Abstract
Purpose: :
To introduce a method for co-registering and averaging fast high-resolution SDOCT volumes of the fovea as well as the optic nerve head (ONH) region, in order to reduce speckle noise and significantly enhance the contrast of microscopic retinal details.
Methods: :
The method is based on high-speed and high-resolution SDOCT capable of recording fast motion artifact free volume series of the human retina. A single volume covers a field of view of 1.5 mm x 1.5 mm for the fovea and 4.5 mm x 4.5 mm for the ONH region. The volumes are acquired at an axial scan rate of 200 kHz. The acquisition time for a densely sampled volume is only 1.25 s. The volumes are co-registered laterally, using the lateral cross correlation of the mean intensity projections in depth. The co-registration is then refined using local weighted mean transformations. The same transformations are accordingly applied to the corresponding full SDOCT volumes. In a next step the laterally co-registered volumes are registered axially to a single reference volume. Finally the perfectly co-registered volumes are used to calculate average volumes. The obtained enhanced contrast foveal volumes are analyzed with a classifier based vessel extraction algorithm.
Results: :
Volume series at the fovea and optic nerve head of healthy volunteers have been recorded with a high speed and high resolution SDOCT system. The resulting average volumes provide significantly reduced speckle noise and unprecedented contrast currently only achieved by tomogram averaging. The enhanced contrast volumes reveal microscopic details such as individual nerve fiber bundles, capillary structures, and even photoreceptor patterns, which cannot be observed in single volume scans. The fovea average volumes show especially strong contrast for the parafoveal capillary network. Using a classifier based vessel filter we were able to extract the complete parafoveal capillary network. Opposed to variance techniques such intensity based vessel filters do not show decorrelation tails below the capillaries, thus allow comprehensive 3D visualization.
Conclusions: :
The enhanced contrast and reduced speckle noise of average volumes significantly boost the value of high resolution OCT. The remarkably enhanced image quality not only simplifies diagnostics but also improves the outcome of additional post processing algorithms, such as vessel segmentation or retinal layer thickness measurements.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • image processing • blood supply