Neuron Tracing
Kalliopi Arkoudi

Kalliopi Arkoudi

Applications Development Engineer
ZEISS Microscopy

M. Abbate

Maurizio Abbate

Senior Application Engineer
ZEISS Microscopy

D. Wiles

David Wiles

Application Engineer
ZEISS Microscopy

From Image to Results

Neuron Tracing

In this series "From Image to Results", explore various case studies explaining how to reach results from your demanding samples and acquired images in an efficient way. For each case study, we highlight different samples, imaging systems, and research questions. In this case study, we trace neurons.

Neuron tracing maps the pathway of neuronal processes, providing critical information to understand cell function or to uncover insights when comparing healthy cells to diseased states. In this case study, neuron tracing is performed automatically using software visualization and then segmentation of neuronal cell bodies imaged using confocal laser scanning microscopy of a large brain slice. Subsequent data analyses are also performed to provide information including neurite number, branch length and tortuosity.

Case Study Overview


GFP-M mouse brain​


Segment the neuronal structures in a large dataset.


Statistical analysis of segmented dataset​.




ZEISS arivis Pro​


Cartoon schematic of antibody internalisation and endolysosomal trafficking

Figure 1: Basic morphology of a bipolar neuron. The key structures Axon, Cell Body and Dendrites are highlighted. 3D-digital reconstruction of the neuronal tree.

Neurons are highly specialized cells responsible for the transmission of electrical and chemical messages across the nervous system. The key morphological parts of an individual neuron are the cell body, the dendrites, and the axon. The dendrites, often referred to as a dendritic tree due to their arborization pattern, are responsible for receiving information from surrounding neurons or sensory cells, integrating the signals, and then transmitting them to the cell body. The degree of arborization of the dendrite and its location is critical as neurons need to be extremely close to transmit signals to each other. The axon is then responsible for the electrical signal initiation and conduction away from the cell body and towards other neurons or targets, such as muscles or glands. The axonal terminal ends at the synapse, the junction between the neuron and its target. The number of axonal terminals and synapses that a neuron creates depends on the neuronal subtype and its connectivity with the environment. Understanding neuronal structures are critical to understanding specialized cell functions or when comparing healthy versus diseased states.

In this case study, the Automatic Neuron Tracer in ZEISS arivis Pro software is used for neuron tracing to create three-dimensional digital reconstructions of neuronal structures, in a large group of neurons, as well as to automatically quantify several neuronal morphological parameters.

Material and Methods

Sample Preparing and Imaging Data Collection

A Thy1-GFP-Mouse brain was perfused and fixed with PFA 4%. After fixation, the brain was sliced with a vibratome into 500 μm thick slices and mounted with Fluoromount G. Imaging of the sample was performed with the ZEISS LSM 980 laser scanning confocal with the LD 40x LCl Plan Apochromat Autocorr 1.2 NA objective using glycerol as an immersion media. The dataset was captured using the preset 1.2x sampling LSM Plus settings to maximize the relationship between image quality and acquisition speed (imaging settings: pixel number 4276 x 4276 px, zoom 0.6x, frame time: 10.51 s, pixel dwell time: 0.49 μs). The size of the z-stack was 1,471 slices. Upon acquisition the dataset was processed with LSM Plus and transferred to ZEISS arivis Pro software for further analysis.

Cell Body Segmentation Pipeline

Cell Body Segmentation Pipeline

Neuron Tracing

Cell Body Segmentation

The unprocessed image data was imported into ZEISS arivis Pro, and the dataset was visualized in 3D. While the dataset can be analyzed at native data resolution (100%), for this application the resolution was scaled down to 25%. Although this reduces the image details, it increases analysis speed and does not affect the quality of the segmentation trace.

The first part of the processing applies a Denoising function to reduce the image noise. Next, two morphology filters are run in sequence. Morphological filters are non-linear operations used to change the morphology of features in an image, such as enlarging or shrinking structures in the image. The erosion operation erodes away the boundaries of regions of foreground pixels, resulting in areas of foreground pixels shrinking in size. The dilation operation gradually enlarges the boundaries of regions of foreground pixels, resulting in areas of foreground pixels enlarging. The combined application of these two filters, results in the erosion of the neurite structures to highlight the cell body, which is returned to its original size by dilation. The result of the processed image is saved for future use and analysis. The Intensity Threshold Segmenter is then applied to the processed image to create segments based on a specific range of intensities. A Splitting operation follows to divide cell bodies touching each other. The cell body segments will be used as starting points for the automatic Neuron Tracer task, which is the core of the application.

Neurite Segmentation Pipeline

Neurite Segmentation Pipeline

Neurite Segmentation

The next step is the tubular shape segmentation of the neurites with the Automatic Neuron Tracer. The Neuron Tracer can either use cell bodies as a starting point if the cell body is already segmented or start without an explicit starting point if a cell body is not available.

The automatic tracing in ZEISS arivis Pro can be done with two different published peer-reviewed algorithms. The two reconstructors differ in their complexity and each is optimal for different type of datasets.
The Threshold-based Reconstructor1 applies a threshold to separate the foreground and background pixels. Then the foreground is searched for connected pixels. These connected pixels are formed into paths and only paths that form a neuron skeleton structure will be kept and used to create a complete trace. The Threshold-based approach is less computationally demanding and thus recommended for larger datasets. Additionally, the Threshold based reconstructor relies heavily on the quality of the data, since it cannot connect gaps in fluorescence resulting from low image quality.

The Probabilistic Reconstructor2 uses a calculation to determine the local tubularity of the image data. This tubularity map is searched for seed points that are used as a starting point for applying a probability function (Monte-Carlo) to detect the trace parts. These trace parts are then merged to create the complete trace. The Probabilistic approach is much more precise in branch detection and can close gaps in traced structures. This results in the algorithm being more computationally intense and more complex to set up.

For this dataset, the Threshold-based reconstructor algorithm was selected due to the large size and the high quality of the dataset. To initiate the algorithm, both the inclusion threshold and the minimum branch length filter must be set-up. Both settings are interactive with the 2D image when the preview feature is selected. The pipeline is executed, and the trace is inspected by the user for validation. Modifications of the reconstructor parameters or manual editing can be used to improve the quality of the trace.

Modalities for Trace Editing

Modalities for Trace Editing

Trace Editing

The neurite trace can be edited in 2D or 3D using the Trace Tool. Trace editing allows the user to join or split traces that are incorrectly created. Skeleton view is used to visualize the trace while editing, allowing you to draw, cut, or merge traces with interactive visualization of the changes in the viewer. Additional traces can be added and connected to pre-existing ones. Once the changes are completed, they can be confirmed by selecting the Apply Changes button.

Software Processing

Neuron Tracing Tutorial

In this video tutorial for neuron tracing, learn how to segment cell bodies and trace neurites with Automatic Neuron Tracer in ZEISS arivis Pro. The operations discussed include Denoising, Morphology filters, Intensity Thresholder, Splitting, and the Threshold-based reconstruction functions of the Automatic Neuron Tracer. Explore the different tools available for setting up the filters.

Neuron Trace Editing Tutorial

In this video tutorial, inspect the trace created, and learn how to adjust the parameters or manually edit the trace.

ZEISS arivis Pro Logo
ZEISS arivis Pro

ZEISS arivis Pro

ZEISS arivis Pro is a modular software for working with multi-channel 2D, 3D and 4D images of almost unlimited size, highly scalable and independent of local system resources. Many modern microscope systems such as high-speed confocal, light sheet / SPIM, super-resolution, electron microscopy or X-ray instruments can produce huge amounts of imaging data. ZEISS arivis Pro handles such datasets without constraints and in relatively short time.


Figure 2: 3D rendering digital reconstruction of traced neuron trees.

Neuronal morphology and visualization

Once the analysis is completed, the reconstruction can be visualized in 3D (Figure 2). The digital reconstruction of neurons can be overlaid with the imaging dataset, allowing correlation of the morphological features of neurons with their functional properties.

Figure 3: Morphology of individual 3D reconstructed neurons.
Figure 3: Morphology of individual 3D reconstructed neurons.

Figure 3: Morphology of individual 3D reconstructed neurons.

Figure 3: Morphology of individual 3D reconstructed neurons.

Individual segmented neurons can be visualized and inspected to evaluate differences in the arborization pattern (Figure 3). Analysis of this pattern can uncover insights into neuronal circuitry and connectivity.

A variety of statistical analysis features are available for traced neurons

Figure 4: A variety of statistical analysis features are available for traced neurons

Object features analysis

The segmented neurons are defined as individual objects, and ZEISS arivis Pro Automatic Neuron Tracer offers a wide breadth of object statistics to support in-depth analyses (Figure 4). The definitions of the key features are described in Figure 5.



A neurite can either be a subtree of a trace, that starts on the surface of the cell body segement, or a trace without cell body.


Starting points, branch points, and terminals are nodes.


A part of a trace between two consecutive nodes.

Starting Point

Each neurite has a starting point. If a trace has a cell body, the starting points of its neurites are positioned on the surface of the cell body segment.


A subtree of a trace that is starting at a branch point.

Branch Point

A node from which a branch starts.


End points of a neurite.

Figure 5: Definition of key analysis features.

In Figure 6, key information is summarized regarding reconstructed neurons. The object table provides easy access and export of the statistical object information.

In this dataset, 60 cell bodies, and thus neurons, are detected, with a total of 663 neurites. The term neurite is often used in a general sense to describe the branching extensions that emanate from a neuron's cell body, without specifying whether they are dendrites or axons. The neurites are highly arborized, splitting in large number of branch points and sections, ending in many terminal points. This type of information can be very useful when comparing neuronal populations between different conditions, e.g., in health and disease. Diseases such as Alzheimer’s and Huntington’s disease result in reduced complexity of the dendritic branching in specific brain regions.


Cell Bodies




Branch Points






Figure 6: Key features of traced neurons.

Radar chart illustrating the number of Neurites per neuron traced

Figure 7: Radar chart illustrating the number of neurites per neuron traced.

Branch Length

Different measurements can be used to quantitatively evaluate the morphological characteristics of the neuronal population. Here you can see a radar chart that highlights the variation in the number of neurites in the neurons traced. On average each neuron has 11 neurites, with most of them having under 20 neurites (Figure 7).

Box and Whisker plot of the average branch length per neuron traced

Figure 8: Box and Whisker plot of the average branch length per neuron traced.

The average branch length for this group of neurons is 10.8 μm (Figure 8). Genetic disorders such as Fragile X syndrome are known to result in shorter neurite branches in the regions of the brain affected.

Box and Whisker plot of tortuosity. The mean tortuosity of the neurites is 1.44.

Figure 9: Box and Whisker plot of tortuosity. The mean tortuosity of the neurites is 1.44.


Another neurite feature critical for neuronal function is tortuosity, which is a measure of the straightness of the neurite. Tortuosity is defined as the ratio of the displacement from root to terminal of the path to the path length. A perfectly straight path has a value of 1, and a tortuous path has a higher value. Neuronal tortuosity is altered in neurodegenerative diseases, which results in reduced connectivity and subsequent loss of function. In this neuron dataset there is variation in the path tortuosity, with the average value being 1.44 (Figure 9).

Path tortuosity as a function of path length

Figure 10: Path tortuosity as a function of path length.

Since tortuosity is heavily dependent on the length of the neurite, it is often expressed as a function of neurite length. This allows a standardized and quantitative approach for characterizing the degree of curvature normalized for the branch length. A high tortuosity to length ratio means that the neurite is more curved relative to its length, suggesting the neurite is more complex and convoluted. On the other hand, a low ratio would suggest a lower curvature relative to length, suggesting the neurite is simpler and more linear. In Figure 10, we can see the ratio of tortuosity to length plotted against length (independent variable). The graph illustrates that the shorter neurites have the highest tortuosity to length ratio, suggesting higher structural complexity. Conversely, as the length of the branch increases, the tortuosity to length ratio decreases, suggesting that the longer neurites are more linear.


This case study for neuron tracing demonstrates the use of the Automatic Neuron Tracer by ZEISS arivis Pro. A high-resolution, high signal-to-noise images was acquired with the ZEISS LSM 980 laser scanning confocal and processed with LSM Plus to create an optimized dataset for effective neuron tracing. Once the data were transferred to ZEISS arivis Pro, the analysis pipeline started with cell body segmentation, and then use of these cell bodies as starting points of the neurites to trace the neurons with Automatic Neuron Tracer. The threshold-based reconstructor used for the tracing was selected based on the dataset parameters of large size and high image quality. Once the segmentation was completed, the data were visualized and a vast number of morphological characteristics were extracted using object features analysis of Automatic Neuron Tracer. Specific parameters that reflect the complexity of neuronal arborization, such as the neurite number, branch length and tortuosity, were evaluated further.

The intention of this use case was not to perform a thorough analysis, but rather to provide a blueprint for neuron tracing.

The data set and analysis pipeline are provided here for individual test analyses of ZEISS arivis Pro and Automatic Neuron Tracer.

Try It For Yourself

Download a trial version of ZEISS arivis Pro and the case study files here

Share this article

  • 1

    Hang Xiao, and Hanchuan Peng "APP2: automatic tracing of 3D neuron morphology based on hierarchical pruning of a gray-weighted image distance-tree", Bioinformatics, 2019

  • 2

    Miroslav Radojevic, and Erik Meijering "Automated Neuron Reconstruction from 3D Fluorescence Microscopy Images Using Sequential Monte Carlo Estimation", Bioinformatics, 2019