Tiling morphologies and segmentation
A detailed quantitative analysis of image data often requires the segmentation to make structures measurable. This is particularly challenging for biological structures, such as tesserae, because their arrangements are typically 3-dimensional and complex (e.g. there can be hundreds of tiles on a single skeletal element, making manual segmentation infeasible). We are therefore developing new methods of automated segmentation of CT data that will require very little user interaction. Because the mineralized layer can be considered a 2-dimensional surface embedded in 3-dimensional space, we can create a simplified representation for better visualization and analysis, with tesserae represented as nodes in a wire mesh network. Derived parameters relating to tesseral size and shape can be mapped onto this representation, facilitating visual analysis, hypothesis testing and comparison with idealized tilings. Since tiling structures are frequent in biology, we believe that our methods can be applied to a variety of image data of natural systems.
For more information on the Zuse Institute’s involvement in the project please see: Website