Computational aberration correction in spatiotemporal optical coherence (STOC) imaging

Abstract

Spatiotemporal optical coherence (STOC) imaging is a new technique for suppressing coherent cross talk noise in Fourier-domain full-field optical coherence tomography (FD-FF-OCT). In STOC imaging, the time-varying inhomogeneous phase masks modulate the incident light to alter the interferometric signal. Resulting interference images are then processed as in standard FD-FF-OCT and averaged incoherently or coherently to produce cross-talk-free volumetric optical coherence tomography (OCT) images of the sample. Here, we show that coherent averaging is suitable when phase modulation is performed for both interferometer arms simultaneously. We explain the advantages of coherent over incoherent averaging. Specifically, we show that modulated signal, after coherent averaging, preserves lateral phase stability, enabling computational phase correction to compensate for geometrical aberrations. Ultimately, we employ it to correct for aberrations present in the image of the photoreceptor layer of the human retina that reveals otherwise invisible photoreceptor mosaics.

In vivo imaging of the human cornea with high-speed and high-resolution Fourier-domain full-field optical coherence tomography

“…However, FD-OCT system was an order of magnitude slower -recording the whole volume in 2.8 s -despite using one of the fastest spectrometers commercially available (of 150 kHz). Furthermore, it was not possible to perform DAC on the volumes acquired with the FD-OCT system, whereas in FD-FF-OCT it was feasible because of the laterally stable phase even when it was scrambled by DM before the interferometer [42].…”

“…We applied the U-Net neural network adopted for the endothelium cells [42]. We introduced minor modifications to the original concept but the architecture of the neural network followed a standard fully convolutional structure of the U-net, with a contraction and an expansion path, acting as encoder and decoder blocks.…”

“…The trained network was then applied on a test image in a sliding window (32 ´ 32 pixels) that was moved across consecutive patches, as explained in Ref. [42]. The final edge probability map was created as a result of averaging a network response over the overlapped patches.…”

“…We have also performed digital aberration correction (DAC) on the acquired 3D data volumes, as described in Ref. [42], which helped to correct for defocus. We thus were able to show that volumes, acquired in a fraction of a second and averaged to increase the signal-to-noise ratio (SNR), allowed capturing a significant portion of the cornea.…”

Simulation of multimodal optical coherence tomography for biomedical diagnostics: comprehensive full-wave description based on ballistic scattering of arbitrary-profile beams

“…3 that both worsening of the PSF-localization in the focal plane and aberration-induced translational shifts of the entire C-scan can readily be simulated. Such OCT images with a priori known aberrations can be used for development and comparison of aberration-correction procedures, which is important, e.g., for improving the quality of OCT-imaging for ophthalmic applications [24,26,27,28] (a special discussion of utilization of simulated OCT images for such applications will be published later). Further, the ability of the model to easily simulate non-Gaussian beams is demonstrated for a Bessel beam (such beam attract significant interest in OCT due to nearly non-diverging character of propagation 29 ).…”