Improvement of phase stability of optoretinograms (ORG) by post-processing in raster scanning systems

Abstract

Purpose : To present the application of a subpixel motion correction method based on FFT upsampling [Thurman et al. 2008, Li et al. 2022] in modeling optoretinograms–light-evoked responses from mouse and human retinas using NIR optical coherence tomography (OCT).

Methods : Albino (Balb/c) mice were imaged in vivo with a custom scanning light ophthalmoscopy / optical coherence tomography (SLO/OCT) retinal imaging system. Before the ORG experiments mice were dark adapted for over 6 hours. We extracted phase-based ORG signals (time-dependent, stimulus-evoked changes of the phase difference (path length) between retinal layers) using 1) the standard phase-based method and 2) applying subpixel motion correction with FFT upsampling. The general concept for subpixel motion correction is presented in Fig. 1.

Results : The phase-based ORG signals, although sensitive to retinal motion and computationally demanding, provided ~nm sensitivity for detecting changes in retina layer positions. The OCT signal decorrelates over time and phase-based information about layer position blurs in the resulting noise. The high phase stability was maintained over seconds with 100 kHz acquisition experiments by introducing motion correction.

Conclusions : Our work demonstrates a successful implementation of phase-based ORG signal analysis in mice with subpixel motion correction. Subpixel motion correction improved signal and phase stability allowing for tracking of retinal x and z motion and extraction of the ORG signal in 10 s experiments.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

 

The motion correction concept. a) The input B-scan S1 reference for motion correction. b) Second B-scan S2 to be corrected. c) Corrected second B-scan using two-dimensional ramp (g). d) Cross-correlation spectrum of S1 (a) and S2 (b). e) Zoomed cross-correlation spectrum with zero position marked by yellow cross and position of maximum cross-correlation marked by the pink cross. The position of maximum cross-correlation defines the vector translation of the second B-scan. f) Result of FFT upsampling to estimate the location of the peak of cross-correlation with subpixel resolution. The phase ramp is based on the position of the cross-correlation peak in (f). The ramp is used to translate signal S2 (b). h) Plot of the axial displacement of a series of B-scans to the reference frame S1 over time. i) Plot of the lateral displacement of the series of B-scans to the reference frame S1 over time.