Improved 3D-OCT Signal Quality and Accuracy in Retinal Pathologies using Image Registration based Motion Correction and Merging of Multiple Orthogonal Raster Scans

Abstract

 

Purpose:

 

To evaluate whether image registration based motion correction and merging of multiple raster scanned OCT volumes of pathologic eyes can help improve the visualization of retinal pathologies as opposed to unprocessed OCT volume scans.

 

Methods:

 

Eyes of multiple subjects were scanned at New England Eye Center using multiple prototype OCT devices such as an ultra-high resolution prototype system developed by MIT, operating at a central wavelength of 850 nm with an axial image resolution of 3 um and an A-Scan rate of 90 kHz. For each eye, multiple raster scans with orthogonal fast scan direction and isotropic lateral sampling of a certain area of the retina were acquired. After acquisition, the orthogonal raster scans were motion corrected in three dimensions on a per A-Scan basis, registered to a common space and subsequently merged. This method produces a single motion corrected volume with improved signal quality from the original data. The motion correction is based on optimizing a global objective function. Each volume is transformed such that the resulting volumes become most similar while penalizing the motion as modeled by the transformation. No additional tracking hardware was used. The resulting volumes were compared with the original volumes using visual inspection.

 

Results:

 

Visual inspection shows that the technique is able to remove motion artifacts and improve signal quality both in normal as well as in pathologic eyes. The visibility of pathologic features such as drusen, neovascularization and retinal detachments is improved. In addition, densely sampled raster scans and removal of motion artifacts make it less likely for pathology to be missed and allow for appreciation of the 3D nature of defects. The attached figure shows corresponding central slices of a macula region volume (6x6 mm, 400x400 A-Scans) acquired with an UHR-UHS-850nm OCT device before and after processing.

 

Conclusions:

 

Results suggest that the proposed method can help improve visualization and 3D analysis of retinal pathology. Since signal quality in pathologic eyes tends to be lower than in normal eyes, signal improvement through merging of multiple data sets is especially useful.  

 

Keywords: image processing • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • imaging/image analysis: clinical