Abstract
Purpose::
To demonstrate automatic blood vessel parameter (3D coordinates, vessel diameter, and blood flow velocity) extraction from OCT scans around the optic disc acquired with an ultrahigh resolution (~3 µm) spectral domain OCT (SD-OCT) system operating at a speed of 24000 A-lines/s and the application of the technique for quantitative blood flow dynamics study.
Methods::
A circular scan pattern, consisting of 56 normal density (1024) and one high density (8192) scans around the optic disc with radius varying from 0.73mm to 2.73mm, was used. A Doppler image is calculated from the high density scan. By using the concept of OCT fundus shadowgram_an intensity distribution can be calculated for each OCT B-scan after removing the surface reflections. Through a series of signal processing including smooth filtering, background correction, and thresholding, the vessel center and diameter can be determined. Upon determination of the lateral position and the vessel diameter, the coordinate in the depth direction of each blood vessel is calculated in combination with the Doppler information for the vessel. The extraction of the vessel coordinates and diameter makes it possible to calculate the orientation of the vessel in reference to the direction of the incident sample light, which in turn can be used to calculate the absolute blood flow velocity and the flow rate.
Results::
The calculated circular fundus image compared well with the corresponding color fundus photograph where the number of vessels on the fundus matched the number of vessels detected. 8 eyes where tested and a total of 125 vessels where detected leading to a 100% detection accuracy in the lateral position. The calculated depth positions of the vessels were compared with the manually determined blood vessel centers on the Doppler image. The offset of the calculated vessel center from the targeted center for 106 blood vessels was less than ½ of the corresponding vessel radius, which leads to a 84.8% detecting accuracy. Large vessels, defined to have a diameter larger than 50 pixels, depth positions were detected with an accuracy of 93.3%. 3D angiogram can then be constructed and 3D blood flow dynamics can be studied with the technique.
Conclusions::
Quantitative 3D fundus angiogram can be generated by combining the information of OCT fundus shadowgram and the Doppler image obtained with spectral-domain OCT. This technique allowed quantitative imaging of the blood flow dynamics around the optic disc, which may benefit the study of eye diseases like glaucoma.
Keywords: image processing • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • blood supply