Abstract
Purpose :
Ophthalmic diagnostics using optical coherence tomography (OCT) is limited by trade-off between field-of-view (FOV) and sampling density and bulk eye motions. While advances in laser technology have significantly increased OCT imaging speeds, faster acquisition reduces image signal-to-noise ratio and necessitates post-acquisition frame-averaging. We demonstrate a novel multimodal ophthalmic imaging system that simultaneously acquires spectrally encoded scanning laser ophthalmoscopy (SESLO) and cross-sectional OCT images using a shared swept-source, imaging optics, and acquisition system. In post-processing, en face SESLO images, which are intrinsically co-registered with the OCT FOV, are used to estimate lateral motions and sequential OCT cross-sections are used to co-register axial motion for volumetric co-registration and mosaicking of multiple three-dimensional OCT datasets.
Methods :
Concurrent en face SESLO and cross-sectional OCT images of in vivo human optic nerve were acquired with 1376 x 500 pixels at 200 frames-per-second. Four volumetric OCT datasets were acquired in 2.5 seconds each, co-registered, and mosaicked in post-processing. Lateral and rotational motion between each cross-sectional OCT image was calculated based on SESLO data using discrete Fourier and adaptive log-polar transform algorithms. Bulk motions were used as spatial offsets for sampling positions in the corresponding volumetric OCT datasets. Multi-volumetric mosaicking was achieved by three-dimensional natural neighbor interpolation between bulk motion offset sampling positions and “ideal” evenly spaced sampling positions. Image intensities of all volumetric datasets were normalized prior to interpolation to reduce edge effects.
Results :
Lateral and rotational motion measured using SESLO shows micro-saccades, fixation drifts, and blinks in each volumetric OCT dataset. These bulk eye motion artifacts are compensated for and multiple volumetric OCT datasets are combined in post-processing (Fig. 1).
Conclusions :
Multi-volumetric co-registration and mosaicking in post-processing removes bulk motion artifacts and enables recovery of missing volumetric OCT data. The proposed methods will enable densely sampled extended FOV ophthalmic imaging that is less susceptible to eye motion and without loss of image quality.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.