Abstract
Purpose :
To demonstrate the possible misdiagnosis of type 1 choroidal neovascularization (CNV) underlying drusen in non-exudative AMD using optical coherence tomography angiography (OCTA) and the importance of appropriate boundary segmentation when interpreting flow images.
Methods :
Patients with drusen secondary to non-exudative AMD were enrolled in a prospective OCT study at Bascom Palmer Eye Institute using the Zeiss Angioplex™ OCT angiography instrument (Carl Zeiss Meditec Inc.) with a center wavelength of 840nm, a scan rate of 68,000 A-scans per second, and the FastTrac™ tracking system. The 3x3mm raster scan pattern was used, with 245 A-scans per B-scan, four B-scans repeated at each position, and 245 B-scan positions. Two kinds of custom slabs were created for visualizing the structure and flow underlying the drusen. The parallel boundaries of one 50μm slab followed the Bruch’s membrane contour, which is known as the RPE-fit line, and included the retinal pigment epithelium (RPE) at the top of the drusen. Using the custom map feature, we generated a second slab with the outer boundary following the RPE-fit line and the inner boundary following the RPE contour. The thickness of this second slab was adjusted so as to optimize the detection of the decorrelation signal from within the drusen. The structure and flow signals from under the drusen were then compared.
Results :
Ten eyes of eight AMD patients with drusen were imaged. In all ten eyes, an artifactual flow signal was seen on the first slab that could be misinterpreted as type 1 CNV. This flow signal was shown to be identical to the overlying retinal microvasculature, and was due to a decorrelation projection artifact of the hyperreflective RPE surface. The second slab that did not include the hyperreflective RPE layer did not show evidence of flow associated with the drusen.
Conclusions :
OCTA decorrelation projection artifacts can be misinterpreted as actual flow within a structure if the segmentation boundaries include a hyperreflective boundary like the RPE. One way to minimize this artifact is to be aware of its existence and to use a custom segmentation algorithm whenever possible that can avoid these hyperreflective boundary layers.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.