Abstract
Purpose :
Optical coherence tomography angiography (OCT-A) is prone to decorrelation tails, apparent flow in retinal tissue due to the blood flow within overlying vessels. Such artifacts can interfere with the interpretation of angiographic results. We present a fast practical approach for reducing these artifacts using slab images generated from two or more retinal layers.
Methods :
A slab image can be generated by linearly or nonlinearly projecting the OCT-A cube data (generated using Optical Micro Angiography (OMAGc) algorithms) within one or more layers. The artifact reduction method assumes that the observed lower slab image (I) is generated by additive mix of a signal from an upper slab (U) and a lower slab without artifacts (Î). The mixing problem can be formulated as Î=I-wU, where w∈[0.1] is the fraction of U added to Îto result in I. If the upper slab and the unknown lower slab with no artifacts do not have similar vessel patterns, w can be solved by minimizing the square of the normalized cross-correlation between U and Îas minwγ2(U,I-wU). The explicit solution is w=Cov(U,I)/Var(U), which is rapidly computed.
Results :
We have tested our method on 32 scans using CIRRUS HD-OCT 5000 with AngioPlex™ OCT Angiography (ZEISS, Dublin, CA), and 25 cases using PLEX™ Elite 9000 (ZEISS, Dublin, CA) with diseases such as Diabetic Retinopathy, wet and dry Age-related Macular Degeneration (AMD), Branch Retinal Vein Occlusion, Sickle Cell Disease, Central Retinal Artery Occlusion, Macular telangiectasia, and normal. Figure 1 and 2 show examples of decorrelation artifacts reduction by our method for two cases of Neovascular AMD.
Conclusions :
We present a practical and fast method to reduce the decorrelation tail artifacts in OCT-A images. This artifact reduction technique also can be used to correct a whole OCT-A cube with a sliding slab that is segmentation-independent. Our approach is specifically useful for visualization and interpretation of a variety of retinal vascular diseases.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.