Purchase this article with an account.
Justin V Migacz, Nripun Sredar, Davis B. Zhou, Maria V. Castanos, Alexander Pinhas, Rishard Weitz, Alfredo Dubra, Richard B Rosen, Toco Yuen Ping Chui; Multi-directional image fusion in AOSLO angiography. Invest. Ophthalmol. Vis. Sci. 2020;61(7):223.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
AOSLO angiography captures video-rate image sequences of the retinal vasculature, including blood cells and vessel walls. By splitting the detection of back-scattered non-confocal light into 4 quadrants, an arbitrary edge contrast direction can be generated in post processing. Any single image rendering, however, does not capture the edge features of every vessel due to the diversity of vessel orientations. In this work, we present a new software method to merge the best edge features of every directional rendering into one fused image.
AOSLO image sequences of healthy subjects and patients with retinal vascular abnormalities, such as non-perfused blood vessels and microaneurysms, were acquired. These sequences were registered, averaged and tiled to generate high-contrast images of the blood vessels. The images were divided into 9-by-9 pixel subregions in which the orientation of the split-detection images were adjusted for maximum pixel variance. The orientation maps were then median filtered to reduce unwanted discontinuities occurring in low-contrast regions of the image. The final fused images were generated by rendering the split-detection orientation at each pixel. This procedure mirrors one employed by Huang et al. (PMID: 21964097) for ranking a series of subregions in an image sequence prior to averaging.
The visualization of parafoveal vessels, such as those shown in Figures1 and 2 a healty normal and a sickle cell patient, respectively, was enhanced by the proposed image fusing method. Some portions of the image showed spurious discontinuities along curved portions of vessels, (see arrows in Fig.1c & 2c). Areas with no vessels and had low variability of intensity did have frequent orientation oscillations but did not distract from the main vessel features within the image.
Fusion of split-detection images of the vasculature resulted in more vessels being visible within a single image than with any separate single-orientation image. Content-rich summary images such as these can lead to faster interpretation and therefore more likely adoption of AOSLO angiography in clinical settings.
This is a 2020 ARVO Annual Meeting abstract.
Normal parafoveal vessel.
Parafoveal capillary occluded by blocked erythrocytes in a sickle cell retinopathy patient.
This PDF is available to Subscribers Only