Abstract
Purpose :
Tracking the axial position of eye with optical coherence tomography (OCT) is useful in clinics for reducing the effect of axial eye motion. Because both real (RI) and mirror (MI) images appear in the reconstruction of Fourier domain OCT, only half-range is usually cropped for display. For eye movement in axial direction or reference (Ref.) arm misalignment, the MI can enter the display window and introduce artifacts. We proposed a new method for z-tracking by utilizing all the information of full-range OCT reconstruction to detect the MI artifact.
Methods :
Due to the unbalanced dispersion between the sample and Ref. arm, the RI becomes sharper than the MI after the numerical dispersion compensation. So the first lag of the autocorrelation function (ACF) of an A-scan is smaller in the RI than the MI portion, which can be the indicator for automatically detecting the RI and MI. When the detected RI enters the other half-range and the MI fully appears in the normal display window, the algorithm can directly display the other half-range with the RI instead of readjusting the Ref. arm to improve the tracking speed.
Results :
Figure 1(a) shows an example of the full-range OCT reconstruction. The ACF is calculated separately for both half-ranges (red for above and green for below the zero-delay line), and the typical result of 1 A-scan [along the dash line in Fig. 1(a)] is shown in Fig. 1(b). In this case, the RI (green) shows steeper ACF and low value at the first lag. Figure 2 shows the flowchart of z-tracking using the full-range information. In this example [Fig. 2(a)] the RI enters the common deleted region (red box) and the MI shows up in the common display window (green box). The value of ACF at the first lag of each A-scan is plotted in the Fig. 2 (b). The display window can directly show the red box region [Fig. 2 (c)] based on the detection without readjusting the Ref. arm to move the RI into the common display window. This method has been tested with 1050 B-scans and successfully identified the presence of MIs in each acquisition.
Conclusions :
We proposed a new method for z-tracking based on the full-range information of OCT reconstruction and the different properties of the ACF of the RI and MI. This method can potentially reduce the total acquisition time due to less rejection of scan data.
This is a 2020 ARVO Annual Meeting abstract.