Data collection

Sample preparation. Normal individuals will be considered at ages 2, 3, 6, 9, 12, 18 and 24 months ; diabetic ones at ages 2, 3, 6, 9 and 12 months. Aortas will be harvested as described (REF). During the aging period, morphological and hemodynamic parameters will be recorded every 3 months on each individual to track its “personal history” : blood pressure measurements, cardiac parameters, vessel diameters, local and global measurements of vascular stiffness, flow velocities, blood viscosity, and classical biochemical parameters.

Data acquisition. Samples will be imaged at the ANATOMIX beamline of Synchrotron SOLEIL. The specimens will be scanned by µCT with a pixel size of 0.65 µm, each scan yielding volumes of 2048x2048x2048 voxels (1.73mm³) representing 32 GB of reconstructed volume data. Full aortas will also be imaged to have a full enveloppe for meshing.

Image preprocessing and segmentation. In MODELAGE, image analysis requires the development of dedicated image preprocessing (filtering) and processing (segmentation) methods and tools. The main challenges are here related to the ability of efficiently considering the specific physical and semantic properties of the images to design well-fitted image processing operators. The developed methods will be at the frontier between segmentation and skeletonization but we will also use topological approaches for extracting elastic lamellae structures. Considering both paradigms jointly will be envisaged by embedding topological priors / descriptors into deep-learning architectures.