Vox2cortex: Fast explicit reconstruction of cortical surfaces from 3d mri scans with geometric deep neural networks
The reconstruction of cortical surfaces from brain magnetic resonance imaging (MRI) scans
is essential for quantitative analyses of cortical thickness and sulcal morphology. Although
traditional and deep learning-based algorithmic pipelines exist for this purpose, they have
two major drawbacks: lengthy runtimes of multiple hours (traditional) or intricate post-
processing, such as mesh extraction and topology correction (deep learning-based). In this
work, we address both of these issues and propose Vox2Cortex, a deep learning-based …
is essential for quantitative analyses of cortical thickness and sulcal morphology. Although
traditional and deep learning-based algorithmic pipelines exist for this purpose, they have
two major drawbacks: lengthy runtimes of multiple hours (traditional) or intricate post-
processing, such as mesh extraction and topology correction (deep learning-based). In this
work, we address both of these issues and propose Vox2Cortex, a deep learning-based …
[PDF][PDF] Vox2Cortex: Fast Explicit Reconstruction of Cortical Surfaces from 3D MRI Scans with Geometric Deep Neural Networks—Supplementary Material
We want to give a brief mathematical intuition why our curvature-weighted Chamfer loss
emphasizes geometric accuracy in high-curvature regions compared to low-curvature
regions. Imagine therefore two ground-truth points a and b with respective curvature κ (a)< κ
(b) and closest predicted points u and v as shown in Figure 1. Furthermore, let the distance
from the prediction to the ground truth be equal in both cases, such that∥ u− a∥=∥ v− b∥.
For the sake of simplicity, we treat the predicted values u and v as the parameters that are …
emphasizes geometric accuracy in high-curvature regions compared to low-curvature
regions. Imagine therefore two ground-truth points a and b with respective curvature κ (a)< κ
(b) and closest predicted points u and v as shown in Figure 1. Furthermore, let the distance
from the prediction to the ground truth be equal in both cases, such that∥ u− a∥=∥ v− b∥.
For the sake of simplicity, we treat the predicted values u and v as the parameters that are …