sofroniewn
commented
4 years ago Hi @DottedGlass thanks for your interest in napari, that sounds like an exciting application! I used to work in neuroscience so I'm always partial to that stuff.
napari itself will probably not directly read an swc file at this point - we still need to think about how we will handle reading more specialized file types, but we definitely want to make it easy for you to visualize this sort of data, and maybe further down the line there can be a plugin for reading and viewing swc files.
Right now, to read an swc file into python you can use the neurom.io.swc.function from the neurom package which you can get via pip install neurom- see their docs here.
Once you can read the swc file you can then pass results to napari.view_shapes and as follows
from neurom.io import swc
import napari
data = swc.read('path_to_your.swc')
napari.view_shapes([data.data_block[:, :3]], shape_type='path', edge_color='red', ndisplay=3)and you get the following - note this is a random neuron I downloaded from neuromorpho:
Unfortunately note that it has two things wrong with it!
- [ ] One is that no radii information has been taken into account. This is actually should be pretty straight forward for napari, but we need to think about the api and then actually go in to the relevant parts of the code and provide the support. I'm not sure about a timeline for this, but it does make sense to me to do this.
- [ ] Two is that the shape has been assumed to be a
pathwhile what we actually have is more like a tree with segments. Because we assumed a path we just connected all the vertices next to each other, but that means that you only have two endpoints instead of a having endpoints at all the places where you need them. You could break the data apart and add many paths, but really I think it would be much better if napari added support for a new shape type called atree. We'd have to think about input data / api, but making that conform to the swc spec or something like it makes sense to me.
How does all this sound to you @DottedGlass?
vtwang
padster
adamltyson
jni
asheibanifard
eyildiz-ugoe
noisysky
🚀 Feature
Neuron traces are used to determine the pathway of axons/dendrites in 3D images of the brain. One common file format for traces is SWC, which stores the traces as a structually connected components (e.g. cylinders)
We would like to have a easy way to load data from a SWC into napari.
example of neuron traces
Motivation
It is of interest to neuroscience researchers to develop and evaluate automatic neuron tracing algorithms. Visualizing SWC neuron traces as a layer on top of the brain image can help researchers test these tracing algorithms.
I am on a project working with the neuron traces from the MouseLight dataset, and we are approaching the automatic neuron tracing problem as a semantic segmentation problem. As a first step we are converting the ground truth SWCs into binary image labels, so that we can use supervised learning methods to generate neuron segmentations/labels. Conversions are done semi-automatically. Labels generated from methods such as Ostu thresholding will need to be manually corrected.
Because napari is built with python, it is easy to visualize the effects of image processing algorithms or segmentation algorithms from common python packages. Furthermore, napari has great tools for manually making and editing labels.
Pitch
A simple way to implement this feature is to read a SWC and render it as a path shape layer. Adjustments will need to be made to path.py to handle an array of edge_widths, as each cylinder component in a SWC can have different widths.