Multi-scale analyses

Multiscale image modelling. We will develop hierarchical (“tree”) image models dedicated to multiscale modelling of the 3D synchrotron images. We will investigate two families of models, developed in the framework of mathematical morphology: component-tree and the tree of shapes models and binary partition trees models. In this context, we will investigate potential ways to take advantage of distributed architecture to develop such approaches. Our objective is here to propose a toolbox composed of various tree data-structures likely to model the images to be analysed at various scales of details.

Statistical analysis and modelling. We will first explore the data according to their tree data-structure while considering different criteria: per individual, considering different groups of individuals, depending on different co-variables such as age, diseases, section of the aorta sampled. The second step will be to extract quantitative and objective features (indicators) about the structural state of the analysed vessel segments that describe their level of degradation, directly related to the age and the mechanical characteristics of the tissues. These approaches will enable us to characterize local level of degradation at different scales, to analyse homogeneity of tissues, for each group and between groups according to co-variables, such as the age or data from the longitudinal follow-up and to obtain explainable description of the vessel tissue in relation with the indicators of interest. These developments will be done locally then upscaled to process large amount of data with the HPC Romeo platform.