Statistical modeling of retinal optical coherence tomography using the Weibull mixture model

Area of Science:

• Ophthalmology
• Biomedical Imaging
• Statistical Modeling

Background:

• Retinal optical coherence tomography (OCT) images possess complex statistical properties.
• Existing models may not fully capture the asymmetry and heavy-tailed nature of OCT intensity distributions.

Purpose of the Study:

• To introduce a novel statistical model for retinal OCT images based on a mixture of Weibull distributions.
• To evaluate the model's performance in image denoising and disease classification.

Main Methods:

• A mixture of six Weibull distributions was proposed to model retinal OCT image features.
• The Weibull mixture model was converted to a Gaussian mixture model via histogram matching for use with spatially constrained Gaussian mixture model (SCGMM) denoising.
• Model parameters and goodness of fit (GoF) were used as feature vectors for support vector machine (SVM) classification.

Main Results:

• The proposed Weibull mixture model demonstrated a better goodness of fit (GoF) with fewer parameters compared to previous models.
• OCT image denoising using the proposed model showed remarkable improvements in contrast to noise ratio (CNR) and texture preservation (TP).
• The model achieved effective classification of healthy retinal OCT images from pigment epithelial detachment (PED) disease using SVM.

Conclusions:

• The novel Weibull mixture model accurately describes the statistical characteristics of retinal OCT images.
• The proposed model significantly enhances OCT image quality and shows promise for OCT-based disease diagnosis.