fiji / snt Goto Github PK

@ctrueden, I was trying to release a new version (as I've done in the past: 1) commit directly to master the version in the pom 2) uploading the jar and 3) increment the version back to SNAPSHOT in pom.

I just did 1) with 2fca2f4 but the build failed. It seems that things changed since I last did this. I also noticed that the github releases page seems to be automatically populated (before I was creating releases there by hand).. What is the new procedure?

For consistency, all views should be spatially calibrated, right now this is not the case. See also improvements proposed in #13.

See discussion in the forum.

From @mhl:

At the moment the points you manually specify are considered fixed, so won't be adjusted by the "Fit Volume" option, for example. This is a limitation that I intend to fix, but it needs a bit of careful work. There are two plans I have for this:

• Allow manual adjustment of where the points are at any stage
• Allowing the fitting to adjust the manually placed points as well as the intermediate points

In the mean time, my suggestion would be to select the intermediate points in a zoomed-in view, using the three-pane view to find the midline of the neuron. There's a short tutorial here explaining what I mean:

http://fiji.sc/Simple_Neurite_Tracer:_Accurate_point_placement

Migrated-From: http://fiji.sc/bugzilla/show_bug.cgi?id=222

Reported in the forum. It is likely related to #18

This is to log the ImageJ2 migration efforts. The work is starting on my scijava branch.

Braking changes:

• Applet support will be dropped (Is anyone using it?)
• IJ1 macro support will be dropped. Was flimsily and not that useful
• All classes no longer in use from the virtual fly brain project will be deprecated
• Support for old 3D viewer will be dropped in favor of SciView (see #22, scenerygraphics/sciview#42)

Other changes:

• Reduce IJ1 dependencies (see also #29)
• Reduce VIB-lib dependencies (no longer maintained?)
• Allow ops filtering (Tubeness and Frangi ops are about to be released, see imagej/imagej-ops#490)
• Allow tracing on any pre-filtered image
• Multichannel support (see #13)
• Proper scripting support
• API documentation (currently there is zero javadoc documentation)

• Re-enable FlyCircuit (done in 65b426e)
• Add option for original/standardize file for NeuroMorpho (done in ec2ebe8)

The UI is not updated when using Keyboard Accelerators. E.g.:

• When pressing M, the tracing canvas(es) should navigate to the clicked position
• When pressing 5, the display choice should be updated in the NeuriteTracerResultsDialog.

This is just a loose list of things that could be improved:

• All commands that offer the option to open Rec. Viewer (e.g., Path Manager's Analyze>Color Coding...) should also ask for to open SciView. Actually, we could just remove that option, as it does not seem to provide much
• One would have to extend whatever we decide above into all the other commands that have that option
• Pressing "Open SciView" should also sync Path Manager changes
• Adding the sync call to SNT#updateAllViewers() would cover 90% of an user's sync needs

@kephale, I will try to look into ssome of these late next week.

@kephale, Subscribers to the NeuroAnatomy & SciView update sites get this:

Exception in thread "Thread-12" java.lang.NoSuchMethodError: sc.iview.SciView.addWindowListener(Ljava/awt/event/WindowListener;)V
	at sc.fiji.snt.SciViewSNT.setSciView(SciViewSNT.java:135)
	at sc.fiji.snt.SciViewSNT.initSciView(SciViewSNT.java:110)
	at sc.fiji.snt.SciViewSNT.<init>(SciViewSNT.java:103)
	at sc.fiji.snt.SNTUI.lambda$32(SNTUI.java:1598)
	at java.lang.Thread.run(Thread.java:748)

This is because the SciView version shipped is older than SciView 7b5befb. Shall SciView be updated on its updateSite, or shall I start uploading the updated jar to NeuroAnatomy?

It would be nice to be able to specify the filename for saveSnapshot() in Viewer3D

Tubular Geodesics (TG) is an incredible extension and the best thing that happened to SNT since its inception. Installation of TG is the single most thing users can do to really improve tracing quality. It seems critical to distribute Tubular Geodesics through an update site, either through the base Java 8 update site, or through the Neuroanatomy update site.

The former is much more demanding since we'll have to abide by the Fiji's contribution Requirements. The latter is much more flexible as there are no really no requirements - just ensuring that we will not break the user installation - as described in Distribution guidelines.

@fethallah, what do you think? Would that be OK for you?

My take would be to use to use the Neuroanatomy site to distribute the binaries and update the TG tutorial accordingly. I am guessing that the instructions on the Tubularity repository are awfully outdated and this will require a lot of legwork. I'm also guessing we'll have to set up a Jenkins job for this, as e.g., I do not have immediate access to a Windows machine.

The current code automatically inserts Path identifiers into the filename. When the tree only has 1 path this is not always the desired behavior.

From a message to the ImageJ mailing list:

Currently in our department we are trying to find a standardized method of tracing neurons received from stacks. We generate the stacks and any substacks within Fiji and once we have a target neuron we import the stacks into Neuronstudio to trace it, as Neuronstudio can automatically trace. After saving these traces in an swc file we try to reload it into Fiji Simple Neurite tracer, however we always encounter an issue with scaling. We always have the correct offset coordinates at hand, previously saved from the original stacks in Fiji, but the offset scaling is always off, sometimes really off. The tracings always appear smaller than the original. With trial and error we can come as close as possible to the correct scaling but this is rather time consuming and the endgoal is to fully automate the whole process via standard procedures for a protocol, that will be used in the lab. Would anybody know if what the precise error is or how to work around it?

Hi, thanks for the great software. I would like to generate a "rasterized" version of a neuronal trace, which is currently in SWC format. It seems that previously this functionality existed in SNT, as described in the Methods section of this paper. However, the procedure described here no longer seems to exist in the SNT plugin menus.

Is there a way to convert from SWC file to a 3D output, where each voxel in the output is zero if it is not an "active" voxel in the network described by the SWC, and 1 otherwise?

I have also posted this question here if it is easier to answer there!

For example, we can make Tree work with SciJava parameters

Some relevant snippets:
https://github.com/scenerygraphics/sciview/blob/janelia-minihackathon/src/main/java/sc/iview/ActiveSciViewPreprocessor.java
https://github.com/scenerygraphics/sciview/blob/a556792cb92499424b785fa3e82448646528181e/src/main/java/sc/iview/DefaultSciViewService.java#L142

This is to log the recent improvements in path management. The user interface chosen for such tasks is the "All Paths" frame. The work is occurring on my scijava branch.

Already implemented features (partially tested):

• Searchable list of paths
• Tag paths with customizable colors
• Custom SWC-type colors
• Export of selected paths as ROIs
• Hierarchy edits through drag and drop dropped

Features needing fixes:

• Number of search hits in search bar is not accurate caused by reliance on outdated jide-oss
• Auto-expand of selected nodes not working?
• SWC-type colors are not applied once they have been customized

The new UI does not rely on IJ1, relying mainly on org.scijava.ui.swing. Currently it looks like this:

Any feedback more than welcome!

Right now the ability to summarize reconstructions statistically is rather poor and limited to the the Export as csv command or the Analyze Skeleton output that runs on digitized paths. There is really no straight forward way to obtain at a glance things such as N. of branch points, Total length, etc... Otherwise, we should at least point users to web-tools such as L-measure or Remod or nat

This relates to #46. While we prepare SNT's release we will have to update the existing documentation and screencasts that are awfully outdated.

The simplest would be to mirror existing wiki pages to a new name e.g., Duplicate the contents of https://imagej.net/Simple_Neurite_Tracer:_Key_Shortcuts to https://imagej.net/SNT:_Key_Shortcuts, then once the release is out, simply redirect the former to the latter. This away no pages get broken in the process. A non-exaustive list of things to be done is:

Screencasts:

• Introduction to tracing in SNT
• Introduction to tracing on filtered images
• Advanced operations: Forking, editing nodes, etc
• Analysis (Sholl, Strahler, measurements, etc.)
• Scripting

Wiki:

• Create a scripting page, including a dedicated one on Python Notebooks
• Create a dedicated page for Reconstruction Viewer
• Move all relevant pages https://imagej.net/Simple_Neurite_Tracer:foo -> https://imagej.net/SNT:foo

Plugin:

• Update Help> Menu to reflect new changes
• Add descriptive tooltips to GUI elements whener relevant (for offline documentation)
  Pinging @carshadi, @tferr

See https://forum.image.sc/t/trouble-saving-paths/22965

See http://forum.imagej.net/t/unable-to-define-radius-as-a-whole-number-in-snt-sholl/12081?u=tferr

Hello Everyone,

I'm new here so let me introduce myself first.
My name is Jesse and I'm a research technician at UMC Utrecht in the Netherlands.
My goal is to learn some basic coding (Python and ImageJ macro) in relation to automated image analyses; I'm very new at this.
My current challenge is an automated object/cell count on this image (and many more).
.

First I want to create a clean F-actin skeleton and use it for counting objects. I was experimenting with gaussian blur, tubeness and thresholding etc. with some quite good results. I discovered the simple neurite tracer and I'm convinced this will give me the best F-actin skeleton if I'm able to create a mask and have to click once... So this is my question: Is it possible for me to write a macro where it traces all F-actin lines and creates a mask without clicking? I know this is not neurite related but this will help me a lot with my macro. It would be much appreciated if someone can help me with this or point me in the right direction!

Thanks and best regards,

Jesse

I'm trying to create a Tree from AA0792.swc, but the soma gets relabeled as an axon after loading.
:root #object[sc.fiji.snt.util.PointInImage 0x21440e36 ( 6169.424625, 1550.220808, 4726.28674 ) [onPath Path 0 [axon]]]

Currently 3D_Viewer is unusable. We need to migrate away from it and adopt ThreeDViewer (see scenerygraphics/sciview#42)

Trees created with GraphUtils.createTree() do not have the proper Strahler order and lead to errors in analysis/statistics

Reported and discussed in the forum. Basically the educated guess performed to automatically populate the plugin's dialog is not working that well in all cases. Also we should ensure we do not sort null arrays here and here

Hi,

I am looking for a way to easily encode the "Three-point soma" within ImageJ/SNT in both .swc and .trace formats, but I haven't found any of it after hours of googling.

Or do I have to edit the .swc file myself after exporting it from ImageJ?

🎉 !!SNT is finally ready to be released!! 🎉
Because the program changed significantly, we would rename it to SNT. The change may sound silly but it highlights the effort that was put into backward compatibility and makes the point that the program is no longer that Simple 😉

Also, I see a couple of issues that we should look into before the release:

• Dependencies: json does not follow SemVer
• Dependencies: jzy3d-api required by us is a SNAPSHOT version
• Impact on other projects: Last time I checked, TrakEM2 was using SNT, so we'll have to make sure things still work there. I will be looking into it, but Pinging @axtimwalde for now, for a heads-up. This is addressed with trakem2/TrakEM2#25

@ctrueden, how can we achieve this and release SNT v0.0.1? (or is it v.0.9.0-SNAPHOT?). I'm assuming it is just not enough to rename the project. Can you help us assemble a todo list fo this?

Dataset consists of multichannel multiframe tiff files.There are n tiff stacks. Each stack has Nimgs where each img is from a depth along the axial plane of the brain, n is the number of channels. Once motion corrected, we are looking to trace the dendrites belonging to a stack in 3d.
Does SNT have any inbuilt motion correction within images and between images in a stack?

Dear SNT developers,
I posted this question on the forum.image.sc but I didn't receive any answer yet, I'm thus posting it here, I hope it can be the right place. I'm still not sure how and where to posts questions about imageJ plugins and I apologize if this is not the place for my kind of question.

I am importing in SNT v3.1.101 some Swc files that were traced in Neutube to perform some measurements.
Yet I’ve two problems,

1. Minor issue: We tagged processes with different type determinants but this information is apparently lost in SNT (both in path manager and exported CSV file).

2. Major issue: Our cells have a number of small lateral processes (here a simple example with only two of them) that stem from a primary dendrites and we would like measure their numbers, lenght etc. We would also like to measure the inter-process distance.
   Yet impored swc are subdivided into paths that always continue past the branch point along one bifurcation, without caring about post-branch path lenght ( I guess they simply follow the order of nodes in the swc file). Rather we would like to split each path between branching points to divide them in segments (branch to branch and branch to termini),

A manual method would be useful (I could not yet figure it out!)
However, as we have many cells and many processes a hint on how to do that programmatically would be much appreciated!
If SNT is not the right software to perform such analyses, can anywone suggest and alternative?

Thank you very much

Federico

Advanced features remain rather cryptic. Most users are not even aware that e.g., a tubeness image can be used. This could be done through the GUI or at least in the documentation page

A good starting point could be https://gist.github.com/tferr/70b358d62d99f1189f8b0cf387a176c0

As discussed in the forum. A simple way to implement it would be to adopt Trackmate's strategy.

from code:

...
            SWCPoint soma = (SWCPoint) loader.getNodes("soma").toArray()[0];
            AllenCompartment thisSomaCompartment = (AllenCompartment) soma.getAnnotation();

            if (thisSomaCompartment.contains(somaCompartment)) {
...

I'm getting (after a bunch are successfully downloaded):

...
Parsing AA0952...
Exception in thread "main" java.lang.NullPointerException
	at sc.iview.neurongrammars.FetchNeurons.main(FetchNeurons.java:77)

It would be nice to be able to specify the filename for saveSnapshot() in Viewer3D

• Syncing multiple times duplicates nodes in SciView (see scenerygraphics/sciview#215)
• After SciView is closed, pressing Sync Changes (or performing an operation triggering a sync call) will continue to sync changes to the closed SciView Viewer.. Actually buttons in the SciView panel should update after SciView is closed.
• Volumes are not aligned (Note: this situation occurs when SNT populates a trace in SciView then "Add Volume" is used to open the image. Suggested alternative is to open a trace with a target image, which can be done via https://github.com/fiji/Simple_Neurite_Tracer/blob/137b36b0be2c9a080f87ffc5e8ce859ff5357e36/src/main/java/sc/fiji/snt/SNT.java#L1939)
• Consider disabling the inspector and disable SciJava events

It is actually fairly easy to have outdated tags: The solution would perhaps be to refresh all tags once commands are run

Problematic when tracing images with e.g., multi-point ROIs specifying landmarks

We should do much more to support interoperability with other ImageJ plugins and other programs outside ImageJ. E.g.:

• Support Vaa3D's ESWC
• Provide a straightforward converter for the .traces format (implemented through the Neuroanatomy update site)
• Allow paths to be exported to the ROI Manager or the image overlay [Implemented in v3.06]

Useful features but dropped requests:

• Start reading comments in SWC headers. Without true standardization among SWC exporting programs, we'll never be able to extract valid metadata from such headers. Unfortunately, this has to be dropped
• Allow import of ndf files. Better would be to make NeuronJ aware of SWC (see NeuronJ#1)

This can be reproduced by e.g., importing the OP_1.swc dataset: Path 0, does not get tagged as primary.

Inconsistencies/Improvements in the UI, as reported by @carshadi:

• Could be useful if the just finished path becomes the selected one e1ad0c3
• A way to re-initialize a "finished" path for additional A* tracing would be helpful 091c566
• There should be a way to select the children a selected path 5a841e5
• "Activate canvas on mouse hover" should be active by default
• In order to make a fork point, you need to hold down ctrl+shift and then move the cursor slightly before "Fork Point" appears b91cecb 49c3fb1
• Deleting a path from the path manager, while still in edit path mode, the delete does not occur until switching back to tracing mode (it occurs in the path manager but the canvas does not update until switching) ead9dc7

When importing a SWC file without an image SNT uses an empty placeholder image. With certain files the dimensions are completely off which makes it fairly easy to run out of memory even with small files. E.g., as in the following example provided by @kephale

Currently if one has to import multiple SWC files, one has to do it one by one. It would be nice to be able to do it in one go

@tferr Since this repository has been superseded, shall we set it to archived mode, to make it read-only?

As is, the "include all paths" option in the Sholl Analysis dialog assumes that the soma has been traced as a single point (zero-length path). This is rather restrictive and not recommended. It would be desirable to allow users to filter out somas defined by multiple paths. While at it, we could allow Sholl to be performed on specified paths filtered through SWC types ("soma", "dendrite", "axon", "end point", etc.), which would allow streamlined profiles from specific compartments.

The issue is discussed in this forum thread.

Restricting tracing to single channel images is problematic: there are several scenarios in which it is important to trace paths in one channel while accessing a counter-staining in another, e.g., a lesion site, other cells (as in brainbow-like imagery), tissue landmarks, etc..

This may be quite involved and may require a major rewrite of InteractiveTracerCanvas and BatchOpener.