We present deep-learning based multi-contrast optical coherence tomography (OCT) imaging methods for the analysis of retinal tissue properties. Two modalities, synthesizing degree-of-polarization-uniformity (syn-DOPU) and scatterer density estimator (SDE), were introduced. Syn-DOPU generates DOPU images from non-polarization sensitive OCT images, and hence eliminates the need for special hardware. SDE provides robust scatterer density estimation irrespective of measurement and ocular medium conditions. The methods were applied to age-related macular degeneration cases, and revealed the detailed abnormality of the retinal pigment epithelium. Additionally, layer and sector analyses of normal cases demonstrated positional and age-related variations of DOPU and scatterer density.
|