RPE layer detection and baseline estimation using statistical methods and randomization for classification of AMD...

Area of Science:

• Ophthalmology and Medical Imaging
• Biomedical Engineering
• Computational Pathology

Background:

• Age-related macular degeneration (AMD) is a leading cause of vision loss in the elderly.
• Optical coherence tomography (OCT) provides high-resolution cross-sectional images of retinal layers.
• Accurate segmentation of the retinal pigment epithelium (RPE) layer is crucial for AMD diagnosis.

Purpose of the Study:

• To develop a method for extracting the RPE layer and baseline from spectral-domain OCT (SD-OCT) images.
• To quantify drusen height for classifying eyes as AMD or normal.
• To achieve accurate AMD classification without requiring prior training data.

Main Methods:

• Contrast enhancement and adaptive denoising for speckle reduction.
• Iterative pixel grouping and elimination for RPE layer segmentation.
• Randomization, polynomial fitting, and drusen removal for baseline estimation.
• Patient-wise classification based on drusen height thresholds.

Main Results:

• The method was validated on a dataset of 2130 images from 30 patients (15 AMD, 15 Normal).
• Achieved an overall accuracy of 96.66% with no false positives.
• Demonstrated superior accuracy and baseline estimation compared to existing methods.

Conclusions:

• The proposed statistical approach offers a robust, training-free method for AMD classification.
• Modified algorithms enhance RPE detection robustness, even with significant drusen.
• The novel baseline estimation technique provides precise retinal image measurements.