Purchase this article with an account.
Madhusudhanan Balasubramanian, Wasi Ahmed; Attenuation Correction and Building 3-D Geometry of Retinal Vasculature from SD-OCT Volume Scans without Contrast Agents. Invest. Ophthalmol. Vis. Sci. 2018;59(9):5066. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Ocular structures with high optical absorption and/or scattering (e.g. blood vessels) significantly attenuate OCT light propagation and cast shadows below them in OCT a-scans. We present an improved attenuation compensation method to minimize shadows and resolve tissue structure by reframing and solving the governing equation of OCT light transport in a tissue. Further, we present signal processing techniques to extract retinal vascular structure without contrast agents.
OCT attenuation correction procedure by Gerard et al IOVS 2011 uses division by an integral with limits from x to ∞. This introduces noise amplification near bottom boundaries of OCT scans. To overcome this, we reframed the governing equation of the received OCT signal as an integral equation with an exponential non-linearity. To further minimize noise amplification, we identified the bottom extent of each a-scan using a cumulative sum (CUSUM) based signal segmentation. Frangi filter was used to extract retinal blood vessel network from the corrected OCT volumes. The location and orientation of retinal blood vessels were identified using eigen analysis of the Hessian matrix at each retinal location. Vessels of varying diameters were identified by conducting eigen analysis at multiple scales using scale-space theory. A 3-D geometrical model of the vessel network was extracted from the scale-space analysis, statistical shape modeling of the retinal vessel and an edegeless active contour model. Simulated a-scans and SD-OCT volume scans (Spectralis, Heidelberg Engineering) of ex vivo porcine eyes (Balasubramanian M et al ARVO 2017) were used for validation.
Fig. 1 shows an example of attenuation compensation for a volumetric SD-OCT scan of an ex vivo porcine eye. Fig. 2 shows retinal vascular architecture extracted from the SD-OCT scan.
By directly solving the integral equation with exponential non-linearity that govern the OCT light transport in tissue, the modified procedure significantly minimizes noise amplification near the signal boundaries and the procedure is adaptive to the signal. Due to poor vessel contrast in SD-OCT scans, it remains a challenge to extract a full retinal vascular network using the Frangi filter alone. Use of a shape tracking algorithm (active contour) provides a more robust and more complete vascular network of the retina from SD-OCT scans.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.
This PDF is available to Subscribers Only