High performance OCTA enabled by combining features of shape, intensity, and complex decorrelation
Huakun Li, Kaiyuan Liu, T. Cao, Lin Yao, Ziyi Zhang, Xiaofeng Deng, Chixin Du, Peng Li
Abstract
Motion contrast optical coherence tomography angiography (OCTA) entails a precise identification of dynamic flow signals from the static background, but an intermediate region with voxels exhibiting a mixed distribution of dynamic and static scatterers is almost inevitable in practice, which degrades the vascular contrast and connectivity. In this work, the static-dynamic intermediate region was pre-defined according to the asymptotic relation between inverse signal-to-noise ratio (iSNR) and decorrelation, which was theoretically derived for signals with different flow rates based on a multi-variate time series (MVTS) model. Then the ambiguous voxels in the intermediate region were further differentiated using a shape mask with adaptive threshold. Finally, an improved OCTA classifier was built by combining shape, iSNR, and decorrelation features, termed as SID-OCTA, and the performance of the proposed SID-OCTA was validated experimentally through mouse retinal imaging.