• aorta
• artificial intelligence
• deep neural network
• graph neural network
• thoracic aortic aneurysm
• wall shear stress

By representing mesh as a graph, it is possible to apply graph convolution to extract features and preserving the spatial coherency. In our work, we focused in the prediction of wall shear stress distribution directly over 3D surfaces. Each surface represent a thoracic aorta affected by aneurysm. All the geometries used come from real segmentations.