Document 404548

Basic 3D reconstruction in Imaris 7.6.1 Task The aim of this tutorial is to understand basic Imaris functionality by performing surface
reconstruction of glia cells in culture, in order to visualize enclosed structures and perform 3D
measurements. Ultimately, we will create a 3D visualization animation of our dataset.
Key contents of tutorial -­‐
-­‐
-­‐
-­‐
-­‐
Visualize dataset
Setting the scale
Mixed Model Rendering
3D measurements
Generating movies
Load stack to Imaris In this example, we will be using a sample 3-channel z-stack obtained on a Zeiss LSM 510 confocal
microscope. The stack is included in one .lsm file. Imaris can read most file formats created by
various microscopes, as well as single TIFF files and many more. To import a stack of single tiff
files, go to “File”, “Open…”, click on the first one image and click Settings. The images should be
exported and saved in a way that Imaris can recognize channels, z position and time point for the
stack to be loaded properly. See Figure 1 and 2.
Figure 1. Imaris main window and dialogue for opening image sequence. Select always the first of the images from
the particular stack, and make sure the file name is formatted in a way where each image is a specific channel of the
multi-dimensional data (blue arrow). To assign specific dimensions to images, click ‘Settings’ (red arrow).
Channel
Stack position
Always check
whether the entire
stack is recognized
Time point
Figure 2. Using the ‘Settings’ dialogue (see Figure 1) the multidimensional dataset can be loaded properly, by
assigning different attributes to images depending on their filename (spatial and/or temporal position on stack, and
channel). This step is of pivotal importance for appropriate multidimensional analysis.
Set the scale The image dimensions or pixel to µm scaling have to be set according to the acquisition settings.
Typically, this information is contained within the image metadata, or is reported by the software
when the acquisition is performed. In certain occasions, with files directly produced by most
microscope manufacturers, Imaris can read the metadata automatically and set the image scale. But it
is always a good practice to confirm that these are correct. To set these, go to ‘Edit’, ‘Image
Properties…’. In the ‘Geometry’ dialogue (Figure 3), use the Voxel size (typically reported with the
same name in the image metadata) to set the µm corresponding to one pixel. In z, typically the
optical slice thickness is reported.
Figure 3. The scale of the stack is set by entering the corresponding µm to 1 three-dimensional pixel (volume
element - voxel) of the stack. Z typically corresponds to the value of the optical slice thickness. Additionaly, when
there is a time-lapse stack, one can set the time point interval.
Verify stack After loading has been completed, use the ‘Slice’ button to verify that the stack has been loaded
correctly. Use the arrow next to the button to select the ‘Gallery’ option in order to visualize the
stack, merging the individual channels (Figure 3). This gives an overview of the stack slices, as well
as the optical thickness and imaging depth, given that the scale is set correctly. The stack can also be
examined as XZ and YZ slices, by selecting ‘Section’ using the arrow of the ‘Slice’ button. In our
case, this visualization can give a first impression as to whether there are structures enclosed in the
cells we are examining, before we proceed to 3D rendering. See Figure 4. At this point, the ‘Display
Adjustment’ dialogue can be used (if it is not active, enable from ‘Edit’, ‘Show Display Adjustment’)
to adjust the channel contrast for best visualization.
Figure 4. Section view of the stack, where XZ and YZ slices can be examined, additionally to the XY view. Planes can
be changed in any view by dragging the position of the white crosshair lines (e.g. blue arrow). Using this view, we can
preliminary determine whether there are potentially intracellular inclusions. The contrast of the various channels can
be adjusted using the sliders in the ‘Display adjustment’ (boxed). This is a histogram based enhancement; it will not
affect your data, and should only be used to improve 3D reconstruction and visualization, not to draw conclusions
regarding the expression or presence of any ligands.
Examine 3D morphology Di
Ad spla
ju y
st
m
en
t
Ob
L i je c
st
t
Vi
ew
Ar
ea
Use the ‘Surpass’ button to examine the 3D morphology of your stack. This is a primary intensitybased rendering of the 3D volume of the stack, and the basis of the 3D reconstruction area. The
workspace is divided in 4 main parts (see Figure 5), namely the view area, the object area and the
object property area and the visualization area. The ‘Display Adjustment’ window is also useful in
this case to adjust for best visual contrast.
Ob
Pr jec
op t
er
t ie
s
Visualisation
Area
Figure 5: The main areas of the ‘Surpass’ view. This view facilitates the image navigation, selection and interaction
with the 3D volume. The ‘Object’ window allows the selection of a combination of various visualisation techniques; by
default, the ‘Volume’ object presents an intensity rendering of the inserted stack. The object properties can be
adjusted from the respective window. The intensity rendering contrast can be adjusted for appropriate visualisation
using the ‘Display Adjustment’, which is essentially a histogram-based adjustment control pane.
Surface Reconstruction Reconstructing the surface of a structure is a useful visualization tool that can provide insight on the
morphology of the structure, its 3D architecture as well as its interaction with nearby structures. In
our example, we will use it to unveil intersections and inclusions between the microglia body (blue),
the amyloid (a protein which they are supposed to phagocytize - red) and the microglia nucleus
(green). The reconstruction will be analytically explained for the blue channel, and the reader may
perform the reconstructions of the other channels as an exercise.
To start, we want to create a new surface, by clicking on the surfaces icon (
toolbar:
) in the Objects
A new object has been added into the Objects list (see figure 5). For the creation of the surface we
will be working at the ‘Object Properties’ workspace (see figure 5); it is advisable to use the Display
Adjustment to enhance the contrast of the channel we are reconstructing, in order to evaluate the
precision of the reconstruction. Once this is done, go to the ‘Object Properties’ workspace (shown
on the left of figure 6). We want to segment the whole image, but there is the option of only
reconstructing a part of the image. This option can also be used to speed up the algorithm, by
applying it to a part of the image in order to fine-tune the parameters, and then apply the final
reconstruction to the whole image. Click the blue button with the arrow on the bottom right of the
dialogue to continue.
Figure 6: The first steps of the surface reconstruction wizard.
In the next dialogue (2nd image in figure 6) we have to select the appropriate channel to be
reconstructed (that would be channel 3 in our example). Then, the level of detail can be chosen. This
of course depends on the resolution (both axial and lateral) of the dataset, and is automatically set
according to the Nyquist theorem to 2x the pixel size, thus 0.140µm in our case. However,
depending on the noise level of the dataset, and other imaging parameters such as oversampling, this
can be set to coarser values, resulting in smoother surfaces. Here, we will set it to 0.21µm, as this
provides with enough detail and does not result in too noisy of a surface.
The volumes to be reconstructed can be segmented using either an absolute intensity threshold
(distinction between foreground and background based on the intensity value) or a more advanced
local contrast method. The local contrast method is recommended for use when there are structures
with uniform illumination and smooth surfaces, of ovoid, ellipsoid or spherical shapes that are
touching each other and have to be segmented and split. The local contrast operates on the volume
by fitting a sphere into the object, of diameter set by the user. In our case, due to the morphology of
the structures, this would result into over-segmentation. It would be more appropriate for use to
reconstruct e.g. the nucleus (green channel). Therefore, we make use of the Absolute Intensity (see
figure 6, 2nd image), and proceed to the next step using the blue forward arrow.
In the next dialogue, the surface coverage is calculated based on the aforementioned threshold,
which can be manually set using the slider (figure 7). At this point, it is advisable to check the
accuracy of the surface coverage (gray surface overlay) by adjusting the contrast of the
reconstructing channel, using the ‘Display Adjustment’ (see figure 5 and 7).
Figure 7: surface coverage adjustment using the absolute intensity measure (threshold). This can be adjusted
manually by dragging the edge of the yellow slider (red arrow), showing the gray values that will be considered for the
reconstruction. The thresholds can also be numerically entered into the fields above (blue arrow). In some cases, it
may be necessary to exclude high intensity structures; the threshold can also be adjusted to exclude high values by
dragging the other end of the yellow slider. Using the Display Adjustment (orange arrow), the contrast can be tuned to
check the reliability of the surface coverage.
Once the threshold has been adjusted, there is the option of segmenting the volume using a region
growing method, which enables touching objects to be split. This is applicable in cases where there
are adjacent structures that are segmented as one by the threshold. In this case, using the threshold
Imaris calculates certain seed points inside each surface (a dialogue appears in the next step, allowing
the user to fine tune this) and applies a region growing algorithm, resulting in distinct split objects.
Again, caution must be exercised as using it with the wrong parameters can result in oversegmentation. In our dataset, there is no need for such a function, so the box is left unchecked.
Proceeding to the next step, Imaris will then calculate for the selected settings the surfaces and will
render the resulting isosurfaces. This step may take some time, depending on the processing power
of the workstation. Once finished, the user has the possibility of selecting which surfaces to keep,
based on a variety of filters, according to the relevance of the structure to the data that need to be
visualized. Figure 8 illustrates this dialogue.
Figure 8: The final surface reconstruction step. The user can exclude or include the rendered isosurfaces to the final
image, using a variety of filters (orange arrow), which can be tuned to the required precision. There is an extensive
comprehensive list of filters which can be used depending on the threshold result of the previous step, and on the
desired data visualisation; here, we make use of the ‘Number of Voxels’ filter, to exclude the extremely small
structures. The number was set 300 (smallest allowed structure would have 300voxels).
After choosing the appropriate volumes to be reconstructed, based on the filters selected, the final
surface will be constructed by pressing the green forward button. The rendering of the surfaces can
be further manipulated using the ‘Object Properties’ dialogue (see figure 5); statistics about the
volumes can also be obtained here. In figure 9 the final rendered volume is presented, with a
different color, refrelctivity and opacity (all modified using the colour type button
); the initial
intensity volume renderings have been removed using the
button.
Figure 9: Final isosurface rendering of the microglia volume. Image on the left depicts only the isosurface; image on
the right includes the rendered volumes of the other channels. Changing the isosurface transparency shows that these
structures are included in the surface.
Volume Statistics Each isosurface created has distinct spatial properties, which upon reconstruction are readily
available in Imaris, and can be exported to Excel. Figure 10 illustrates these for the central microglia
cell (selection shown in yellow).
Figure 10: Each isosurface object has certain spatial
properties and can include components from all
channels (e.g. shown by the blue arrows). These
statistics can be easily exported for processing.
Using XTenstions Imaris has the possibility of incorporating MATLAB and ImageJ code to the analysis. There are
already some MATLAB extensions implemented, which can be found under the (
) button.
Using these requires a MATLAB license as well. For every type of reconstruction (surface, filament
or spots) there is a variety of readily available XTensions, which can be added to the volume
statistics. User-written MATLAB and ImageJ plugins can also be included in Imaris (see Quick Start
Tutorial 7.6.1 included in Material folder).
Short description of Mixed Model Rendering Isosurfaces is but one of the many possibilities for modeling a volume in Imaris. Depending on the
structures that need to be visualized, Imaris can produce also fitted spheres (
filaments including spine reconstruction (
), neuronal
), or even create a customized intensity volume
rendering (
). These can all be included in the same image to produce a mixed model of the data
stack. For illustrating the mixed model, in our example (figure 11), a sphere had been fitted to the
nucleus (channel 2) and the red structures had been intensity rendered, using the ‘Blend’ function in
the Volume object pane.
Figure 11: Image on the left illustrates an intensity rendering of the original stack; image on the left shows a mixed
model rendering of the dataset, including isosurfaces (blue channel), fitted ellipsoid (green channel) and blend mode
intensity render (red channel).
Animation A short description will be presented here illustrating the potential of Imaris when it comes to
creating animated presentations of the dataset. For this, we will use the reconstructed volumes
shown in figure 11, changing transparencies, cutting through the volume and adding structures as
the animation proceeds. The final movie can be obtained from within the material of the website.
Switch to the Animation tab(
), go to the lower right part of the menu, and click ‘Settings’. On
the ‘Surpass’ pane, all object creation buttons must be enabled (apart from those marked as Custom
Object at the bottom), and the Frame Rate must be set to 24 fps, or higher to ensure smooth
animation. Figure 12 shows the interface of the ‘Animation’ tab, and includes a description of the
basic functions when creating an animation.
Figure 12: The ‘Animation’ tab. The animation effects are based on the addition of Key Frames (red arrow). Each key
frame locks the current properties of the objects (set for each object using the Object Properties – orange frame,
corresponding to the object pointed by green arrow), as well as the rotation of stack. The stack can be manually
rotated using the Navigate cursor, but for increased precision and smoothness of move, rotations of specific angles
can be added (blue arrow). The number of Animation frames (purple arrow) can be increased to make animation
smoother and slower, according to the number of key frames too. At 24fps, it is advisable to select more than 240
frames for illustrating the volume better. In our example, 1000 frames have been selected.
We will create a full rotation in the Y axis, modifying the transparency of the glia body, and
subsequently sectioning it, making the nucleus visible. Additionally, the red channel can be either
used as an intensity volume rendering, or can be reconstructed as an isosurface, which will make the
animation smoother. The colours will also be changed for better contrast. Start by setting the colour
of Surfaces 5 to green, transparency to 0%, colour of Spots 1 to Blue, and click add a key frame.
Rotate the camera by 900, using Custom Rotations (blue arrow, figure 12), around Y axis. This
automatically adds a key frame. To look at the key frame, place cursor on the animation slider, and
click on the line created. Create another rotation like this; select created key frame, change the
transparency of Surfaces 5 to 75% and press ‘Modify’. Create another two 900 rotations. Click the
last created frame, and restore transparency to 0% and press ‘Modify’.
Now, to create a slicing plane: Create a new rotation of 60% along the X axis. Select the newly
created key frame. Add a new Clipping pane, using the respective Objects button (
). Place the
new Object above the Surfaces 5, and below Volume; move spots above it too, so that it will clip
only the glia surface. Figure 13 depicts the clipping plane.
Figure 13: The Clipping Plane object, within the Animation tab. The position of the object in the list (red arrow)
determines which objects are going to be sectioned. The object can be manipulated using the cursor (change from the
Navigate to Select - green arrow) in all directions, or precise clippings can be performed using the Object Properties
tab (orange frame).
Move the plane using the Select cursor, so that there is no clipping. Click Modify Key frame. Then
move the manipulator almost to the end, clipping all glia isosurface. Click Add Key frame. Bring the
clipping plane back to the top, without cutting any surface, and add another Key frame. Finally,
select the very last key frame, make sure everything is back to original position, and if necessary click
Modify.
Set the animation to 480 frames, and click the ‘Play’ button to preview the animation. If no
modifications are necessary, press the ‘Rec’ button ( to create a movie with the animation). A variety
of options is available for the user to select, depending on the desired quality and file size. In our
example, mp4 format was selected, with low compression, of size same as the window size.
Postscript As with every piece of software designed for advanced applications and offering great versatility,
only the tip of the iceberg has been described, from the multitude of features that Imaris includes.
We urge the reader to try out different options when following this tutorial; explore the menus and
the great variety of Image processing and analysis tools offered by Imaris.
Further reading Website: http://www.bitplane.com/
Imaris tutorials: Youtube Channel
https://www.youtube.com/playlist?list=PLB6C0D473A06E9AB4
List of XTensions: http://open.bitplane.com/Default.aspx?tabid=237
A selection of documents explaining the basic principles is also included in the Material folder.
Acknowledgement We would like to acknowledge Dr Jonas Neher of the Hertie Institute for Clinical Brain Research,
Tübingen, Germany, for providing the dataset for this tutorial.