Speed performance of each module in our framework and the current state-of-the-art techniques from the literature.

* Because the current state-of-the-art techniques from the literature use 1 GPU, we compare the execution time of our framework selectively using CPU and 1 GPU with their execution time.† Both our system and Jian et al.’s use GTX680 and Zhang and Kang’s [19] uses GTX590 that include 2 GTX580.‡ In order to compare with Jian et al.’s algorithm [20], we simplified our algorithm to reconstruct FD-OCT images like what Jian et al.’s algorithm did, for example, like performing FFT only once.§ While Jian et al. support 3D rendering [21] as well, they did not report the execution time of only 3D rendering. However, because our algorithm is also based on volume rendering algorithms from NVIDIA [22] as theirs, we believe that both algorithms have similar speed performance.Speed performance of each module in our framework and the current state-of-the-art techniques from the literature.