Real-time bulk motion insensitive flow segmentation algorithm for Doppler spectral optical coherence tomography

We present a simple and efficient numerical technique for segmentation retinal and choroidal blood vasculature with bulk motion correction in functional Doppler Spectral Optical Coherence Tomography (Doppler SOCT). The technique uses local variance of velocity tomogram which is higher in the areas of the tomogram with internal flow. The resulting variance map reveals the position of vessels. This can be used either for vessel segmentation purposes or for masking the vessels on velocity tomograms. The remaining velocity information is connected only with static structure velocity offset. As only Fourier transformations are used in calculations the algorithm removes the bulk motion from velocity tomograms and provides images of segmented vessels with speed of 80 000 lines/s. The algorithm is shown to work with velocity tomograms obtained by joint Spectral and Time domain OCT (STdOCT).