It is helpful to understand the model data structure before trying to create a model. A model is basically a collection of points that refer to or mark interesting locations in an image. These points are stored in groups that are called contours. In the simplest case, a contour is a closed boundary around a region in an image. However, you should think of a contour more generally as a set of associated points, which may or may not be connected by lines. You will typically want to keep track of more than one type of feature in an image. This is facilitated by allowing contours to be grouped together into separate objects. Each object can be given its own name and all contours in the object share common attributes such as color and line thickness. This table shows how objects, contours, and points are organized in a hierarchy in the model.
| Model header: includes pixel size, Z-scale, and information relating model to image coordinates. |
||||||||||||||
|
|
|
||||||||||||
To create a model for a MRC image file called example.mrc
run the following 3dmod command.
This command will cause a new model to be created, ready to edit.
3dmod example.mrc
If a model has already been saved but needs further editing run the
command
3dmod example.mrc example.mod
When this command is run a window similar to the one below will open
up.
A secondary window containing the image of the image file and model
file is opened after the data has been loaded.
Note that the files used for the examples shown below, dual.rec
and example.mod, are among the sample
data files available from our web site.
The title bar of the Info Window shows the name of the image file and also contains controls for minimizing the window or closing it, which is a quick way to exit the program. Below the menu line are some buttons that control window behavior, with the name of the current model file shown on the right. The peg button allows you to keep the Info Window on top of windows from all other programs. The raise button (two up arrows) will bring all of the windows of the current 3dmod above windows from other programs. The buttons with the curved arrows will undo or redo changes to the model. The "Show point" button will be explained in a moment.
Below these buttons are a set of controls called spin boxes, which are now used extensively in 3dmod. A spin box shows a current value of some parameter, and lets you change that parameter either by increasing or decreasing it with the up and down arrow buttons, or by typing in a new value. The three spin boxes on the left show the current model object, contour and point, as well as the color of the current object. The model being edited above has 4 objects; the 4th object is currently selected. The 4th object has 11 contours and the 5th contour is currently selected. This contour has 20 points and the 10th point is the current point.
Generally, when you change object, contour, or point with one of these spin boxes, the program will show the new current point in the image windows and change the section being displayed if necessary. Sometimes you will not want this to happen, such as when you want to switch between objects without having the displayed section change. In that case, just turn off the "Show point" checkbox.
The three spin boxes on the right are the image position controls. They show that the image dimensions are 572 by 378 and that there are 31 sections. The current image point is at the location (237,259) with the origin at the lower left corner of the image. The current section is number 16. If you change the Z value, a different section will be shown in image windows. You can change X and Y values to move the current image point. The current image point is not the same as the current model point. It is marked by a cross when the program is not in model mode, and it determines what coordinates are displayed in some of the image windows.
At the bottom of the Info Window are the Black and White sliders, which you can use to adjust the contrast and brightness of the displayed image. Pixel values between the selected black and white levels (147 and 234) are stretched out to occupy the full dynamic range of the display. Moving the sliders toward each other will increase contrast, while moving one or both to the left will increase brightness.
These sliders have features common to nearly all of the slider controls in 3dmod. Clicking to one side of the handle with the left mouse button will change the value by one. Clicking with the middle mouse button will jump the slider to the position of the mouse pointer. Dragging the handle by clicking and holding the left mouse button will cause the image contrast to be continuously updated until you release the button. In other words, these sliders are continuously active, or "hot". Sometimes the program will not perform well when a slider is continuously active, such as with very large images or models. You can keep a slider from being hot by holding down the Ctrl key (Apple key on the Macintosh). When you do this, the numeric value will be continuously displayed but the image will not be updated until you release the slider.
Below the sliders are several controls related to setting the contrast.
The "Movie" and "Model" radio buttons allow a user to select one of the two modes. When "Movie" is selected, pressing the middle or right mouse button will cause the image to movie back and forth. When "Model" is selected, the three different mouse buttons will edit the model, as described below.
The bottom panel of the Info Window contains additional information and status messages. You can expand the window vertically to see more of the information at once.
 |
 |
This display of two nearby slices from a tomogram (sections 12 and 16) illustrates the three kinds of model objects: closed contours, open contours, and scattered points.
The type of contours in an object, and the symbols displayed at individual points, can be selected with the Object Edit dialog box. Here are two examples.
 |
 |
On the left is a box for the open contour object that is used to track microtubules between sections. A circle is selected for the symbol display, and the symbol size has been set to 7, so that it will be easy to see the symbol when there is only one point on a particular section. On the right is a box for the scattered point object used to model small vesicles; here no symbol is selected but the sphere radius for points has been set to 14.
Model points are added or changed using a 3-button mouse with the program in Model mode. The table below gives an overview of the mouse controls in Movie and Model mode.
| Mouse Button | Movie Mode | Model Mode | ||||
|---|---|---|---|---|---|---|
| Left | Select position or drag mouse to pan image. | Attach to nearby model point or drag mouse to pan image. | ||||
| Middle | Movie image forward, or stop movie. | Add new model point or drag mouse to add several points. | ||||
| Right | Movie image backward, or stop movie. | Modify the current model point or drag mouse to modify several points. | ||||
When you first start a new model or a new object in a model, your first model point will be added to an empty contour. There are two basic ways to add model points to a contour. You can click the middle mouse button to add each individual point after moving to a desired position. The other way is to hold down the middle button while dragging the pointer over a desired path, which will add closely spaced points along the path.
Once you have finished a contour and want to start another one, you need to create a new contour. This can be done with the menu entry Edit-Contour-New, but is much more conveniently done by pressing "n", which is one of many hot keys available to assist modeling. If you are modeling with closed contours, a new contour is created automatically when you switch to a new section and press the middle mouse button. When you are modeling with open contours, this does not happen automatically unless you select the option to "Start new contour at new Z" seen above in the Object Edit dialog box. This option is not turned on for modeling microtubules, but it is useful for drawing open contours that are supposed to be confined to one section, such as the ones along the chromosome boundary.
As you model, the last point added is left as the current model point. You can change the current model point by clicking with the left mouse button near a point in any contour that is present on the section. This is one way to change the contour or object being modeled. If you click with the left button at a position that is far from any model points, the current point and contour become undefined. If you then add a model point, it will be placed in a new contour, so this is yet another way to begin a new contour.
The right mouse button is used to modify existing points by either moving or deleting them. Use this button alone to move one or more points; hold down the Ctrl key to delete points under the cursor while clicking or holding down the right button.
The toolbar for the Zap window has a number of useful controls.
3dmod has many hot keys, which are listed both in the man page for 3dmod and in a help page that can be accessed from the Help-Hot Keys menu entry. These hot keys will work from all of the 3dmod image display windows, and even from the Info Window or other dialog boxes. There are also hot keys active only in particular windows, which are described in the help for those windows. The most commonly used hot keys are listed here:
PageUp Show the next section (increase Z)
PageDown Show the previous section (decrease Z)
Home Show the last section
End Show the first section
Delete Delete current model point
- Zoom down
= Zoom up
s Save the model
n Start a new contour
Ctrl-S Snapshot an image window to a Tiff file
Shift-S, Ctrl-Shift-S Shapshot to non-Tiff file (usually JPEG and PNG)
t Toggle model drawing off or on
, Slow down movie or model rotation
. Speed up movie or model rotation
 |
 |
The XYZ Window shows not just the XY plane but also the two
orthogonal
planes, YZ and XZ, in separate view boxes. It allows you to movie
through the volume in any
of these planes independently, and also displays the intersection of
the model with each of the three planes. In the example, the XY plane
is in the lower left, the XZ plane is above it, and the YZ plane is to
the right.
The current image point governs which planes are
displayed; its position is marked by a small yellow cross in each
image panel. In the figure on the left, you can see the large modeled
vesicle in the XZ (top)
plane, and
an oblique slice through the microtubule marked by the small yellow
cross in the YZ (right) plane.
When the XYZ Window is first opened, the window size, the zoom, and
the division of the window among the three view boxes are set so that
the whole image will appear in each of the view boxes. You can
then resize the window if desired, and the program will maintain the
same proportion among the view boxes. If you make the window
smaller or zoom up, then not all of the image will appear, and you can
pan with the left mouse button in any one of the view boxes to bring
features of interest back into view. Panning in one plane will
typically shift the image in another plane so that all three planes
will always show the same feature. If you find that the top and
side view boxes are too narrow, which can easily be the case after
zooming up, then you can reapportion the window area among the three
boxes by dragging the light blue square diagonally with the left mouse
button.
The toolbar of the XYZ Window has zoom controls and a low/high
resolution button as in the Zap Window. There is a centering
button for bringing the current point as near to the center of the view
boxes as possible. In movie mode, it will center the current
image point marked by the yellow crosses; in model mode it will center
the current model point. The toolbar also has sliders for
riffling through the X, Y, or Z coordinate.
The colored marker lines in the margins of the images also indicate the current point coordinates. They contain small handles that can be grabbed with the mouse to riffle through the images in one plane, as an alternative to the toolbar sliders. The color of the marker line matches the color of the border around the image that is controlled by it; for example, the red line controls the plane selected in the upper panel.
The display of the model in the XY plane is essentially the same as in the Zap Window. In the other planes, scattered points will show up as circles, and other segments of contours will appear to the extent that they intersect the plane.
In Movie mode, the mouse buttons behave the same as in the Zap Window, and they act separately in the three image panels for selecting the current image point or starting a movie in that panel. Basic modeling can also be done in the XY plane, again with the same mouse button functions as in the Zap Window.
 |
 |
The Slicer Window provides a much more flexible view of structures than the XYZ Window because it can display a slice at an arbitrary orientation through the image volume. The orientation is set by adjusting the X, Y, and Z rotation sliders in the second toolbar. In the example, one of the oblique microtubules was centered in the window, then the Z rotation was adjusted to orient the microtubule vertically, then the X rotation was adjusted to bring the microtubule into longitudinal section. Small changes in the X and Y sliders usually cause big changes in the image appearance, whereas the Z slider will primarily rotate the image clockwise and counterclockwise unless X or Y angles are large.
When a Slicer is opened, the center of the displayed slice is
initially at the current
image point, but it can then be panned in the plane of the slice using
the left mouse button, just as for other windows. You can step
through slices in the direction perpendicular to the displayed slice
using the PageUp and PageDown hot keys. The center of the window
is marked by a red cross; this is the center of rotation when the slice
angles are changed.
There is an alternative method of changing the slice angles which
can often be more convenient and intuitive than adjusting the
sliders. When you hold down the Shift key and the middle mouse
button, you can rotate the volume around an axis perpendicular to the
direction of mouse motion. For example, the image on the right
shows the result of dragging with the middle mouse button in the
direction of the red arrow until some of the microtubules come into
good cross-section. You can also rotate in the plane of the slice
by holding down the Shift key and dragging with the right mouse
button. The Shift key also converts the numeric keypad hot keys
from operating on an individual axis angle to rotating the volume
around X, Y or Z axes relative to the current slice; in the latter
case, these keys operate the same as for the model view window shown
below.
The most useful items in the Slicer toolbars are the following.
If the Ctrl key is held down, the middle and right mouse buttons
will start a movie in the Slicer
window, as in the Zap window. The program will movie through
slices in a direction prependicular to the current slice, and will even
adjust its direction if you change the rotation sliders while it is
movieing.
It is possible to model in the Slicer, again with the same mouse
button functions as in the Zap window. This provides a convenient way
to track trajectories in 3D. Points
can be drawn continuously by moving the mouse while holding the second
mouse button down. The PageUp and PageDown keys will move
the center point by one pixel perpendicular to the plane being viewed,
which allows you to draw evenly spaced
contours at oblique angles. See the Slicer help page
for the more details.
The power of the Slicer comes at a cost - it is the first window that we have seen where the computation time can easily make the display unacceptably slow. Display time depends on four factors: image thickness, high versus low resolution, image size, and the number of processors available. The program will take advantage of up to 4 processors automatically. If you need to average through many slices with a high resolution display, be sure to make the image as small as possible. If you must work with a very slow slicer window, remember that the hot sliders can be tamed with the Ctrl key.
The Multi-Z window provides another way to view 3D data, by presenting a gallery of slices side by side. With the controls in the toolbar on the right, you can choose how many slices to show, whether to show them in a row, a column, or an array, and the spacing between successive slices. You can also control whether the model is drawn in the middle panel (in the example, it is not), and whether it is drawn in the other panels. This window is a variant of the Zap window and the first toolbar has a subset of the features in the standard Zap toolbar.
The Z-Scale is the ratio of the thickness between successive images
to the pixel size. Its two most common uses are to adjust for
section thickness in serial section reconstructions and section
thinning
in tomography. For serial sections, divide the section thickness in
nanometers by the pixel size at which the data were digitized, again
in nanometers. For tomograms, divide the original thickness at which
the section was cut on the microtome (in nanometers) by the actual
thickness of
material in the tomogram (pixels of material times pixel size in
nanometers). In either case, if you have an estimate of how much
the specimen shrank laterally due to beam exposure, you should multiply
by the shrinkage factor (e.g., 0.9 for 10% shrinkage). For tomography, the
factor should also be adjusted for lost material if you have an estimate
of that; see the
Model Header help page for a formula.
The Pixel Size field should contain both a value for the size of the pixel in the digitized data, and the units for that value. Units can be one of the following; A, nm, um, mm, m, km. For example, you could enter "0.015 um" or "15 nm".
 |
 |
 |
These are different views of the example model in the Model View
Window, which can be opened while running 3dmod by selecting Model
View from the Image menu, or by typing the hot key "v".
This same window can be opened in
a standalone mode with
3dmodv example.mod
The model has been
tilted around the X and Y axes using the arrow keys on the keypad.
The model can also be rotated with the middle mouse button, or using
arrows and text boxes in the
Control dialog box that can be opened using the Edit-Controls menu
entry, but the arrow keys are often more convenient. A single keystrike
rotates the model by one step. Continuous rotation is started by
pressing the keypad Enter key followed by the desired arrow key;
another press of the Enter key stops the rotation.
On the left is the initial view of the model, before any of the default viewing parameters have been changed. These parameters are initially set to show open and closed contour objects with contour line tracings, and scattered points as wire baskets. The middle view shows two changes that can be made immediately, without having to compute surface meshes for the contours. The microtubule lines (green) have been made thicker and the scattered points are displayed as lighted surfaces. On the right is the display after making meshes for the objects (as explained below) and setting all objects to display as lighted surfaces. These display settings are all changed in the 3dmodv Objects dialog box, opened with the Edit-Objects menu item or the O hot key.
 |
 |
 |
The Objects dialog box contains controls for selecting the current object to be edited, buttons for turning individual objects on or off, and a set of panels for setting different object properties or controlling the overall display. A panel is selected by clicking on its name in the list.
The example on the left shows how the appearance of the green lines was changed for the middle model view panel above. The "Lines" panel was selected, the 3D line width was increased from 1 to 3, and "Anti-alias rendering" was turned on to smooth out the jagged edges of the lines.
The example in the middle shows how the scattered points were displayed as solid objects. The selection for "Draw Data Type" was changed from "Contour" to "Mesh". 3dmod automatically changed "Drawing Style" from "Line" to "Fill" to produce the most commonly desired appearance.
The example on the right shows a change that was required after meshes were computed and displayed for all objects. By default, objects are lit only on one surface, which is always the outer surface for closed contour objects. (Note the dark surfaces on the interior of the microtubules.) However, a sheet like the chromosome boundary (yellow) does not have an obvious interior surface, and its lit surface happened to face downward, toward the chromosome. This problem was solved by checking the "Light Both Sides" option in the "Material" panel. The other method of solving this is to go back to the 3dmod Object Edit dialog box and change the "Front Face" from "Outside" to "Inside", then recompute the mesh.
The appearance of the magenta scattered point was improved between
the middle
and right views of the model by pressing the "g" hot key twice to
increase sphere drawing quality. When there are many spheres, this
degree of quality could
significantly increase drawing time for the display, so it is not the
default.
The sphere drawing quality can also be controlled in the "Points"
panel, which allows one to change the global quality setting (just as
with the "g" and "G" hot keys) as well as set a quality for individual
objects.
When the Model View window is opened from 3dmod, it is possible to
display the image data in 3D in this window in two different ways.
One way is to draw one or more planes of the image along with the model.
Use the Edit-Image menu entry to open a dialog to control many aspects
of this image
projection. Multiple planes are drawn with enough transparency so that
you can see through to the rearmost image plane; you can add more transparency
to see the model through the images as well. The following pictures show
an example of this display.
 |
 |
The other way to display image data is to draw a surface called an isosurface at a particular threshold level. The middle image below shows an isosurface rendering of the two basal bodies whose cross-section appear in the Zap window on the left.
 |
 |
 |
Select Isosurface from either the Image menu of the Info window or the Edit menu of the Model View window (or use the U hot key) to start the isosurface display and open the dialog to control it. The isosurface display is most useful when looking at macromolecular structures in averaged subvolumes from tomograms, but may also be useful for examining tomograms directly, particularly if the data are binned and small unconnected fragments are deleted from the display. The threshold can be adjusted to isolate features of interest, and the other sliders in the dialog can be used to select the position and the size of the volume being rendered.
There are two ways to relate positions on the isosurface to features in the image data. First, if you click on the isosurface with the right mouse button, the selected point will become the current image point in 3dmod. Second, if you turn on the display of the current point in the View menu of the Model View window, you can click on a point in the Zap or Slicer window and see that position marked in the Model View window with a red sphere (as seen above). Both of these methods will work better if you uncheck Link to global X, Y, Z so that the position of the volume being rendered does not change when you click in either place.
The display properties of the isosurface can be adjusted from the Edit-Objects dialog just as for actual objects in the model. The isosurface is kept in something called an "extra object" and it is assigned a number higher than the number of actual objects. The isosurface will disappear when you close the dialog, but you can use the Save button to move it into a new actual object in the model.
Skinning is the name used for calculating a three-dimensional
surface from contour data. This surface, once calculated, is
represented by
a mesh of triangles that can be stored within each object. If any contour
points within that object are edited then the mesh
for that object must be recalculated in order for it to match
the contour data. There are two ways to calculate the mesh from the
contour data within a
model: with the imodmesh program, and
with the Meshing panel of the Objects dialog box. Using imodmesh
provides a wide range of
options for controlling the meshing; whereas the Meshing panel provides
a more intuitive interface with the most important options.
To skin the example model, this command was used:
imodmesh -C -t 1 -d 10 example.mod
By default, imodmesh will construct meshes only between pairs of contours, which will leave holes at the top and the bottom of a closed contour object because there is nothing to connect the highest and lowest contours to. The "-C" option tells imodmesh to fill in these holes or "Cap" them off.
In the model view on the right, the lines of the microtubules appear
as tubes because imodmesh created cylindical meshes around them. It
did this because of the
"-t" option, which is followed by a list of open contour object numbers
that should be given tubular meshes
(here, just object 1). The "-d" option is then used to specify the
diameter of the tubes, in pixels (10).
imodmesh -C -o 4 -R 0.75 -i 2 example.mod
Here, a new mesh is created only for object 4 (-o 4), and
points will
be skipped, but the surface defined by the mesh will always be within
0.75
pixel of the original data (-R 0.75). In addition, the "-i 2" makes
imodmesh connect every other Z plane (i.e., at intervals of 2 in Z),
reducing the mesh size by 2.
Each time you run imodmesh, the existing file is saved as a backup file with a "~" added to its name. You can also erase all of the mesh data from your model using the "-e" option to imodmesh.
Imodmesh also has an option (-l) to store a low resolution mesh
alongside a
regular mesh, so that a big model can be manipulated rapidly in
low-resolution mode then switched to high-resolution mode for studying
details.
Other commonly needed options are -s and -P to connect contours across
gaps; see the imodmesh man page for
details.
All the options just mentioned are presented in the interface in the
Meshing panel:
 |
 |
Whenever an object is meshed, either with imodmesh or through this
interface, the options used to mesh the object are stored in the model,
and their values will be displayed when this panel is opened.
This makes it easy to remesh objects without having to remember exactly
what options were used previously. On the left, the "Tol" and "Z
inc" settings reflects the values entered with the -R and -i options to
imodmesh, and the "Cap" checkbox indicates that the -C option was used
for the vesicle object. On the right, the "Tube" selection and
the
diameter are shown for the microtubule object. When the "Make
Mesh" button is pressed, the meshing process is started in a separate
thread so that you can continue working in 3dmod while the mesh is
being computed.
If you do mesh with imodmesh while you are modeling in 3dmod,
be sure to save the model, run imodmesh, then immediately load the
model back into 3dmod with the Reload Model entry in the File
menu.
To display a scale bar in any of the image windows, or in the Model View window, select the Scale Bar entry from either the Edit menu of the Info window or the View menu of the Model View window. The following dialog will appear.
Scale bars will be drawn in the various windows as long as the dialog is open and Draw scale bars is checked. You can select whether the bars will be white or black in the image windows; the bar will always be white in the Model View window if the background there is black. The various controls allow you to adjust the size, location, and orientation of the bars. At the bottom is a list of the scale bar sizes in the various kinds of windows; you will need to record this information when you take a snapshot.
Because IMOD does not produce files that can be used in movie
players, if you want to make a movie of your images or models, you need
to make a series
of snapshots of a window and import those images into movie-making
software. (For Windows, we recommend Videomach from
www.gromada.com for making movies). There are two dialogs that help manage this task:
 |
 |
On the left is the dialog box opened with Edit-Movies from the 3dmod
Info Window menu and used to control movieing and automatic
snapshots in image display windows (primarily Zap and Slicer).
The sliders at the top allow you select the range and increment between
slices to be snapshot. Here, every slice from 6 to 56 will be
snapshot. The Looping
radio buttons control how many pictures will be taken when you start a
movie. With Round trip
selected, pictures will be taken going in both directions, which is
useful if your movie playing software will not loop back and
forth. If your movie software will loop back and forth, you can
select One way and get
pictures going in just one direction. The Start at radio buttons allow you to
choose whether the set of pictures will start and end at the limits in
Z (Start/end) or at the
currently displayed section (Current pt).
To get a series of pictures, select one of the file formats in the Snapshot box and start a movie in a
Zap or Slicer window with the middle or right mouse button. The
window will stop movieing when the set of pictures is finished.
On the right is the corresponding dialog for movies, opened with
File-Movies from the Model View Window menu. In the most general
case, you can adjust the
model display to the desired starting position, orientation, and size,
and press Set Start; then
adjust the display to the desired final state and press Set End. Often one simply
wants the model to spin all the way around, which is provided by the
buttons Full 360 X and Full 360 Y. When you press Make, you will see the model go
through its sequence of positions. If it is doing the right
thing, select a file format and check Write
files, then press Make
again.
The different image formats differ in their degree and type of
compression. TIFF files are uncompressed and can be quite large,
since they are 24-bit color. PNG files have a loss-less
compression that is quite effective for most model snapshots but much
less effective for image snapshots. JPEG files will give the most
compression, with some loss of information that can be controlled by
the percent quality factor, selectable in the 3dmod Preferences dialog. For images
with no model overlay, 80% quality is usually adequate for making
movies. If there is a model overlay, you need 90% or higher
quality to make it look decent in the individual snapshots (images here
were shot at 90% quality). It might be preferable to take such
snapshots as PNG's and let your movie software take care of the
compression.
Note that both of these dialogs have a montage option for capturing
high-resolution snapshots that are larger than the image window
itself. In each case, the program zooms up the display and pans
around to capture a montage of images, then assembles these into a
single image and writes a TIFF file. A movie of such montaged
snapshots can be saved from the Zap window, but only a single frame can
be saved in high-resolution mode from the model view window.
 |
 |
 |