Guide to Making 3D Models

Large volume 3DEM datasets provide endless opportunities to illustrate new aspects of  cell and organelle organization.  Simple views of the stack with and without cutaway elements provide literal insights into cell operation.  3D renderings of the organelles based on tracings can be turned into videos using software such as Blender, and can be both scientific and educational.  This article provides a start to using your own datasets to illustrate your findings.

One simple workflow consists of:

Workflow_3DEMv02

  1. Make Sketches and choreography.

  2. Find suitable area and then

  3. Crop out a working volume at original size.

  4. Make CutAway Models in which geometric parts are blacked out.

  5. Make Painted OutModels in which the things you want gone are painted black

  6. Make Segmented Models in which clever algorithms do the painting

  7. Resize all of the above to make scaled working stacks that are pseudo-isotrophic.
  8. Combined Models containing both cutaway and segmented volumes, colors, etc.

  9. Import into ImageJ 3DViewer or Amira and make sophisticated views and videos. Can also export as surfaces to Blender to make shaded 3D images and movies.

Below are some of the images that we have made using free software based on stacked data.

 

 

1. How do I sketch or choreograph what I want?

sketchForUse.jpg

With a pen, pencil, colored pencils or crayons.

Ask yourself “What is it that I am trying to illustrate using a 3D model, that can’t be shown using a simple slice image?”

If making videos – ask “What is it that I am trying to illustrate with movement?   Would a simpler, still image suffice?”

For example, in single slices, the endoplasmic/axoplasmic reticulum looks like randomly placed short threads.  In contrast, 3D reconstructions show its interlinked, net-like arrangement inside each axon.   By “flying” the camera along the axon centerline, you can see that the AR is not just around the cell periphery but includes central axoplasm as well.

So start out with a pencil and paper and sketch/storyboard the illustration or demonstration video that you hope to produce.  Then sit back and criticize it as if you were someone who is simply looking to find fault – be “Reviewer Number 2”.

 

2. What do I look for in an ideal example to use a working volume?

If you have scaled versions of the original stacks, pull them up and look through them for cells (or whatever) that show the features you care about.  Save drafted versions and make notes about where each is.

Pick or find 3-10 examples that suit – write down stack name, slice and region that has your cell and duplicate and save the best slice in a scrapbook (eg PPT doc).  Step backwards and forwards through each stack and consider the following:

Imagine each possible model in other orientations:  rotated, in reverse z- order, with more or less surrounding cell material included.  Which would look satisfying:  ie balanced artistically, visually simple, unambiguous, and with components cleanly demonstrated.

Are there technical defects?  Is there skipping in the stack, flecks, fixation artifacts, incomplete cell, examples of contradictory or confusing profiles in surrounds.

How big or small a region will satisfactorily capture the structures of interest?  Higher magnification means less work for you and less difficulty to comprehend, but lower mag provides better perspective.  Often both are needed.

The goal is to keep the job manageable.  Do as little tracing as possible.  But do not crop so tightly that you end up with a finished model that is not sufficiently big to work with.  Once you pick an area, it is difficult to go back and change it without repeating a lot of tracing.

3. How do I make a working stack?

After selecting a stack and cell that you want to work with:

a) Open the full size stack in ImageJ/Fiji.  If RAM is limiting, open as a virtual stack.

b) Locale the start and end of the cell in the Z-direction.

c) Box up the cell with the rectangle (marquis) tool.

d) Duplicate that part of the stack:

Image>Duplicate> stack, slices x-y, name=OriginalStack.

e) Register the duplicated stack.

Image>Adjust>Canvas>(set values 10%-25% larger)
Plugins>Registration>Linear Stack Registration With SIFT> octave=5, model=Translation,

f) Crop the stack so that the edges are straight.

Image>Crop

Save this stack out as slices (It may be large), into a folder called

File>SaveAs>

myExample_WorkStack_Full

 

 

3. How do I Make CutAway Models in which geometric parts are blacked out.

 

How do I make a resized working stack?

Decide on the resolution you need. Usually 20nm voxels is good for most jobs. Remember that you will add additional slices in z, and convert RGB color as well (a 3-fold increase in file size) if you plan to do much work. So smallest size that still shows the features is best. If it seems to small, consider having two projects – one to show overview and another to show a small area in detail. The compromises are worth sorting out at the start because it is hard to go backwards, unless you love doing arithematic.

First ensure that the properties are set correctly for your files – they often are not.

Image>Properties>Unit: micrometers or um, PixelWidth: 0.005, PixelHeight: 0.005, VoxelDepth: 0.075

To go from 5nm (x,y) and 75nm (z) to 20nm xyz, you need to downsample in x,y and upsample in z (interpolate).

A simple, one-step strategy would to scale x,y by 0.25 and z by 3.75

However, as the z step needs to “create” information, if it does it all in one step, the newly created slices will suffer from fuzzy interpolation. Better to do it as scaling by 2, and then another scaling as 1.875. The size is the same but the result is better.

Image>Scale>X Scale: 0.5, Y Scale: 0.5, Z Scale: 2, …

then

Image>Scale>X Scale: 0.5, Y Scale: 0.5, Z Scale: 1.875, …

Save this stack as something like myExample_isoStack_00.tif using ImageJ stack format, or as series of single slices. I like single file format for compactness, but Amira will need the slices.

You should also brighten and contrast the resulting stack eg slice-by-slice automatically

Process>EnhanceContrast>Saturation:0.3%,Normalize, ProcessAllSlices

How do I view my isoStack as a 3D “Cube”?

With the iso stack open on screen use the (FIJI) plugin 3D Viewer:

Plugins>3D Viewer

OK – at the time of writing, there is a big issue with the 3D Viewer caused by Java or something (Read about it here) and while it works on some of my machines, on others not so much. It has to do with graphics card compatability, and it is a pain in the butt. Others have asked about other alternatives that are free – and there are a couple (see another article). You can also use Amira or other commercial software. Or Blender if you are game.

You can open multiple views of the same stack in seperate viewers and syncronize them so that they all are at identical angles.

older

  1. Sketch and choreograph what it is that you want to see at the end. Figure out what you will need and make a shopping list of images or image sequences (video)

Model Making. Uses ImageJ/FIJI

  1. Make a working stack from which all your models will be derived.

  2. Resize it: the resized working stack is scaled to be isotrophic, manipulable, and big enough.

  1. Duplicate the resized working stack and make

  2. Cut_Away_Models in which geometric parts are blacked out that will be invisible later

  3. Painted_Out_Models in which the things you want gone are painted black in each slice, or the thing you want to keep is painted in colored.

  4. Segmented_Models in which clever algorithms do the painting

  1. Composite_Models in which the views above are used against one another or added to each other. Eg the mitos in only one of the cells, plus its nucleus.

  1. Then those models can be imported into ImageJ 3DViewer or Amira and sophisticated views and videos created, or surfaces exported to Blender or other software to make shaded 3D images and movies and printable 3D objects.