OCTA flow signal enhancement by reducing residual structural signal

Abstract

Purpose : The image quality of OCT angiography (OCTA) strongly depends on the contrast between static tissue and moving scatterers, i.e. blood cells travelling through capillaries. Simply calculating the difference or variance between multiple acquisitions at the same location attenuates signal from static tissue and highlights perfused capillaries. However, the high contrast OCTA images clinicians are used to, undergo significant image processing before they are displayed. We present a fast, practical approach for further suppressing signal from static tissue and hence enhancing the contrast of OCTA images.

Methods : We generate OCTA B-scans using the Optical Micro Angiography (OMAG) algorithm, taking the complex OCT signal as input. In such images (Fig. 1b), signal from static tissue is, in contrast to signal from the capillaries, already significantly attenuated. One can however still observe some residual signal from static tissue. We therefore assume that this image corresponds to the additive mix (I_l) of a fraction of the structural intensity signal (ωI_u) and the pure flow signal (x), i.e. I_l =x + ωI_u. Because one can assume the desired flow signal, x, with all structure removed, to look fundamentally different from the structure image, I_u, ω can be solved by minimizing the square of the normalized cross-correlation between I_u and I_l as argmin_ω γ²(I_u, I_l - ωI_u). Its explicit solution is ω = cov(I_u, I_l) / var(I_u), which can be rapidly computed.

Results : We have tested our method on scans from CIRRUS^TM HD-OCT 5000 with AngioPlex OCT Angiography (ZEISS, Dublin, CA). Fig. 1a) shows the structural intensity B-scan I_u, b) the mixed structure-flow image I_l and c) the resulting flow image x. A clear improvement in static tissue suppression can be observed. Fig 2 shows OCTA en face projection images before (a) and after (b) the processing described above. A substantial contrast enhancement can again be appreciated. Further, a suppression of the horizontal lines in a) can be observed, which are a common OCTA artifact typically caused by eye motion.

Conclusions : We present a practical and fast method to improve the contrast of retinal OCTA images. Testing on clinical data demonstrated superior signal suppression of static tissue.

This abstract was presented at the 2019 ARVO Imaging in the Eye Conference, held in Vancouver, Canada, April 26-27, 2019.

 

Fig. 1 a) structural B-scan; b) direct output of the OMAG algorithm; c) resulting flow image after proposed processing

 

Fig 2 a) direct output of the OMAG algorithm; b) resulting flow image after proposed processing