Speaker: Jakob Troidl (193-02 Computer Graphics)

Recent high-resolution electron microscopy imaging allows neuroscientists to reconstruct not just entire cells but individual cell substructures (i.e., cell organelles) as well. Based on these data, scientists hope to get a better understanding of brain function and development through detailed analysis of local organelle neighborhoods.

However, in-depth analyses require efficient and scalable comparison of a varying number of cell organelles, ranging from two to hundreds of local spatial neighborhoods. Scientists need to be able to analyze the 3D morphologies of organelles, their spatial distributions and distances, and their spatial correlations. This thesis's central premise is that it is hard to provide a one-size-fits-all comparative visualization solution to support the given broad range of tasks and scales. To address this challenge, we have designed NeuroKit as an easily configurable toolkit that allows scientists to customize the tool's workflow, visualizations, and supported user interactions to their specific tasks and domain questions.
Furthermore, NeuroKit provides a scalable comparative visualization approach for spatial neighborhood analysis of nanoscale brain structures. NeuroKit supports small multiples of spatial 3D renderings as well as abstract quantitative visualizations, and arranges them in linked and juxtaposed views. To adapt to new domain-specific analysis scenarios, we allow the definition of individualized visualizations and their parameters for each analysis session. This configurability is tied in with a novel scalable visual comparison approach that automatically adjusts visualizations based on the number of structures that are being compared. We demonstrate an in-depth use case for mitochondria analysis in neuronal tissue and analyze the usefulness of NeuroKit in a qualitative user study with neuroscientists.




20 + 20
Supervisor: Eduard Gröller