The „Compressed Sensing“ of neuronal morphology

The most striking feature of neurons is their diverse morphology that manifests mainly in the variable number of dendrites and their branching pattern. The maturation of the dendritic arbor during development or its alterations during degenerative diseases influences the functional aspects of neurons.

Morphometric analysis of nerve cells during physiological conditions and impact of disease relevant factors were tested in mouse organotypic hippocampal slices. Cells were visualized by cLSM or 2P-microscopy after virus-mediated expression of EGFP.

Compressed Sensing in combination with adapted dictionaries such as wavelets and shearlets is a powerful method for data separation that had been previously used in astronomical image processing.

We were adopting this method to perform 3D reconstruction of neurons. The basic idea was to separate the micrograph in two new images, one containing the separated spines, the other the dendritic structures. This method succeeds provided that the structures to be separated are sufficiently morphologically distinct – a condition, which spines regarded as belonging to the class of point like structures and dendrites regarded as curve like structures do fulfill.

With the curve data we were able to reconstruct a three dimensional representation containing just the backbone of the neuron using an edge detection algorithm. These reconstructions were finally analyzed for their branching pattern and the thickness of dendrites in different locations.