Multiple frames–based OCT angiography is susceptible to BTM and requires an increase in the number of acquired frames to improve the SNR. To minimize motion artifacts, we used two sequential frames and averaged the amplitudes in both the lateral and depth directions. The algorithm that we developed for this purpose is given below by Eq. (1): Display Formula
(1)where is the amplitude of the OCT image and , , and are indices for the frame, transverse, and depth pixels, respectively. is the number of averaged depth pixels and is the number of averaged A-lines. Pixel-shifted images were used to correct for BTM. This correction is accounted for in the modified version of Eq. (1), given below in Eq. (2): Display Formula
(2)where and are the pixel translations of the lateral and depth directions, respectively. Because we could not predict the direction of tissue motion, we calculated the squared difference of two OCT images by changing and , to find the shift values that minimized the sum of the pixel values. Generally, larger and values for the averaging will increase image SNR; however, it also increases the computation time and results in a “blurring” and loss of smaller vessels. The larger and values for the image registration will increase the computation time. To display BTM-corrected OCT angiographic images in real time, the averaging procedure used and ; the parameters for image registration were and in this work.