Abstract
Purpose :
AOSLO images contain intraframe distortion due to eye motion during acquisition, which can affect measurements of the cone mosaic. Algorithms designed to correct intraframe distortion (dewarping) assume that fixational eye movements are spatially random, however this may not hold for patients with nystagmus. Here, we evaluated the performance of a dewarping algorithm in patients with variable degrees of nystagmus.
Methods :
Metrics of distortion (interframe phase correlation coefficient (PCC) or intraframe Fourier correlation) were measured in confocal AOSLO videos at 1° temporal in 10 healthy controls, 10 subjects with achromatopsia with high nystagmus (ACHMhigh) & 10 subjects with achromatopsia with low nystagmus (ACHMlow). For each video, 2 minimally-distorted and 2 representative (median) frames were selected according to each distortion metric. Additionally, 2 other frames were randomly selected, resulting in a total of 5 frame pairs per video. All 10 frames were separately used as a reference frame for strip-registering and averaging the video. The normalized cross-correlation (NCC; inversely related to distortion) between the resultant image pairs was measured before and after dewarping. This process was then repeated using only the PCC representative frame pair and the change in NCC after dewarping was assessed using a range of 2-10 frames included in the distortion estimation.
Results :
Dewarping resulted in an increased NCC between image pairs in 147 (98%) comparisons. Compared to controls, mean NCC after dewarping was similar for the ACHMlow group (p=0.73) but significantly lower in the ACHMhigh group (p=1E-4). Minimally-distorted frames according to PCC had significantly higher NCC after dewarping compared to randomly selected frames (p=0.016) but were not significantly different than the other frame selection methods. With fewer than 5 frames, 66% of videos failed to register, but when using between 5-10 frames, there was no significant effect of the number of frames included in the distortion correction on change in NCC.
Conclusions :
Selection of a minimally-distorted reference frame and dewarping provides a superior approximation of an undistorted AOSLO image compared to strip-registration alone or random frame selection plus dewarping, even with as few as 5 registered frames. Dewarping may not fully correct distortion in videos with nystagmus due to non-random eye motion within the imaged area.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.