.

Main Methods:

• Comprehensive literature review of OCT-based ocular disease prediction techniques over the past 20 years.
• Evaluation of methods using metrics like accuracy, sensitivity, and specificity.
• Assessment of public OCT datasets and analysis of adversarial attack vulnerabilities.

Main Results:

• Identified key diseases including glaucoma, diabetic retinopathy, cataracts, amblyopia, and macular degeneration.
• Compared the efficacy and computational demands of various feature extraction approaches.
• Demonstrated the vulnerability of OCT diagnostic systems to adversarial attacks and proposed an effective defense strategy.

Conclusions:

• Combining classical and deep learning methods with robust defenses enhances the security and reliability of OCT diagnostics.
• The proposed defense strategy offers practical integration and superior performance.
• Findings provide guidance for future research and clinical implementation of OCT-based diagnostic tools.