Question icon
Your Current Search
Choose below to refine your search
Research Topic
Download abstract book

Download the NI2012 abstract book here. The page numbers in the index are clickable for easy browsing.


Robust automated protocol for extraction and comparison of single neuron morphology


Hiroyuki Ai (Fukuoka University), Stephan Haupt (University of Tokyo), Philipp Rautenberg (Ludwig-Maximilians-Universität München), Michael Stransky (Ludwig-Maximilians-Universität München), Thomas Wachtler (Ludwig-Maximilians-Universität München), Hidetoshi Ikeno (University of Hyogo)

Neuronal morphology is highly individual and a key element determining information processing and transmission in the brain. In order to carry out a systematic and theoretical analysis of neural mechanisms and to understand the role of individual neurons, it is necessary to construct models based on experimentally acquired neuronal branching patterns. We have developed a robust automated protocol for producing neuron models based on real neural morphologies acquired from confocal laser scan microscope (LSM) data.
LSM image stacks containing the entire morphology of single neurons are first subjected to a two-step segmentation. In the first step, brightness and contrast are adjusted to compensate for differences in noise and background levels among individual data sets, and binarization is applied. In the next step, extracted branching structures are traced based on the SSDT method using our software SIGEN (Yamazaki et al., 2008, doi: 10.1016/j.neucom.2005.12.042). SIGEN does not extract a wire model but also determines cylinder diameters for extracted segments. In this step, actual neuronal branch elements and false positive elements are still intermingled. Detected segments are then scrutinized and connected to the main branch based on two parameters, volume threshold (VT) and distance threshold (DT), finally resulting in cylinder models of the neurons. The final radius of cylinder elements corresponding to the thickness of a neurite segment is assigned by averaging the number of extracted pixels in a direction perpendicular to the skeleton center line within an element.
We applied our method to an identified interneuron in the honeybee auditory system. We compared the number of branches and estimated axial resistances of cylinder segments of neuron models extracted manually to results from our automated extraction protocol. We also investigated the effect of VT and DT on branch extraction success. The number of branches, especially in the fine dendritic areas, was clearly increased (up to 23%) by tuning of DT and VT. Our findings demonstrate how using well-defined parameters permits repeated and reproducible extraction of neuron morphologies and minimizes variability in reconstruction resulting from differences in the extraction process.
This research was supported by Strategic International Cooperative Program, Japan Science and Technology Agency (JST) and the German Federal Ministry of Education and Research (BMBF, grant 01GQ1116).
Preferred presentation format: Poster
Topic: Computational neuroscience

Andrew Davison
Andrew Davison says:
May 11, 2012 02:24 PM
Automated extraction of neuronal morphologies from imaging data will be a key technology in coming years, and this seems like a solid approach.