Purchase this article with an account.
Vivien VASSEUR, Martine Mauget-Faÿsse, Laurence Salomon; 3D Image Processing and Image Display of the Vascular Network in OCT-Angiography images. Invest. Ophthalmol. Vis. Sci. 2019;60(9):3085. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The release of new systems to capture OCT-Angiography images has opened a broad new field of image quantification directly performed on 3D volume of data. Starting from a series of 2D images, a method was developed to reconstruct the total volume in 3D, to visualize it in a custom 3D viewer, and then perform measurements of specific areas of interest of the volume.
Images were captured by the Spectralis system (Heidelberg Engineering, Germany). A set of 30 volumes from 15 patients was used to develop and validate the method. Volumes were exported in RAW format and opened in a software that was developed for the project. The volume was then automatically displayed in the user interface 3D viewer. Window synchronization between image windows helped the user to visually see the position of pathologies in one image versus other images of the same visit and patient. Since the quality of Angio OCT images is not excellent, due to the presence of artifacts generated by the acquisition sensor, they need to be processed. First, to eliminate the noise, a Kriging filter is performed on the OCT-Angiography volume. Second, to enhance the display of vessels and generate accurate results, a true 3D skeleton transformation is performed on the volume. It thins the vessels to facilitate the detection of crossing points and branches, and to compute measurements. The vessel density is computed and displayed in a color map, the measure of the tortuosity of the vessel network is computed, the count of microaneuryms is manually performed with validation in the volume, and the ischemic area surface is evaluated. Fourth, a set of display tools help looking at pathologies in detail. For example, microaneurysms are selected in the 2D infrared image, and then displayed in 3D.
A whole set of 30 volumes has been analyzed, measurements were computed, and then visually validated. In addition, a set of videos of the volumes was generated to better understand the physiology of the retina.
The automatic quantification of OCT-Angiography images was found to be very reproducible. Computing measurements directly on the volume in the 3D space proves to be more reproducible compared to a 2D analysis. In addition, the method that was developed takes benefit of the availability of 3D data, and opens a wide range of new analysis.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
This PDF is available to Subscribers Only