Method note: Adipose tissue quantification in vivo

Quantifying adipose tissue (fat)
in a mouse or rat by in-vivo
microCT
Method note
Page 1 of 26
2
Method note: Adipose tissue quantification in vivo
1. Introduction
1.1. X-ray interaction with biological soft tissues
Fat or adipose tissue is exceptional among animal soft tissues in having a
different x-ray attenuation to other soft tissues, and thus having a distinct
density in microCT images (figure 1).
In microCT scanners employing x-ray accelerating voltages in the range
20-100 kV, with corresponding x-ray photon energies mostly in the range
15-75 keV (after aluminium filtration to reduce beam hardening and
radiation dose), the principal modes of x-ray attenuation are photoelectric absorption and Compton scattering. Which mode predominates
depends on the photon energy and the atomic number (Z) of the
absorbing element according to the relationship shown in figure 2 [1].
Photoelectric absorption predominates at low photon energies and in
heavier elements, while higher photon energies and lower Z nuclei favour
Compton scattering. Photoelectric absorption in material is proportional to
Z4 where Z is the atomic number of the element – this represents an
acute sensitivity of photoelectric absorption to elemental composition.
However unlike photoelectric absorption, Compton scattering has unitary
proportionality to Z.
Figure 1. A reconstructed crossection of a microCT scan of a mouse thorax. The
four tissues which are most readily distinguished in microCT images on the basis
of differing x-ray attenuation, are bone, lung, fat and other “lean” non-fat soft
tissue. All four are clearly visible in this image, the adipose tissue is shown as the
darker grey regions near the periphery of the thorax. Note that in this image
lighter color means higher x-ray attenuation and “density” while darker color
means low x-ray attenuation. (Bone is saturated to white by narrowing of the
reconstruction contrast limits to enhance soft tissue visual contrast.)
Page 2 of 26
3
Method note: Adipose tissue quantification in vivo
This difference is important; it means that photoelectric absorption gives
much stronger absorption image contrast, based on material element
composition, than Compton scattering.
The elemental composition of soft tissue is predominantly hydrogen,
carbon, nitrogen and oxygen. Elemental ratios in some biological tissues
are shown below in table 1. The content of the two major constituent
elements C and O in fat is 57% and 30% while in muscle (typical of soft
tissues) the ratio is C 12% and O 73% [2]. The large difference in the
ratio of carbon to oxygen between fat and lean tissues accounts for the
difference in x-ray absorption at x-ray energies where photoelectric
absorption is the predominant interaction mechanism, that is, at photon
energies less than about 30 keV.
Note from figure 2 that C, N and O are close to the boundary between
predominance of photoelectric absorption and Compton scattering, at
photon energies less than 50 keV. They can be attenuated by both
modalities, lower photon energies will favour photoelectric while higher
energies will favour Compton interaction. (Hydrogen (Z=1) is attenuated
only by Compton scattering.) This is the physical basis for the increase in
tissue contrast achieved by decreasing x-ray photon energy: where
photoelectric absorption is favoured, Z-based contrast will increase.
How do we know what x-ray photon energy is being used during a scan?
And is photon energy the same as the applied voltage in the x-ray source?
No. It is important to note that there is a big difference between applied
voltage, and the resultant x-ray photon energy. Sharp energy peaks from
characteristic x-rays combine with the broad spectra of bremsstrahlung
(braking radiation) to give a wide and complex energy spectrum. The
applied voltage marks the maximum of the spectral range, so that the
mean photon energy is a lot less than the applied voltage. Figure 3 below
shows the spectra of x-ray photon energy obtained with different
combinations of filter and applied voltage (“filter” is simply a sheet of
metal placed in the x-ray beam path, to remove the lowest energy xrays.) Table 2 shows the mean x-ray photon energy obtained with several
settings of filter and voltage relevant to in vivo microCT scanning in the
SkyScan 1076, 1176 and 1178 in vivo scanners.
Table 1. Biological tissue elemental ratios, by mass percentage [2]. A massweighted average Z is also given.
H
C
N
O
Na
Mg
P
S
K
Ca
Z
1
6
7
8
11
12
15
16
19
20
Water
11.2
Muscle
10.2
12.3
3.5
72.9
Fat
11.2
57.3
1.1
30.3
6.4
27.8
2.7
41
Bone
Average Z,
mass-weighted
7.2
88.8
0.08
0.02
0.2
0.5
0.3
.007
6.1
.006
0.2
Page 3 of 26
7
0.2
7.1
14.7
9.1
4
Method note: Adipose tissue quantification in vivo
Figure 2. The interaction modes of x-rays with matter depending on elemental
atomic number Z and the x-ray photon energy hv. Biological soft tissue elements
C, N and O (Z=6,7,8) are near the boundary of photoelectric absorption and
Compton scattering, at photon energies less than 50 keV used in microCT. Since
the sensitivity of absorption to Z is much higher in photoelectric absorption, in
practice contrast between fat and lean tissue is improved by decreasing x-ray
photon energy. (From The Atomic Nucleus, Evans 1955.)
2. Method
2.1. What does this mean for in vivo microCT imaging of adipose
tissue in small animals? What settings to use?
This means that fat can be imaged well by microCT imaging of rats and
mice, providing appropriate x-ray energy is used. But, importantly, x-ray
settings should not cause excessive radiation dose to the animal.
These considerations work in opposite directions. Generally lowering the
mean photon energy will increase contrast between fat and lean tissue
(more photoelectric absorption, less Compton scattering). However
decreasing photon energy will also increase radiation dose from a scan.
Therefore a balance or compromise is required. Scan settings are
recommended which both optimise tissue contrast and reduce radiation
dose.
For the mouse we recommend 1mm Al filter and applied voltage of 40-50
kV. From table 1, this setting corresponds to an average x-ray photon
energy of 25 keV, taking account of camera detection efficiency with x-ray
energy.
Page 4 of 26
5
Method note: Adipose tissue quantification in vivo
Note that a superior contrast between fat and lean tissue would be
obtained using the 0.5mm Al filter with the same 40-50 kV applied
voltage. However the x-ray dose would be almost double that of the
corresponding scan with 1mm Al filter. Thus if an in vivo scan is done just
before sacrificing an animal, where radiation dose is not such an issue,
then the 0.5mm Al filter setting could be used. However if serial in vivo
scans are done in a longitudinal in vivo study design, then the 1mm Al
filter should be used.
If exceptionally low dose is required for mouse scanning for soft tissue
analysis, then the Cu+Al filter (equivalent of 2mm Al) can be used with an
applied voltage of 40 kV. The low applied voltage helps to prevent too
much loss of lean-fat contrast due to the increased x-ray energy due to
the higher filtration.
Table 2. The mean x-ray photon energy from a microCT x-ray source,
depending on filter and applied voltage, with and without a correction for the
decreasing x-ray camera detection efficiency with increasing x-ray photon energy.
Applied
voltage
50kV
Filter
from "effective" spectrum
compensated for X-ray
camera efficiency
Not compensated - as
emitted from
source+filter
15.71
19.27
Al 0.5mm
22.53
26.52
Al 1mm
25.03
28.82
27.47
31.14
No filter
18.02
24.12
Al 0.5mm
26.91
32.64
Al 1mm
30.31
35.32
33.95
38.04
No filter
20.77
29.39
Al 0.5mm
31.26
38.42
Al 1mm
35.23
41.24
39.41
44.05
Al 2mm
90 kV
Mean keV
No filter
Al 2mm
70kV
Mean keV
Al 2mm
1
1
1
1
In the SkyScan 1076 or 1176 in vivo scanners, the “Cu+Al” filter is equivalent to about
2mm Al filtration (40 microns of Cu with 0.5mm Al).
Page 5 of 26
Method note: Adipose tissue quantification in vivo
Different filters with 50 kV applied voltage
1.E+17
No filter
Ti 25µm
Count (LOG scale)
Al 0.125mm
Al 0.25mm
1.E+16
Al 0.5mm
Al 1mm
Cu 38µm+Al 0.5mm
Cu 0.125mm
1.E+15
1.E+14
0
10
20
30
40
50
60
photon energy keV
Different filters with 100 kV applied voltage
1.E+17
No filter
Ti 25µm
Al 0.125mm
Count (LOG scale)
6
Al 0.25mm
1.E+16
Al 0.5mm
Al 1mm
Cu 38µm+Al 0.5mm
Cu 0.125mm
Cu 0.4mm
1.E+15
1.E+14
0
20
40
60
photon energy, keV
80
100
Figure 3. The spectrum of x-ray photon energy for different combinations of xray filter and applied voltage. The large peak of low energy x-rays at about 10
keV is characteristic of the tungsten x-ray source target, however this peak is
effectively removed with filtration with 0.5mm Al or more.
Page 6 of 26
7
Method note: Adipose tissue quantification in vivo
For small rats up to about 300g, 1mm Al filter can also be used, with 6080 kV applied voltage. For larger rats over 300g, the Cu+Al filter in the
SkyScan1176 (which is equivalent to about 2mm Al filter) should be used
with applied voltage of 70-90 kV. Recommended scan and reconstruction
settings for mice and rats, for in vivo quantitative assessment of adipose
tissue, are shown in table 2 for the SkyScan 1176 in vivo scanner.
For the SkyScan1178 in vivo scanner, there is a fixed 0.5mm Al filter. For
assessment of adipose tissue in the mouse, 40-50 kV should be used, and
the maximum 65 kV for the rat.
Table 3. Recommended scan and reconstruction settings for mouse and rat
microCT in vivo scans to quantify adipose (fat) tissue. This is for the SkyScan
1176 in vivo scanner.
Mouse
Offset scan (double width)
X-ray filter
X-ray applied voltage2
Pixel size, microns
Scan rotation 180 / 360
degrees
Rotation step
Frame averaging
Rat ≤300g
Scan parameters
No
Yes
1mm Al
1mm Al
Rat >300g
40-50 kV
35
360
60-80 kV
35
360
Yes
“Cu+Al”
(=2mm Al)
70-90 kV
35
360
0.6-0.9
1-2
0.4-0.7
1-2
0.4-0.7
1-2
Reconstruction parameters
3-5
3-5
3-5
Yes if large bed No (unless part Yes
is used
of animal goes
outside FOV at
any part of
scan rotation)
Ring artefact reduction
2-4
3-5
3-5
Beam hardening correction
30
30
30
%
Defect pixel mask (advanced 5%
5%
5%
tab)
Contrast limits (attenuation
Minimum: 0
values for grey 0 and 255)
Maximum: adjust for soft tissue contrast, allow
bone to saturate (max line at about half way point
on spectrum)
ROI
Apply ROI to include all animal but minimise outer
ambient space.
Smoothing kernel radius
Object > field of view
2
The source current setting is variable; if synchronisation is not done (not
required for adipose assessment) then current can be reduced with a reciprocal
increase in exposure time, to lower radiation dose – see the corresponding
method note.
Page 7 of 26
8
Method note: Adipose tissue quantification in vivo
The parameters for scanning and reconstruction for mice and rats (smaller
and larger) are shown in table 3. In general there is not much variation in
scan parameters for mice or rats in vivo, whether one is imaging the
animal to study bones or soft tissue (the animal remains the same in
either case). The same filter and voltage parameters are appropriate for
bone scanning, however higher resolution is needed, of 18 or 9 micron
pixel.
2.2. Mounting animals for scanning
Mice and rats should be placed supine (back down) on the animal bed.
Mice can either be scanned in the 3 cm diameter carbon bed, or
alternatively in the 6 cm diameter carbon bed with the appropriate
polystyrene foam insert (outer diameter 6cm inner diameter 3cm). This
latter arrangement has the advantage that the reconstructed crossections
in the mouse are surrounded by very low density polystyrene foam so that
the animal body is easier to automatically segment. By contrast the
carbon fiber bed has x-ray density closer to animal soft tissue so if the
animal lies directly against the carbon bed, it can require extra time and
effort during analysis to separate the animal body from the bed.
Figure 4. Upper: the 6 cm carbon bed mounted in the SkyScan 1176 in vivo
scanner can be used to mount either a mouse, using the appropriate polystyrene
foam insert (lower left) or a rat (lower right).
Page 8 of 26
9
Method note: Adipose tissue quantification in vivo
The reason that we recommend supine placement of the animal is that
this reduces the transmission of breathing movement to the spine and
ribs, which is greater if the animal is placed prone (back uppermost).
2.3. Quantitative analysis of the rodent body for fat
(adipose) tissue: steps in CT-Analyser
The first step in fat analysis (as with any analysis) is how to define,
standardise and select your volume of interest.
A methodology will be presented here for selecting and analysing fat from
the thorax of the mouse, where the fat needs to be separated from lung
tissue. Depending on the scan method used, particularly whether
breathing synchronisation has been applied – and what type of
synchronisation – there is a varying amount of overlap in density between
the two tissues lung and adipose. Delineation of fat from abdominal
regions follows the same essential methodology, but is simpler in one
respect, in that no separation from lung tissue is needed. (However other
challenges can occur with abdominal fat delineation, such as air bubbles in
the gut, dense calcified dietary particles in the gut and movement
artefacts from the gut.) An overview of a methodology for adipose
delineation in the abdominal region is also given.
2.3.1. Set a standardised and referenced range of crossection
slices
The first part of the definition of the volume of interest (VOI) is setting the
range of crossectional slices. The anterior border of the VOI can be
defined relative to a landmark, such as the first appearance of the lungs,
or the first division of the two bronchi (see figure 5 i and ii respectively).
Since there is not much fat at the anterior lung, an “offset” should be
applied, so that your volume of interest begins something like 100-200
slices posterior-wards of the anterior landmark reference as defined in
Figure 5 (i) and (ii). The resulting offset crossection is shown in figure 5
(iii).
Then extend the VOI a fixed number of crossectional slices in the posterior
direction, something like 150-250 slices (figure 5 iv). The point chosen in
this way for the posterior end of the VOI range should be close to the
posterior margin of the lungs (the diaphragm). In the CTAn screenshot
the clear band in the shadow projection is the selected region and the
green regions are out-of-range. The selected VOI is shown in figure 6 as a
3D model (CTVol, Marching cubes type surface rendering).
Page 9 of 26
10
Method note: Adipose tissue quantification in vivo
(i)
(ii)
(iii)
Figure 5. The start of the lungs (i) or the branching of the trachea into bronchi
(ii) can be used as landmarks for referencing the VOI. Adding an offset in the
posterior direction (iii) commences the VOI in a region with more subcutaneous
fat. (iv) the VOI should end after a fixed number of crossections, at a point close
to the posterior margin of the lungs.
(iv)
Figure 6. The appearance of the VOI segment of the mouse thorax selected for
fat analysis relative to a lung landmark (branching of trachea), in CTVol as a set
of 3D tissue models. The adipose tissue is shown in green, lungs blue, bone gold.
Page 10 of 26
11
Method note: Adipose tissue quantification in vivo
2.3.2. Select and note tissue binary thresholds.
In CTAn, go to the binary page. (The ROI page can be skipped since the
VOI will be set using tools in the custom processing page.) Observe the
density histogram corresponding to the different tissue components. Note
the histogram is easier to interpret if set to log scale with this button
.
At this stage in the fat analysis, choose the threshold ranges
corresponding to different tissue. Density ranges for (a) fat and (b) all
tissues should be chosen. Optionally, thresholds for bone lung can also be
identified.
2.3.2.1 Fat threshold
Set an upper and lower threshold limit similar to the ones shown in figure
7 to delineate fat. Depending on how the lung was imaged, i.e. if
synchronisation was used, the lungs will overlap with the adipose tissue in
density either partially or fully.
Figure 7. A threshold density range selection (above) will segment the fat tissue
and either part or all of the lung tissue also (below) – segmented image area is
shown in red, non-segmented in green. Note that here adipose tissue density
broadly overlaps with that of the lung.
In the example shown in figure 7, the threshold limits chosen were 32-84
greyscale.
Important note: do not try to copy exactly the numerical threshold values
in this document for your analysis. With different scans and
Page 11 of 26
12
Method note: Adipose tissue quantification in vivo
reconstructions these numbers will be different, you need to find these
values yourself.
Optimal lung images without movement make lung-fat separation easier
Please note: figure 7 shows the grey level obtained for lung in vivo when
the scan is done either without breathing synchronisation or with
synchronisation [3] on a free-breathing mouse (either prospective or
retrospective) where movement suppression will be only partial. However
in the case of scanning with lung ventilation, using the Scireq FlexiventTM
for example, or else in the case of scanning a freshly sacrificed mouse
with lungs kept inflated and static [3], the scan quality will be superior so
that the reconstructed density of the lungs will be realistically closer to its
true, low value. In this case there will be a separation in density between
fat and lung, as shown in figure 8. This means that only a narrow
periphery of the lung is binarised, not the whole lung, making it easier to
remove the binarised pixels from the lung from the much thicker binarised
adipose regions.
Figure 8. A threshold density range selection (above) above for fat tissue in a
scan with movement artefacts almost fully eliminated by synchronisation with
controlled ventilation, or post-mortem scanning with lungs inflated. Here the
overlap between lung and fat is much less, restricted to the lung periphery only.
Page 12 of 26
13
Method note: Adipose tissue quantification in vivo
2.3.2.2. Threshold for all tissues
Now choose a threshold range which will binarise the whole crossection of
the mouse body (figure 9). This can simply be the lower fat threshold up
to the maximum 255 greyscale. The purpose of this is the create a volume
of interest (VOI) corresponding to the mouse body boundary. It may not
be possible to find a threshold that does not leave some white objects
outside the mouse periphery and some black pores inside it. This is OK,
these objects and pores can be removed later by image processing. In the
case shown in figure 9, the threshold range for all soft tissues is 32-255.
Figure 9. A threshold density range selection (above) above for all soft tissues.
2.3.3. Go to custom processing to set a task list to select and
analyse the adipose tissue
With threshold ranges chosen for fat and all tissues, we can now proceed
to the custom processing page – the 5th and final page of CTAn. Here we
will perform a series of thresholding and image processing steps to
delineate both a volume of interest (VOI) of the whole mouse outline and
a binarised 3d image of the adipose tissue. This will allow us the measure
fat tissue volume and thickness, as a percent of all tissues. Furthermore,
it will be shown that this method can easily and simply be adapted to
binarise and analyse the lungs also.
Tables 4 and 5 illustrate the workflow of the task list on image and ROI.
Page 13 of 26
14
Method note: Adipose tissue quantification in vivo
1. Binarise all soft tissues.
Use the threshold previously found for all soft tissue, i.e. from figure 9,
32-255.
3.2. Clean the binary mask of the animal outline.
Use the despeckle plug-in, to remove both white dots in the surrounding
space around the animal, and to remove black spaces within the animal
outline. The two operations SWEEP and REMOVE PORES can be used in
two runs of despeckle, 3D and 2D respectively and “apply to image”.
Page 14 of 26
15
Method note: Adipose tissue quantification in vivo
3.3. Copy this mask of the animal shape to the ROI
Run Bitwise operations, select the COPY operator and apply the
expression:
ROI = COPY IMAGE
This will make the ROI become a copy of the binary mask of the animal
shape.
3.4. Reload the image
Run the RELOAD plugin, and select IMAGE:
3.5. Binarise the reloaded image with the adipose threshold range
Run the threshold plugin a second time. Select global thresholding, and
this time set the threshold range appropriate for adipose tissue (see figure
7 above, in this case 32-84).
Page 15 of 26
16
Method note: Adipose tissue quantification in vivo
A binary image will be obtained including both adipose tissue and varying
amounts of lung tissue. The image will also contain noise dots.
3.6. Erode two pixels from the ROI boundary
Run the plugin MORPHOLOGICAL OPERATIONS and select ERODE. Choose
2D space, a round kernel and a radius of 3 pixels (apply to ROI).
3.7. Purge the image of out-of-ROI parts.
Run Bitwise operations and choose the AND operator. Run the expression
IMAGE = IMAGE AND ROI:
This “purges” the image by removing the part of the image outside the
ROI. (It simplifies subsequent image processing.)
3.8. Clean up the image with opening and despeckle
To improve the quality of the binarised image of fat and lung, run first the
open-close morphological operations to remove any remaining thin
peripheral artefact. Then despeckle to remove small noise objects, both
white and black objects less than (say) 100 voxels in size.
Firstly, run morphological operations and OPENING, in 3D space, with a 1
pixel radius, round kernel (apply to image). Then repeat the same
morphological operation but with CLOSING selected:
Page 16 of 26
17
Method note: Adipose tissue quantification in vivo
Secondly, run DESPECKLE, remove white pixels in 3D space, <100 voxels,
and apply to image. Then run the same operation but to remove black
voxels <100 voxels:
Page 17 of 26
18
Method note: Adipose tissue quantification in vivo
3.9. Remove the lungs (and leave the fat)
Finally, the lungs can be removed from the image (but leaving the adipose
tissue in place). This is done with the DESPECKLE operation selecting
SWEEP, in 3D space, with “remove largest object”, applied to image:
Note: alternatively at this stage, the opposite operation could be
performed, selecting “all except the largest object”. This would have the
effect of leaving the lungs in place but removing the adipose tissue
regions. Therefore this method and task list is applicable both to the study
of adipose tissue and lungs.
Page 18 of 26
19
Method note: Adipose tissue quantification in vivo
Please note also – that, while in a “normal” mouse in which the lungs are
mostly binarised together with the fat, the lungs will constitute the largest
object. However there will be two exceptions to this. One is where
synchronisation or scanning of post-mortem inflated lungs means that the
lungs are reconstructed with true density of their air content without any
movement artefact, the majority of the lung content will be of low enough
density to separate from the fat, so that only a thin periphery of lung will
be binarised (see figure 8). The other exception is where the mouse is
obese with a very large fat tissue volume. In both these cases, the volume
of binarised fat will be larger, not smaller, than the binarised lung volume.
In this case, simply reverse the logic of the SWEEP operation and remove
all but the largest, not the largest object.
3.10. Measure the volume and other 3D parameters of the lungs
Now run 3D analysis, and choose all parameters except structure
separation – also called “trabecular separation”. (To change between bone
and non-bone parameter names, go to file / preferences and the “general”
tab, and select under “nomenclature” either “Bone ASBMR” or “General
Scientific”).
Structure separation measures the separation between binarised (white)
structures, within the VOI, or to put it another way, it measured the
thickness of the black spaces within the VOI. In this case this
measurement is unnecessary and can take some time; so it can be
deselected.
Note that for analysis of multiple animal scan datasets, you can select
under “save results as” the “both” option so that, as well as the standard
results report, a summary table will record all analysed dataset results
line-by-line for easy statistical analysis in a spreadsheet program.
Page 19 of 26
20
Method note: Adipose tissue quantification in vivo
3.11. Finally, save binarised image datasets of the image and VOI
The plug-in “save bitmaps” is a useful way to allow checking of the
outcome of the task list after the analysis. Both the image and the ROI
(saved as a binary image mask) can be inspected in a “quality control” so
that you can see exactly which images were analysed and what VOI
applied.
Run the “save bitmaps” plugin twice, once to save the image inside ROI
and the second time to save the ROI. In each case the format should be
BMP, and the first three tick boxes selected only, for “convert to
monochrome” (for smaller image file size), “copy shadow projection” and
“copy dataset log file”.
Page 20 of 26
21
Method note: Adipose tissue quantification in vivo
Page 21 of 26
22
Method note: Adipose tissue quantification in vivo
Table 4. The fat analysis steps in custom processing, applied to the mouse
thorax. In SkyScan CTAn in the custom processing page, you work with several
image “channels”, that include “IMAGE” and “ROI”. At any given time the IMAGE
and ROI are represented by different images, either greyscale or binary. These
parallel images are shown at every stage of the fat analysis process.
Operation
IMAGE
1. Load dataset into custom processing.
2. Binarise the body selecting a threshold to
include all soft tissue.
There may be black pores in the tissue
region and white objects outside it.
3. Despeckle to remove white dots:
- SWEEP all except largest object, 3d
Despeckle to remove black pores:
- Despeckle: remove pores, 2D
4. Copy the resultant binary mask to
become the ROI
Bitwise: ROI = COPY IMAGE
5. Re-load the grey-scale image while
keeping the ROI mask of the animal
outline.
Reload: IMAGE
6. Run the threshold plugin to binarise the
image in the fat density range, which will
also include the lung, either boundary
only or the full volume of the lung.
Note that an artificial white line
boundary will appear around the animal.
Page 22 of 26
Result
ROI
23
Method note: Adipose tissue quantification in vivo
7. Clean-up operations:
Erode the ROI boundary by 2 pixels, purge
the image (remove image outside ROI):
- Morphological operations, ERODE, 2D,
round kernel, 2 pixels, apply to ROI
- Bitwise operations: AND
IMAGE=IMAGE AND ROI
8. Clean up the binary image of fat with:
- Morphological operations, opening, 3D,
round kernel, 1 voxel, to IMAGE
- Morphological operations, closing, 3D,
round kernel, 1 voxel, to IMAGE
- Despeckle: Remove white speckles < 100
voxels, 3D
9. Now remove the lungs. In most cases
except very obese mice, the lungs
together will constitute a larger binarised
object than the various fat bodies.
- Despeckle, SWEEP, 3D, remove the
largest object
10. Now run 3D analysis
Also “save bitmaps”, image and ROI
3D analysis result including fat volume, fat
volume as percent of all tissue volume, and
fat thickness.
NOTE: alternatively, one can run the last
SWEEP operation in the reverse way,
removing everything except the largest
object. This will select the lungs and
remove all adipose tissue regions.
Thus the task lists for selection of both lungs
and fat are very similar, differing in only
this final step.
11. Run “save bitmaps” to output a copied
directory of the full set of binarised
image crossections and ROI crossection
masks.
- Save Bitmaps, image inside ROI
- Save bitmaps, ROI
Page 23 of 26
24
Method note: Adipose tissue quantification in vivo
Table 5. The fat analysis process for the abdominal region, without the need to
separate fat from lung.
Operation
IMAGE
1.
Result
ROI
Start (no operation)
2. Binarise the body selecting a threshold
to include all soft tissue.
There may be black pores in the tissue
region and white objects outside it.
3. Despeckle to remove white and black
dots:
- SWEEP all except largest object, 3d
- Despeckle: remove pores, 2D
4. Copy the resultant binary mask to
become the ROI
Bitwise: ROI = COPY IMAGE
5. Re-load the grey-scale image while
keeping the ROI mask of the animal
outline.
Reload: IMAGE
6. Binarise with thresholds to select fat,
and exclude denser lean tissue and bone.
Threshold plugin, global
7. Clean up the binary image of fat with:
Morphological operations,
opening, 3D, 1 voxel, to IMAGE
Despeckle:
Remove white speckles < 100 voxels
8. Now run 3D analysis
Also “save bitmaps”, image and ROI
3D analysis result including fat volume, fat
volume as percent of all tissue volume, and
fat thickness.
Page 24 of 26
25
Method note: Adipose tissue quantification in vivo
3. Conclusions
This set of operations shows the high degree of flexibility that exists in
SkyScan CTAn for defining both the image and the volume of interest for
analysis, starting with binarised images with different thresholds, and
processing these images with a set of standard image processing steps
(e.g. despeckle, morphological operations etc.) and logical operations
(copy, and, not, sub etc.).
In this way, any material phases which are visibly resolved from each
other within a microCT image dataset in terms of density, can be
separated from each other for analysis. Analysis can also be done with
flexible and powerful control of the volume of interest.
Figure 10. Adipose tissue – shown in green in the lower surface rendered model
image, can be segmented and visualised on the basis of its lower x-ray density
than other “lean” soft tissues. Other tissues which are visualised on the basis of
contrast agent, such as liver, spleen, blood vessels etc., can also be segmented
using very similar image processing steps.
Page 25 of 26
26
Method note: Adipose tissue quantification in vivo
References
1. Evans RD, The Atomic Nucleus. McGraw-Hill, 1955.
2. Johns HE, Cunningham JR, The Physics of Radiology, 4th Ed. Charles C
Thomas, Illinois, USA, ISBN 0-398-04669-7; 1983.
3. De Langhe E, Vande Velde G, Hostens J, Himmelreich U, Nemery B, et
al. (2012) Quantification of Lung Fibrosis and Emphysema in Mice Using
Automated Micro-Computed Tomography. PLoS ONE 7(8): e43123.
doi:10.1371/journal.pone.0043123
Page 26 of 26