Creating 3D surface models of spiny neurons
© 2022 EPFL
Understanding the functional aspects of our brain implies understanding the structure-function relationship of billions of individual interconnected neurons; a particularly convoluted problem as the structure of each neuron affects the whole network and its dynamics. Neurons have visually complex structures, multiple classes of shapes (morphologies) and electrophysiological classifications, making them complicated computational objects to study. Accurate visualization of simulation results depend on highly accurate “3D surface” (mesh) models, where each component can be mapped; however the generation of such models is challenging. Here, the EPFL Blue Brain Project presents a simple and effective method to create mesh models of spiny neurons (neurons with small spikes along their surface) from their corresponding morphologies.
Neuronal morphologies are either digitally traced from optical microscopy images, or digitally synthesized in computer simulations. In both cases the result is a skeleton of sample points from which the cell body and neuron branches can be reconstructed; neuronal morphologies are then processed to remove any artifacts that arise from the reconstruction process. Finally, reconstructed skeletons can be used to perform detailed compartmental simulations to model their electrical behavior.
Creating surface models of neurons from their morphological graphs is comparatively challenging due to their complex branching structures. “Neurons are characterized by an extremely large extent for a very small total volume” says author Marwan Abdellah, Scientific Visualization Expert at the BBP, “Obtaining realistic cell body shapes and having smooth and continuous connection with the branches is also a major concern”.
Some methods ensure that meshes are topologically-optimized and watertight. Others are more concerned with creating smooth, realistic and low-tessellated surfaces that can be used to visualize simulated activity and give high-quality content for scientific dissemination. These methods can be difficult to implement, have a limited branching and cell body quality, and may take impractically long computing times. Others simply lack the ability to reconstruct neuron shapes with geometrically accurate branching.
In this work, the Blue Brain Visualization team presents a simple method for reconstructing surface meshes of spiny neurons from their morphological graphs using so-called union operators. Prof. Felix Schürmann, Blue Brain Computing Director, explains the advantages of the approach: “The method enables the generation of smooth surfaces with plausible cell bodies and natural-looking branch geometries allowing high quality scientific media production. We are happy to be able to make the implementation freely accessible and available within the previously released NeuroMorphoVis package, which is an already widely used tool within the neuroscience community for visualization and analysis” he concludes.
The paper was presented in the Eurographics UK Computer Graphics and Visual Computing CGVC 2022 conference which was held virtually, and published in the Eurographics digital library.
Abdellah, M., Garcia Cantero, J. J., Foni, A., Román Guerrero, N., Boci, E., & Schürmann, F. (2022). Meshing of spiny neuronal morphologies using union operators. In P. Vangorp & M. J. Turner (Eds.), Computer Graphics and Visual Computing (CGVC) conference proceedings (Graphics section). The Eurographics Association, UK. https://doi.org/10.2312/cgvc.20221168