Abstract
Purpose :
To develop a reliable and accurate approach for choriocapillaris (CC) visualization and design meaningful indices for CC quantification.
Methods :
Five repeated 3mm x 3mm scans were collected from five volunteers using SS-OCT Plex-Elite 9000 (Carl Zeiss Meditec, Dublin, CA). A linear OMAG algorithm was used to generate OCTA data and CC was defined as a 15 um thickness slab under Bruch’s membrane using a semi-automatic segmentation software. A novel registration algorithm consisting of translational, affine and B-spline registration was developed to average multiple CC scans. A novel CC flow voids (FV) quantification algorithm using local fuzzy C means clustering for segmentation was developed to calculate CC FV percentage, the ratio of CC FV area to the whole area, averaged CC FV size, the averaged CC FV size of all detected FVs, CC FV number, the number of all detected FVs, CC FV complexity index, CC FV perimeter divided by 4 pi and CC FV area. Registered and averaged CC scans were analyzed and evaluated for averaging performance using global entropy, standard deviation, local texture correlation and peak signal-noise-ratio. Central foveal CC and peripheral CC were compared quantitatively, excluding projection artifacts.
Results :
After registration and averaging, CC angiograms showed good visual correlation with SEM images. Global entropy, standard deviation, local texture correlation and peak signal-noise-ratio all show significant improvement with the increase of repeated number scans for averaging (linear or log fitted, all R2>0.91, p<0.01). Statistical analysis determined that the optimal scan-number for averaging is 4; and more repeated scans does not show significant advantage in terms of CC FV percentage. Using 4 repeated CC OCTA images, central foveal CC showed significantly lower CCFV percentage (p=0.01), fewer CCFV number (p=0.001) and less complex FV shape (p=0.007), and no significant difference in FV size, compared to peripheral locations.
Conclusions :
CC can be reliably visualized and quantified by co-registering and averaging multiple SS-OCTA scans. More repeated scans yield better image quality by reducing random speckle noises. Our quantitative analysis CC showed significant differences between central foveal CC and peripheral CC, giving new opportunities for reliable and accurate assessment of CC involvement in diseases.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.