- Linux: Most linux distributions contain versions of these packages in their package manager that can readily be installed.
- Windows and Mac: There are "scientific python" bundles that are shipped with all the aforementioned packages; most famously from Anaconda or Enthough. Alternatively check the Ipython installation.
The following instructions are for Linux and Max OSX systems and only use command line tools. Please follow the appropriate manuals for Windows systems or tools with graphical interfaces.
Check out the git repository and adjust your
git clone https://bitbucket.org/btorb/btmorph.git cd btmorph export PYTHONPATH=$(pwd):$PYTHONPATH
The above commands will temporarily set your
$PYTHONPATH. Add the appropriate path in your
.bashrc to make add the package permanently.
Test the installation by running the tests (see :ref:`unit_testing`):
nosetests -v --nocapture tests/structs_test.py nosetests -v --nocapture tests/stats_test.py
Neurons not only look like the branching of trees, their structure is, mathematically speaking a tree structure because they can be represented as graphs without cycles. More precisely, when disregarding the soma, a neuron is a binary tree. That is a tree with at most two children at any node. As such, a tree data structure provides an intuitive representation of a morphology and can be easily probed to calculate morphometric features.
The tree is implemented as a linked list data structure (STree2). Each item in the list/tree is a node (SNode2) and contains pointers to its parent (
get_parent) and its children (
get_children). Each node can store something in its designated
content container. By design, the content is a Python dict and in this library it has at least one key:
'p3d', a P3D2 object. Obviously, this tree data structure resembles strongly the structure of an SWC file.
Schematically, it looks like this:
A small set of library containing an efficient data structure and routines to quickly analyze morphometric features of neuronal morphologies.
The internal representation is based on a tree data-structure (rather than an adjacency matrix as in the TREES toolbox).
Atomic functions are provided to allow usage in scripting and enable the user to built more complex morphometrics on top of the provided functionality. The code is serial (i.e., non-parallel) because single neuron morphometrics are fast to compute. When analyzing a batch of morphologies a parallel wrapper can be written (e.g., using Pythons's multiprocessing module or more fancy using MPI).
The input is a digital representation of a neuronal morphology in the SWC format. This is the current de facto format also used on the curated NeuroMorpho.org website.org database. It is expected to use the standardized SWC-format that follows the three-point soma description (see here). Analysis is based on the whole neuron but subtrees can be selectively analyzed based on the value of the SWC-type field.
Morphometrics can be either scalar (= one value per morphology) or vector / distributed (= a distribution of values per morphology). For vector morphometrics, the features can be measures either a branching point, terminal points or both. Other 'points' specified in the SWC file are only used for the internal representation of the geometry.
Simple wrappers are provided to analyze single neurons, populations thereof and compare two populations.
- Scalar: (one per morphological structure under scrutiny)
- total size: total length of the neurite
- # stems
- # branch points
- # terminal points
- width (without translation; absolute coordinates; potential extension along the first 3 principal components)
- max degree (of neurites sprouting at the soma)
- max order (of neurites sprouting at the soma)
- partition asymmetry (can/cannot be measured at the soma?)
- Vector: (for each point, bifurcation point or terminal point):
- segment path length (incoming)
- segment euclidean length (incoming)
- contraction (euclidean / path; incoming)
- partition asymmetry
- fractal dimension (of path between soma and PoI)
- Clouds: save x,y,z coordinates for post-hoc histograms analysis or other scalar (e.g., moments) or vector properties (e.g., PCA)
(simple, using matplotlib):
- 2D/3D plot as wires and/or with diameters
- Three 2D projections for improved visual inspection
In the top directory of the package (
ipython --pylab and issue the command below.
ipython you can use the magic function
%paste to paste a whole code block. Copy the code below and type
%paste at the
import btmorph import numpy import matplotlib.pyplot as plt swc_tree= btmorph.STree2() swc_tree.read_SWC_tree_from_file("examples/data/v_e_moto1.CNG.swc") stats = btmorph.BTStats(swc_tree) # get the total length total_length = stats.total_length() print "total_length = %f" % total_length # get the max degree, i.e., degree of the soma max_degree = stats.degree_of_node(swc_tree.get_root()) # generate and save the dendrogram btmorph.plot_dendrogram("examples/data/v_e_moto1.CNG.swc") plt.savefig('examplar_dendrogram.pdf')
|||Cannon et al. An online archive of reconstructed hippocampal neurons., J. Neurosci. methods (pubmed http://www.ncbi.nlm.nih.gov/pubmed/9821633).|
If you use this software, please cite the following peer-reviewed news item published in the Neuroinformatics journal.
B. Torben-Nielsen, An efficient and extendable Python library to analyze neuronal morphologies. Neuroinformatics, 2014, online first (<a href "http://link.springer.com/article/10.1007/s12021-014-9232-7">here</a>)