Simulation-Based Segmentation of Blood Vessels in Cerebral 3D OCTA Images

TL;DR

This work extracts patches from vessel graphs and transforms them into synthetic cerebral 3D OCTA images paired with their matching ground truth labels by simulating the most dominant 3D OCTA artifacts, enabling annotation-free blood vessel segmentation in cerebral 3D OCTA images.

Abstract

Segmentation of blood vessels in murine cerebral 3D OCTA images is foundational for in vivo quantitative analysis of the effects of neurovascular disorders, such as stroke or Alzheimer's, on the vascular network. However, to accurately segment blood vessels with state-of-the-art deep learning methods, a vast amount of voxel-level annotations is required. Since cerebral 3D OCTA images are typically plagued by artifacts and generally have a low signal-to-noise ratio, acquiring manual annotations poses an especially cumbersome and time-consuming task. To alleviate the need for manual annotations, we propose utilizing synthetic data to supervise segmentation algorithms. To this end, we extract patches from vessel graphs and transform them into synthetic cerebral 3D OCTA images paired with their matching ground truth labels by simulating the most dominant 3D OCTA artifacts. In extensive experiments, we demonstrate that our approach achieves competitive results, enabling annotation-free blood vessel segmentation in cerebral 3D OCTA images.

Figures (8)

• Figure 1: 3D renderings of our proposed synthetic cerebral 3D OCTA images.
• Figure 2: Overview of our proposed method. First, we extract patches from vessel graphs and transform them into a vast amount of voxelized volumes; second, we transform the voxelized volumes into synthetic cerebral 3D OCTA images by simulating the most dominant image acquisition artifacts; and third, we use our synthetic cerebral 3D OCTA images paired with their matching ground truth labels to train a segmentation network.
• Figure 3: Slices of real (top) and synthetic (bottom) cerebral 3D OCTA images. It should be highlighted that we accurately match 3D OCTA-specific artifacts, resulting in synthetic images almost indistinguishable from real images.
• Figure 4: a) Visualization of sampled patches in a graph representation of a vascular corrosion cast; b) exemplary manual annotation, including regions provided to determine vessel size-specific segmentation performance (all, small, large).
• Figure 5: Qualitative results. The U-Net trained on our synthetic data (right) accurately segments vasculature, alleviating the need for labor-intensive manual annotations.