The research fields Human Brain Mapping and Connectome make use of different 2D and 3D imaging modalities, including ultra-high resolution images based on histological sectioning. Reconstruction of large, consistent 3D volumes is a basic requirement for structural and functional analysis of such data.
Our project group works on segmentation, registration and reconstruction of high-resolution multimodal image data. Key problems are anatomically consistent elastic registration to compensate for nonlinear defects and artefacts caused by the sectioning process, multiscale correspondence problems and the development of statistical methods for reconstruction and inference from such large and difficult datasets.
Considering that the size of typical datasets exceeds the capacity of conventional computers even today, we develop highly efficient parallel algorithms for use on high performance computing platforms, and investigate data storage and management solutions for datasets in the order of petabytes. This is done in close collaboration with the Jülich Supercomputing Center.