Synchrotron X-ray multi-spectral imaging is a novel imaging modality that may allow tracking cells at high resolution in small animal models. The data volume generated by such technique can be of hundreds of Gigabytes for one animal. Automatic, robust and rapid pipeline is therefore of paramount importance for large-scale studies. The goal of this article is to present a full image analysis pipeline ranging from the CT reconstruction up to the segmentation of nanoparticleslabeled-cells. Experimentally, rats that had received an intracerebral transplantation of gold nanoparticles-labeled cells were imaged in vivo in phase contrast mode (propagation-based imaging technique) at two different energies strategically chosen around the k-edge of gold. We apply a dedicated phase retrieval technique on each projection (out of 2000 for complete 2π rotation) before CT reconstruction. Then, a rigid registration is performed between the images below and above k-edge for accurate subtraction of the two data sets, leading to gold concentration maps. Due to the large number of specimens, the registration is based on the automatic segmentation of the cranial skull. Finally, an automatic segmentation of gold-labeled cells within the brain is performed based on high spots of gold concentrations. An example of an in-vivo data set for stroke cell therapy is presented.

No abstract