# A software tool for the analysis of neuronal morphology data

- Julia Ledderose
^{1}, - Luis Sención
^{4}, - Humberto Salgado
^{2}, - Oscar Arias-Carrión
^{3}and - Mario Treviño
^{4}Email author

**7**:6

https://doi.org/10.1186/1755-7682-7-6

© Ledderose et al.; licensee BioMed Central Ltd. 2014

**Received: **7 January 2014

**Accepted: **11 February 2014

**Published: **17 February 2014

## Abstract

Anatomy plays a fundamental role in supporting and shaping nervous system activity. The remarkable progress of computer processing power within the last two decades has enabled the generation of electronic databases of complete three-dimensional (3D) dendritic and axonal morphology for neuroanatomical studies. Several laboratories are freely posting their reconstructions online after result publication *v.gr.* NeuroMorpho.Org (*Nat Rev Neurosci*
**7:**318–324, 2006). These neuroanatomical archives represent a crucial resource to explore the relationship between structure and function in the brain (*Front Neurosci*
**6:**49, 2012). However, such 'Cartesian’ descriptions bear little intuitive information for neuroscientists. Here, we developed a simple prototype of a MATLAB-based software tool to quantitatively describe the 3D neuronal structures from public repositories. The program imports neuronal reconstructions and quantifies statistical distributions of basic morphological parameters such as branch length, tortuosity, branch's genealogy and bifurcation angles. Using these morphological distributions, our algorithm can generate a set of virtual neurons readily usable for network simulations.

### Keywords

Neurons Dendrites Morphology Tortuosity Simulations## Results and discussion

Electronic databases of complete 3D dendrites constitute a valuable tool to explore the morphological structure of single neurons [1]. The data acquisition of those structures comprises a multi-step process from tissue collection and staining to the extraction of neuronal structural information via a variety of imaging techniques. To date, the majority of dendritic and axonal morphology reconstructions are based on bright-field microscopy mainly because of its broad compatibility with histological staining methods [2]. Digital tracing of neuronal morphology converts large amounts of imaging information into a simple and compact representation which can be easily visualized, quantified, archived, and shared [3], thus maximizing the opportunity to exploit the full potential of unrestricted morphometric analyses [1, 4].

There are multiple ways to digitize neuronal morphology once it has been visualized by optical microscopy. One effective way to describe the treelike branching of axons and dendrites can be achieved by using a sequence of interconnected cylinders. In this 'vector' representation, each uniform segment in the arbor can be characterized by five values, consisting of the three Euclidean 'x', 'y' and 'z' coordinates, a diameter of its ending location, and the identity of the 'parent' segment from which each new segment originates. Thus, by definition, compartments are segments represented as cylinders with a given diameter and the coordinates of the extreme points. Branches are formed with one or more compartments between the soma, the bifurcations, and the tips. Bifurcations are defined as the points where a branch splits into 'daughter' branches.

This 'Cartesian' description of neuronal structures constitutes a complete mapping of dendritic morphology but bears little intuitive information. To extract quantitative measures of neuronal morphology, we developed a software tool written in MATLAB (MATLAB R2012a, MathWorks, Inc.) that reads these 3D dendritic reconstructions and computes morphological parameters from a large and representative set of neurons. This tool is freely available upon request.

To implement and test our algorithms, we used digitally reconstructed hippocampal neurons from different repositories, grouped as follows: Group 1: 6 Dentate Gyrus 'aged' Granule Cells (referenced as n270-n275 from the Duke-Southampton archive; http://neuron.duke.edu), Group 2: 4 CA3 'young' Pyramidal Cells (referenced as l10, 148a, 160b and 164 from the Duke-Southampton archive), Group 3: 15 CA1 'aged' Pyramidal Cells (referenced as n170-n184 from the Duke-Southampton archive), and Group 4: 18 CA1 'young' Pyramidal Cells (referenced as a series of pyramidal cells from the Gulyás CA1 repository; http://www.koki.hu/~gulyas/).

Our software imports the cells, which are stored in a non-proprietary *.swc format including a header of comments where each line is preceded by a '#' sign. These lines describe the program used for neuron tracing (usually Neurolucida, MicroBrightField, Inc.), localization of the neuron (region, Field/Layer, etc.), type of cell, contributor, reference, soma area, shrinkage correction, number and date versions. Followed by these remarks, the *.swc file consists of a [*n x 7*] matrix which contains the following fields: Index (Column 1), a user defined flag denoting the specific part of the structure (cell body, apical dendrites, basilar dendrites and axon; Column 2), 3D coordinates (x, y, z, in μm; Columns 3–5), radius (r, in μm; Column 6), and parent index (Column 7). As two points connected by a straight line constitute a segment, then each neuronal reconstruction with *n* points has *n-1* total segments, where *n* is the maximum row size of the matrix. The import file function deletes the header and stores the coordinates into a [*n x 7*] matrix in MATLAB’s workspace for further analysis. In all cases the 'x', 'y' and 'z' values were corrected for shrinkage and lens medium refraction.

_{ i },y

_{ i },z

_{ i }) with (x

_{ j },y

_{ j },z

_{ j }) where '

*i*' is the parent index and '

*j*' is the index. A branch is therefore conformed by one or more segments. Two connected segments belong to the same branch only if: i) the parent index of the second segment equals the index of the first segment, ii) both segments share the same structure flag stored in column 2 and iii) there is no other segment in the whole matrix that has the same parent index. If these restrictions applied, a sequential numerical label was assigned to the two segments in order to specify that all segments with the same name belonged to the same branch. At the end of this procedure, the set of rows that share the same first-name conform a branch whose first-name is a unique numerical label (Figure 1A-C). Once each branch has its own first-name it is possible to track the path of its ancestors back to the soma (Figure 1D). Consequently, the 'full-name' of a branch is computed as follows: the parent index in each row points to another row, which belongs to a branch with an already assigned first-name. Unless both rows belong to the same branch (

*i.e.*having the same first-name), this name corresponds to the row’s parent name. The row’s parent-name (

*i.e.*a full name) is stored and this process is repeated until the whole branching genealogy is generated. Thus, at the end of this process, each row has an associated full-name which can be traced backwards to its ancestors (Figure 1D). Here again, the set of rows that share the same full-name conform a branch. Full-name length (number of ancestors+1) equals the path-length and this measure was further used to compute topographic measurements (see below). Our algorithm successfully identified branches and assigned a full-name to each of them. Plots assigning a pseudo-random color to each branch confirmed that all branches were correctly identified, including those that generated dichotomous bifurcations (Figure 1E). Additionally, the plotter was able to represent vector’s width equal to segment’s diameter and/or assign four different colors depending on the specific cell structure (cell body, apical dendrites, basal dendrites or axon) according to the user-defined flag stored in the second column of the matrix (not illustrated).

The analyzed cells from the Duke-Southampton archive exhibited a tendency to deviate repeatedly from the straight direction and return to the initial orientation after considerable meandering ('zigzags' towards the z-axis, Figure
2A, B). It is known that the acquisition and assembly procedures introduce morphological noise in any representation of digitized neurons
[9–11], which makes it difficult to carry out a meaningful statistical analysis. We implemented a feature to perform z-coordinate smoothing aimed to diminish morphological parameter miscalculations due to such extensive amount of noise in the 'z' axis (Figure
2B, C). Specifically, this function smoothes the 'z' data of each branch using a moving average filter. As the spatial distribution of points defining segments is not uniform, their distance projected in the 'xy' plane is used as predictor data for the z-smoothing. However, applying this procedure to each independent branch would result in local 'z-jumps' at bifurcations and spatial continuity between related branches is required. To solve this problem, the smoothing was performed concatenating two additional points located at the branch’s endings: the first point given by an average of the final z-coordinate of the parent plus the initial z-coordinate of the sister. The latter represents the average of the initial z-coordinates of all daughters (if present). Aside from these issues, it is important to mention that the criteria for sampling data points for a morphological structure are subjective and to some extent arbitrary. Due to the complexity of dendritic morphology, the very same neurons mounted on microscope slides, and traced by different researchers or on different reconstruction systems, can result in considerably different digital files
[10]. In this context, the issue of quality control for morphological data is extremely important and should be taken carefully into consideration in any morphological study before interpreting the results. In other words, digital files of dendritic morphology are rarely accurate representations of biological structures; they constitute only an approximation of the neuron. Nevertheless, if a digital data set is internally consistent (*v.gr.* correct indexing, no '0-length segments', etc.), then the mathematical problem of its quantitative representation is independent on the data quality.

*m*come together, the third edge contiguous to that branch belongs to a branch of order [

*m+1*]. When a branch of order

*m*meets a branch of order

*n*, where

*n*exceeds

*m*, the third edge at the branch is a continuation of the branch of order

*n*(Figure 3A). There are no arbitrary decisions in Strahler ordering, and it is purely topological. We tested the Strahler ordering function for correct indexing with a reconstructed neuron. Interestingly, all the pyramidal CA1 cells from the Gulyás repository had very low maximum Strahler indexes. This can be explained by the high number of terminal branches along their paths (Figure 3B).

*i.e.*along segments) and the linear distance between endpoints, whereas the sum of angles metric (SOAM) tortuosity integrates total curvature along a curve and normalizes it by path length [14, 15], thereby handling tight coils better. Thus, for any point P

_{k}we defined the vectors T

_{1}= P

_{k}-P

_{k-1}, T

_{2}= P

_{k+1}-P

_{k}, and T

_{3}= P

_{k+2}-P

_{k+1}. The in-plane angle at point P

_{k}(IP

_{k}) and the torsional angle (TP

_{k}) were given by the following equations, where TP

_{k}, IP

_{k}∈ [0,π]:

Note that this expression normalizes the total curvature with respect to total curve length. This means that SOAM values can be compared between two branches of different length.

Bifurcation angles are taken as the angle between a daughter and its parent branch. To compute bifurcation angles between a daughter and its parent branch it is first necessary to calculate directional vectors at the beginning and at the end of each branch. The bifurcation angle between two branches does not only depend on its terminal segments. For this reason, when computing directional vectors it is necessary to consider several segments and then proceed with a 3D linear regression from where a normalized directional vector will be extracted. We used five segments per branch for directional vector computation (although we believe that the number of segments used should be justified in terms of their tortuosity).

Once accumulative morphological parameters are stored in the last row of each branch, it is possible to make a 'vertical compression' to produce a new matrix with each row corresponding to information from a branch with its respective accumulative morphological parameters.

Elucidating the complex organization of the brain will require synthesis of information about neuron types, the spatial patterns of their dendritic and axonal arborizations, cell numbers and densities, as well as synapse number and location [17, 18]. In the central nervous system, the shape of the dendritic arbor is related to the cell-type specificity and to the large number of synaptic inputs. The extent of dendritic arbors, at least in sensory neurons of the peripheral nervous system, physically defines their receptive fields [19], and axonal topology is known to affect synaptic output [20]. Discoveries that many dendrites conduct input signals actively, back-propagate action potentials, and integrate synaptic inputs by means of time-dependent nonlinear summation provide indisputable evidence that dendritic morphology is a key aspect of the neuronal machinery underlying signal processing and integration [21]. Dendritic structure contributes significantly to neuronal information processing [22, 23] and computational models have shown that dendritic geometry can be responsible for producing an entire spectrum of firing patterns displayed across different cortical neuronal types [24], and also within a single class of hippocampal neurons [25]. The importance of dendrites for neuronal activity is evidenced by the influence of dendritic morphology on network connectivity [26] as it is constantly reshaped by the dynamic remodeling of both dendrites and axons, which is crucial in determining the pattern of synaptic formation among neurons [27].

Here, we developed a prototype of a MATLAB based software package to characterize neuronal dendrites on the basis of the statistical distributions of morphological parameters. From a merely morphological point of view and assuming that cells located on a specific site and under strict experimental conditions share similar morphological properties, the neuroanatomy of a cell class can be measured and compressed by quantifying statistical distributions of relevant morphological parameters. Such an approach is important for understanding the heterogeneity of the different neuronal groups, as well as for unveiling the relationship between neuronal structure and function. Hence, this tool can be applied for comparative anatomy, developmental neurobiology and medical diagnosis [4]. The resulting statistical descriptions of neuronal morphology can be further used to create an unlimited number of non-identical virtual neurons (data not shown). Virtual generation of axonal and dendritic arbors is useful to explore mechanisms of growth [28, 29] and to construct biologically realistic neural networks [28, 30].

## Declarations

### Acknowledgments

We thank Dr. Esmeralda Matute Villaseñor for constant support. M.T. was supported by the 'Programa de Repatriación' from CONACyT.

## Authors’ Affiliations

## References

- Ascoli GA:
**Mobilizing the base of neuroscience data: the case of neuronal morphologies.***Nat Rev Neurosci*2006,**7:**318–324.PubMedView ArticleGoogle Scholar - Halavi M, Hamilton KA, Parekh R, Ascoli GA:
**Digital reconstructions of neuronal morphology: three decades of research trends.***Front Neurosci*2012,**6:**49.PubMed CentralPubMedView ArticleGoogle Scholar - Meijering E, Jacob M, Sarria JC, Steiner P, Hirling H, Unser M:
**Design and validation of a tool for neurite tracing and analysis in fluorescence microscopy images.***Cytometry A*2004,**58:**167–176.PubMedView ArticleGoogle Scholar - Costa LF, Zawadzki K, Miazaki M, Viana MP, Taraskin S:
**Unveiling the neuromorphological space.***Front Comput Neurosci*2010,**4:**150.PubMed CentralView ArticleGoogle Scholar - Bradke F, Dotti CG:
**Establishment of neuronal polarity: lessons from cultured hippocampal neurons.***Curr Opin Neurobiol*2000,**10:**574–581.PubMedView ArticleGoogle Scholar - Trevino M, Vivar C, Gutierrez R:
**Beta/gamma oscillatory activity in the CA3 hippocampal area is depressed by aberrant GABAergic transmission from the dentate gyrus after seizures.***J Neurosci*2007,**27:**251–259.PubMedView ArticleGoogle Scholar - Trevino M, Vivar C, Gutierrez R:
**Excitation-inhibition balance in the CA3 network–neuronal specificity and activity-dependent plasticity.***Eur J Neurosci*2011,**33:**1771–1785.PubMedView ArticleGoogle Scholar - Kulik A, Vida I, Lujan R, Haas CA, Lopez-Bendito G, Shigemoto R,
*et al*.:**Subcellular localization of metabotropic GABA(B) receptor subunits GABA(B1a/b) and GABA(B2) in the rat hippocampus.***J Neurosci*2003,**23:**11026–11035.PubMedGoogle Scholar - Horcholle-Bossavit G, Gogan P, Ivanov Y, Korogod S, Tyc-Dumont S:
**The problem of the morphological noise in reconstructed dendritic arborizations.***J Neurosci Methods*2000,**95:**83–93.PubMedView ArticleGoogle Scholar - Kaspirzhny AV, Gogan P, Horcholle-Bossavit G, Tyc-Dumont S:
**Neuronal morphology data bases: morphological noise and assesment of data quality.***Network*2002,**13:**357–380.PubMedView ArticleGoogle Scholar - Hamam BN, Kennedy TE:
**Visualization of the dendritic arbor of neurons in intact 500 microm thick brain slices.***J Neurosci Methods*2003,**123:**61–67.PubMedView ArticleGoogle Scholar - Horton RE:
**Erosional development of streams and their drainage basins.***Bull Geological Soc Am*1945,**56:**275–370.View ArticleGoogle Scholar - Strahler AN:
**Hypsometric analysis of erosional topography.***Bull Geol Soc Am*1952,**63:**1117–1142.View ArticleGoogle Scholar - Bullitt E, Gerig G, Pizer SM, Lin W, Aylward SR:
**Measuring tortuosity of the intracerebral vasculature from MRA images.***IEEE Trans Med Imaging*2003,**22:**1163–1171.PubMed CentralPubMedView ArticleGoogle Scholar - Hart WE, Goldbaum M, Cote B, Kube P, Nelson MR:
**Measurement and classification of retinal vascular tortuosity.***Int J Med Inform*1999,**53:**239–252.PubMedView ArticleGoogle Scholar - Rall W:
**Branching dendritic trees and motoneuron membrane resistivity.***Exp Neurol*1959,**1:**491–527.PubMedView ArticleGoogle Scholar - DeFelipe J:
**From the connectome to the synaptome: an epic love story.***Science*2010,**330:**1198–1201.PubMedView ArticleGoogle Scholar - Parekh R, Ascoli GA:
**Neuronal morphology goes digital: a research hub for cellular and system neuroscience.***Neuron*2013,**77:**1017–1038.PubMedView ArticleGoogle Scholar - Hall DH, Treinin M:
**How does morphology relate to function in sensory arbors?***Trends Neurosci*2011,**34:**443–451.PubMed CentralPubMedView ArticleGoogle Scholar - Sasaki T, Matsuki N, Ikegaya Y:
**Effects of axonal topology on the somatic modulation of synaptic outputs.***J Neurosci*2012,**32:**2868–2876.PubMedView ArticleGoogle Scholar - Eilers J, Konnerth A:
**Dendritic signal integration.***Curr Opin Neurobiol*1997,**7:**385–390.PubMedView ArticleGoogle Scholar - Mel BW, Ruderman DL, Archie KA:
**Translation-invariant orientation tuning in visual "complex" cells could derive from intradendritic computations.***J Neurosci*1998,**18:**4325–4334.PubMedGoogle Scholar - Stuart G, Spruston N, Sakmann B, Hausser M:
**Action potential initiation and backpropagation in neurons of the mammalian CNS.***Trends Neurosci*1997,**20:**125–131.PubMedView ArticleGoogle Scholar - Mainen ZF, Sejnowski TJ:
**Influence of dendritic structure on firing pattern in model neocortical neurons.***Nature*1996,**382:**363–366.PubMedView ArticleGoogle Scholar - Krichmar JL, Nasuto SJ, Scorcioni R, Washington SD, Ascoli GA:
**Effects of dendritic morphology on CA3 pyramidal cell electrophysiology: a simulation study.***Brain Res*2002,**941:**11–28.PubMedView ArticleGoogle Scholar - van Ooyen A, Duijnhouwer J, Remme MW, van Pelt J:
**The effect of dendritic topology on firing patterns in model neurons.***Network*2002,**13:**311–325.PubMedView ArticleGoogle Scholar - Wong WT, Wong RO:
**Rapid dendritic movements during synapse formation and rearrangement.***Curr Opin Neurobiol*2000,**10:**118–124.PubMedView ArticleGoogle Scholar - Eberhard JP, Wanner A, Wittum G:
**NeuGen: a tool for the generation of realistic morphology of cortical neurons and neural networks in 3D.***Neurocomputing*2006,**70:**327–342.View ArticleGoogle Scholar - van Ooyen A:
**Using theoretical models to analyse neural development.***Nat Rev Neurosci*2011,**12:**311–326.PubMedView ArticleGoogle Scholar - Koene RA, Tijms B, van Hees P, Postma F, de Ridder A, Ramakers GJ,
*et al*.:**NETMORPH: a framework for the stochastic generation of large scale neuronal networks with realistic neuron morphologies.***Neuroinformatics*2009,**7:**195–210.PubMedView ArticleGoogle Scholar

## Copyright

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.