3d virtual histology of neuronal tissue by propagation-based x-ray phase-contrast tomography

Deciphering the three-dimensional (3d) cytoarchitecture of neuronal tissue is an important step towards understanding the connection between tissue function and structure and determining relevant changes in neurodegenerative diseases. The gold standard in pathology is histology, in which the tissue is examined under a light microscope after serial sectioning and subsequent staining. It is an invasive and labor-intensive technique which is prone to artifacts due to the slicing procedure. While it provides excellent results on the 2d slices, the 3d anatomy can only be determined after aligning the individual sections, leading to a non-isotropic resolution within the tissue. X-ray computed tomography (CT) offers a promising alternative due to its potential resolution and large penetration depth which allows for non-invasive imaging of the sample's 3d density distribution. In classical CT, contrast formation is based on absorption of the x-rays as they pass through the sample. However, weakly absorbing samples like soft tissue from the central nervous system give nearly no contrast. By exploiting the much stronger phase shifts for contrast formation, which the sample induces in a (partially) coherent wavefront, it can be substantially increased. During free-space propagation behind the sample, these phase shifts are converted to a measurable intensity image by interference of the disturbed wave fronts. In this thesis, 3d virtual histology is performed by means of propagation-based x-ray phase-contrast tomography on tissue from the central nervous system of humans and mice. A combination of synchrotron-based and laboratory setups is used to visualize the 3d density distribution on varying lengths scales from the whole organ down to single cells. By comparing and optimizing different preparation techniques and phase-retrieval approaches, even sub-cellular resolution can be reached in mm-sized tissue blocks. The development of an automatic cell segmentation workflow provides access to the 3d cellular distribution within the tissue, enabling the quantification of the cellular arrangement and allowing for extensive statistical analysis based on several thousands to millions of cells. This paves the way for biomedical studies aimed at changes in cellular distribution, e.g., in the course of neurodegenerative diseases such as multiple sclerosis, Alzheimer's disease or ischemic stroke.