slicermorph / slicermorph Goto Github PK

Need a mechanism to save the GPA module analysis, reload and continue from where it is left. While analysis is fast for demo sets, it takes a while for larger samples sizes, and customization (labels, etc) are lost.

Discussion: https://discourse.slicer.org/t/volume-rendering-produces-block-of-white-noise/10572
Code: https://gist.github.com/lassoan/4d0b94bda52d5b099432e424e03aa2b1

Add ability to create heat maps of shape variance on models across extremes of PC. Suggestion from SlicerMorph office hours 3/25.

@smrolfe
This will facilitate visualization of PC scores without needing a model. This would be a nice dynamic visualization supplement to the lollipop plots, which gives a static view of PC vectors.

"Export as..." function in Data module right-click menu is not convenient to use, because:

• last used file extension is not remembered: most of the time, user wants to use the same file extension; it is hard to always choose the correct one, especially when there are many (for example, when exporting a volume)
• last used output directory is not remembered: most of the time, user wants to use the same output directory as last time; it is quite inconvenient that you need to re-select the directory every time
• most of the volume output formats are not working (e.g, 2D formats, such as bmp, jpg, png do not work for 3D volumes), not relevant (Analyze, GIPL, MRC, ... are long dead), or duplicate (tif, tiff, TIF, TIFF, png, PNG, jpg, JPG, jpeg, JPEG)
• export options cannot be specified (compressed/uncompressed, etc.)
• need a way to specify if user wants the node to be exported in world coordinate system (parent transform hardened) or in the node's coordinate system (without applying the parent transform)
• nice to have: export all nodes in a subject hierarchy folder with default (or previously used) export settings into a folder

It would be better to display save dialog:

slicer.app.coreIOManager().openDialog("VolumeFile", slicer.qSlicerFileDialog.Write)

with some improvements, such as showing options similarly to how it is done for for reading data:

slicer.app.coreIOManager().openDialog("VolumeFile", slicer.qSlicerFileDialog.Read)

Last "ExportAs" file path could be saved in an attribute the subject hierarchy item, which would save both the folder and file type. But maybe it would be better to save the last used output folder for the node type or for all the node types.

After choosing one of the predetermined layouts (e.g., chart, or Red slice), we cannot revert our custom layout without clicking a particular button (e.g., plot gpa). This is not trivial to new users. Either have a button in GPA, or register the custom layout with the layout manager (preferred choice).

Stopped functioning corectly after #78

It unselect everything.

After doing GPA and loading the associated ply and fcsv file, mean shape points are not on the surface of the mean model.

This causes alignment issues if two versions (e.g., partial sequence and full sequence is loaded).

I am not sure what how to approach this. Somehow we have to know the geometry of the full volume, but I can't think of a way to this consistently. But a way to keep these lined up correctly would be important. Especially if and when we implement the ROI based partial volume import too...

ExportAs should allow users to save their volumes, markups and models from Data module using the right-click context menu. It appears this functionality works 100% of the time for Markups, if it is invoked within Markups module. However, functionality within Data module seems hit and miss, with no clear indication of why and what platforms are being affected. We have tested on 4-5 windows 10 machines, and 4 of them are not working as intended in Mac, we have 50-50 split. On linux two computers where tested and it works 100%. (Update not correct anymore. At least one Linux system seems affected).

I already made the changes about removing the control stroke for points, toggle persistence and lock (they are now just p, t, and l).

But for some reason cycleEffect doesn't work I am getting this error:
NameError: name 'cycleEffect' is not defined
Traceback (most recent call last):
File "", line 91, in
NameError: name 'cycleEffect' is not defined

When a 3D model and corresponding set of landmarks are selected to visualize 3D deformations, and the toggle mean shape landmarks is enabled, the mean landmarks shown in viewer #2 (deformed model) should the deformed mean shape one. Otherwise they just interfere with the visualization and make it harder.

Let's have a button to export the warped 3D model. This would be the reference sample warped to mean shape if the sliders are at (0,0), and then subsequent applied PC values. This button will sit just above the 'create animation' section.

Please make the default export option to the disk (not into the scene), and in PLY format.

PLaceholder for brainstorming a simpler UI for microCT datasets:

There multiple recurring issues that can be easily addressed.

1. We need a reasonable transfer function automatically calculated. This might be a simple ramp that covers the 90% of the data range.
2. Current initial guess does an OK job, when the volume is rendered very first time in the scene. Say if you want to generate a new volume property, it starts from this solid rectangle that it you can’t make it anything into (at least not with a lot of work). Even worse, the only way to recover is to reset the scene.
3. And of course data size…

I kind of like the idea of having a SlicerMorph specific volume rendering module that is integrated to the Data Module (right click on it and show it in 3D. Right now you have to have the right click on the eye icon). Or modify the data module to have specific 3D rendering colume (in addition to 2D visibility).

Example of useless volume property generated when the user manually chooses "Create New VolumeProperty".
The initial:

Subsequent ones:

Currently registered layouts do not contain any maximized plot or table views. Consider adding single table and plot view (for maximing viewing area) and also register our custom slicermorph layour.

Testing with this data:
http://www.bic.mni.mcgill.ca/uploads/ServicesAtlases/macaque_25_model-MNI.nii

Error is:

Switch to module: "ImportFromURL"
Could not download data. Not a supported file type.
Could not import data. Please confirm that the URL and file name is valid.

We need to make sure the reference landmark's visibility is automatically turned off after it is selected in GPA module. Otherwise the node gets hidden and it becomes impossible to hide it manually.

When people are using ImageStacks to convert multiple stacks to volume back to back, it is easy to overwrite previously imported volume when they are not careful.

1. Import the sample_CT_stack from SlicerMorph Sample data
2. Keep the volume node none, which create a new one from the file name
   3 Import the image stack
3. Clear the files from previous stack and get ready to import the next stack
4. Note that Volume Node points to the volume generated in the previous import operation, and easy to overwrite.

Solution:

1. Not give user the flexibility to create a volume, and always import data into a new volume node using filenames.
2. Set the volume node to back to none, right after the import is completed.

My preference would be the second.

To give user more control over the resampling of the intermediate curves drawn between patches.

It doesn't do much. We never use it to download anything directly from MorphoSource. MorphoSourceImport covers all useful use cases.

From Murat:

I really would like to see an automated conversion of vector2scalar volume during the data import phase for bmp/jpg/png stacks (tiff doesn’t suffer from this). The way I envision is an extra checkbox in options for people who installed SlicerMorph, if we cannot do it default for everyone. If we can get pass that point, things will be a lot more easy for the folks. What’s happening is they are using 3 times as much memory for volumes that are already fairly big to begin with (almost everyone worked with datasets >1GB, sometimes 5-6GBs), and almost all volumes coming from bruker/skyscan microcts are bmps (which is the default output option in their post-processing software). And there are a lot of data coming from those scanners.

With otolith data I see 40 cores being utilized for almost a minutes. Seem excessively high for just plotting...

Sample data path.
/home/kelly/otolith/auto3dgm_test/output_l_s/phase2/aligned_landmarks

From an user email during the workshop:

Hi Dr. Maga,

I used a module called ReadLandmarkFile and changed the IDAV Landmark export file to .pts. It was successfully imported! Thank you! Is this the Isaac Landmark reader? And I encountered another problem. The imported landmark cannot correspond to the .ply file after I centered them. It seems they are mirror image of each other. Do you know how to correct this?

I think we just need to LPS2RAS to read them correctly.

Restrict the parsing to the specified folder.

As requested from Shan Shan

Change from import tool to conversion tool (save file instead of loading into scene).

Currently the 3D point cloud option under the Landmark variances tab plots in original space, convert this to GPA aligned coordinates.

Currently imagestacks uses the order OS provides. THere are reports of cases where MacOS does not seem to sort the files by filenames by defaults, or skips files during selection of large number of files.

A filename parser that will sort filenames will be good to mitigate these issues.

To replicate,
Enable 3D rendering for MRHead.nrrd, adjust volume property.
Mouse mouse to 3D rendering window, hit p, see that you cannot create a point without moving the mouse around.
Replicates all the time on latest windows stable.

While this is not a terrible bug, I think we should change the behavior of P keystroke, not to go into Place mode, but actually place the point where the cursor is. It is more convenient (and intuitive) to adjust your rendering first, position your cursor, and create the point at that location.

Please revise to work with any data type Slicer accepts (3D models, fiducials, transforms, etc) not just models, including zip files. If it is a zip, action should be just unzipping to the temp folder, do not load into the scene.

There is now a preference in the SlicerMorph preferences to set the temp location for both cache and Temp folder the same location.

How about renaming the section of GPA module from "Setup 3D Visualization" to "Setup Dynamic Visualization", or "Setup Interactive Visualization" Lollipop plots are also 3D.

This can occur after making several changes to viewing options, PC warping, and mean shape display.

FOr large data
lms.tar.gz

I cannot seem to control the scale of the landmark label with the Markups module display tab in viewer #2 (the one that deforms). I can change it in slice views and viewer #1 but not in viewer #2

Set the curve setting to linear instead of the default Spline. WIth spline it is not possible to control the shape of the curve.

This is a due to an incompatibility with glibc libraries.

It renders ALPACA unusable in most Linux systems including Centos8 or Ubuntu 20.04. We will either have to update the ALPACA to use open3d >=0.10 on all platforms, or disable ALPACA for Linux.

I propose to open a separate development branch for migration to open3d 0.10.0

If the Dcm2niix plugin is active, then Slicer crashes when trying to load apparently any volume. See this video here:
https://1drv.ms/u/s!Ao_a-dPPX98ZhLskvBgBvirQjDDI3A?e=fOwOZP

For the record, the dataset used here is the TinyPatient dataset from the SlicerRT sample data store (please note that I only selected the CT):
https://github.com/SlicerRt/SlicerRtData/tree/master/plastimatch_tiny-rt-study

this is with 4.11.0-2020-07-07 r29207 / dad86b0

• also Export As still appears only in the Markups.

Loading Slicer RC file [C:/Users/murat/.slicerrc.py]
Volume nodes will be stored uncompressed by default
Segmentation nodes will be stored uncompressed
parent qSlicerScriptedFileWriter (QObject 0x00000212AAA2A870)
Write {'fileName': 'C:/Users/murat/Desktop/C.fcsv', 'nodeID': 'vtkMRMLMarkupsCurveNode1'}
Traceback (most recent call last):
  File "C:/Users/murat/AppData/Roaming/NA-MIC/Extensions-29207/SlicerMorph/lib/Slicer-4.11/qt-scripted-modules/MarkupsFcsv.py", line 58, in write
    self.parent.writtenNodes = [markupNode.GetID()]
AttributeError: 'writtenNodes' does not exist on qSlicerScriptedFileWriter and creating new attributes on C++ objects is not allowed
Could not save C to C:/Users/murat/Desktop/C.fcsv

Add selection between visualizing with points only vs a reference model

From Andras's response on discourse:
Custom settings panel
https://discourse.slicer.org/t/application-startup-script-open-option-is-confusing/9596/2

Create mock tabbed layout for GPA module for each step of the analysis (as oppose to scrolling down)

Need to optimize PCA speed. Same dataset takes 4-5 sec to align and PCA in R (either geomorph or Morpho).

Couple usability suggestion for the next version of alpaca

1. Color the target mesh lighter to make the points more visible (I think this is done, I report it for the posterity).
2. Implement a new tab to calculate a Slicer affine transform (so that ALPACA can be used as general purpose alignment tool).
3. For both single and batch processing, give an option to write the setting and re-read them. This will facilitate consistent multiple runs easier.

If it is, execute the analysis, if not return an error message about not being able to save results (gpa doesn't work properly if can't save).

`Loaded 44 subjects with 500 landmark points.
9
Scale Factor: 0.14121616650561752

Traceback (most recent call last):
File "/home/maga/.config/NA-MIC/Extensions-28927/SlicerMorph/lib/Slicer-4.11/qt-scripted-modules/GPA.py", line 473, in onLoad
os.makedirs(self.outputFolder)
File "/home/maga/Downloads/Slicer-4.11.0-2020-04-03-linux-amd64/bin/../lib/Python/lib/python3.6/os.py", line 220, in makedirs
mkdir(name, mode)
PermissionError: [Errno 13] Permission denied: '/home/kelly/otolith/41clupeiformes_2/auto3dgm_test/l_s_output/phase2/aligned_landmarks/2020-04-24_13_33_45'`

When loading a reference model for the GPA module, user has to manually switch between vtk, ply, obj and other formats, and sometimes thinks that files are missing. Change the dialog box to include all acceptable model formats at the same time

Remove the SlicerMorph Labs section, remove the community contributed modules (it is not functional anyways).

I/O modules:
ImageStacks:
SkyscanReconImport:
MorphoSourceImport:

Geometric Morphometrics modules:
Generalized Procrustes Analysis (GPA):
CreateSemiLMPatches:
PseudoLMs:
ALPACA
Auto3Dgm:
LMTemplate:

Convenience Modules (or a better alternative)
Animator:
ImportSurfaceToSegment:
MorphologikaLMConverter:
PlaceSemiLMPatches:
ProjectSemiLMPatches:
VolumeToModel:
SegmentEndoCranium:

Currently any downloaded images through the tool is written to the disk, which is good. But you can only discover this if you try to save the elements. Make this is a customizable option that user can specify (e.g., on windows it saves to Documents folder, which I never use).