Adaptive Forward-Reverse Filter using Interpolation Methods for Artifact Suppression in Retinal Prostheses.

Electrical stimulation on retinal ganglion cells (RGCs) induce the short-latency (directly-evoked) and long-latency (indirectly-evoked) responses of RGCs. The artifact suppression and isolation of direct RGC spike is required for proper analysis of visual information. Adaptive forward-reverse filter (FR filter) using interpolation method is proposed and evaluated. On selected over 1.6 ms waves, which is suspected as artifact, 2 new data points are linearly interpolated between the recorded data points. After that, the interpolated data are filtered by frequency-based FR filter (500 Hz). The proposed algorithm shows the best true positive rate (0.7629) comparing with the SALPA and the simple FR filter without the interpolation method. In point of view of the false positive rate, the proposed algorithm demonstrates the second-best performance (0.0047), not better than the SALPA (0.0006).