paluchlabucl/deformingmesh3d-plugin

An imagej/fiji plugin for interactive segmentation of 3D surfaces

deformingmesh3d-plugin's Issues

Interactive track adjustment.

Using the T and B keys can be used to track the mesh between frames. By holding down "shift" the mesh in the next frame can be clicked to be selected. "ctrl" will copy the mesh from the current frame to the next frame. "L" will attempt to link automatically to a candidate mesh.

BUG: Selected mesh stays selected when tracks are cleared.

This BUG affects users when they, load new meshes, use the mesh manager, or start new meshes.

The selected track will be removed from the display, but the track will still be selected. Tracking the mesh will cause it to be added, in it's entirety, to the segmentation.

Volume displayed in the volume viewer doesn't update scale when a new image is opened.

When a new image is open the image can look distorted if it has a different scale than the previous image.

Work-around. Show/hide the volume.

Record snapshots is broken when the volume is showing.

The output image is only the first frame.

Make ctrl-s save the current meshes.

Add an accelerator to save meshes using ctrl-s. Concerns, selecting a file if one hasn't been selected before. Adding some indication that the file is actually being saved.

Control the furrow plan from the 3D canvas.

The commands to move the furrow, center, and split a mesh all exist and just need to be added to the appropriate controls with hot keys.

f - centers the furrow
f+shift - starts the choose rotation
f+ctrl flips the furrow direction
up moves the furrow forward
down moves the furrow backwards

Mesh scale factor

Hi,

Thank you for this awesome plugin to create 3D cell annotations. I want to use existing nuclei segmentations as mesh initialization. To achieve this I load the label images in Python and convert them to meshes with skimage.measure.marching_cubes. Then I shift the obtained meshes such that the origin corresponds to the image center. However, I have to apply an additional scale factor to the obtained vertices to get the meshes to fit.

In code this is what I am doing for each label-mask:

verts, faces, _, _ = marching_cubes(mask, spacing=spacing)
        
# Shift to origin.
verts[:, 0] -= ( labeling.shape[0] // 2 ) * spacing[0]
verts[:, 1] -= ( labeling.shape[1] // 2 ) * spacing[1]
verts[:, 2] -= ( labeling.shape[2] // 2 ) * spacing[2]
# Scale.
verts *= factor

So far I couldn't figure out how this factor is computed. By trial and error I figured out that the following combinations work:

• spacing = (0.6, 0.216, 0.216), factor=0.0225
• spacing = (1.0, 0.25, 0.25), factor=0.0185
• spacing = (0.5, 0.25, 0.25), factor=0.0195
• spacing = (1, 1, 1), factor=0.004875

I tested this by loading the mask image (which has the spacing set) and then loading the mesh. Then I checked by eye if the outlines fit. I guess the re-scale factors are good enough that I can't see it anymore by eye. However, I would not bet that the values are 100% correct.

Could you enlighten me what the correct formula would be to obtain this scale factor?

Thank you and have a nice day!

Colors are linked to names.

Colors are linked to the name of a mesh track. That is ok for autogenerated mesh names, but if a mesh name has been set, it should retain that name. It should also be represented in such places as the status pane, and the mesh track manager.

sphere initialization handles are not scaled properly.

The handles for controlling the size of an initialization sphere are too large for large images.