Imaging retinal nerve fiber bundles at ultrahigh-speed and ultrahigh-resolution using OCT with adaptive optics

Ultrahigh speed line scan detectors based on CMOS technology have been recently demonstrated in ultrahigh resolution spectral-domain optical coherence tomography (UHR-SD-OCT) for retinal imaging. While successful, fundamental tradeoffs exist been image acquisition time, image sampling density, and sensitivity, all of which impact the extent of motion artifacts, visualization of fine spatial detail, and detection of faint reflections. Here we investigate these tradeoffs for imaging retinal nerve fiber bundles (RNFBs) using UHR-SD-OCT with adaptive optics (AO). Volume scans of 3°x3° and 1.5°x1.5° were acquired at retinal locations of 3° nasal and 6° superior to the fovea on a healthy subject. Dynamic AO compensation across a 6 mm pupil provided near-diffraction-limited performance. The acquisition rates were 22.5k lines/s and 125k lines/s with A-lines spaced at 0.9 μm and 1.8 μm and B-scans at 1.8 μm and 9 μm. Focus was optimized for visualizing the retinal nerve fiber bundles (RNFBs). En face projection and crosssectional views of the RNFBs were extracted from the volumes and compared to images acquired with established conventional CCD-based line-scan camera. The projection view was found highly sensitive to eye motion artifacts, yet could only be partially compensated with coarser sampling, since fine sampling was necessary to observe the microscopic features in the RNFBs. For the cross-sectional view, speckle noise rather than eye motion artifacts limited bundle clarity. The highest B-scan density (1.8 μm spacing) coupled with B-scan averaging proved the best combination. Regardless of view, the higher line rate provided better RNFB clarity.