TISSUE ENGINEERING: Part C
Volume 17, Number 4, 2011
ª Mary Ann Liebert, Inc.
DOI: 10.1089/ten.tec.2010.0417
Standardized Three-Dimensional Volumes of Interest
with Adapted Surfaces for More Precise Subchondral
Bone Analyses by Micro-Computed Tomography
Catherine Marchand, Ph.D.,1 Hongmei Chen, Ph.D.,1 Michael D. Buschmann, Ph.D.,1–3
and Caroline D. Hoemann, Ph.D.1–3
Micro-computed tomography can be used to analyze subchondral bone features below treated cartilage defects
in animal models. However, standardized methods for generating precise three-dimensional (3D) volumes of
interest (VOI) below curved articular surfaces are lacking. The aims of this study were to develop standardized
3D VOI models adapted to the curved articular surface, and to characterize the subchondral bone specifically
below a cartilage defect zone in intact and defect femoral trochlea. Skeletally mature rabbit distal femurs (N ¼ 8
intact; N ¼ 6 with acute debrided and microdrilled trochlear defects) were scanned by micro-computed tomography. Bone below the defect zone (3.5 mm width, 3.6 mm length, 1 mm deep) was quantified using simple
geometric rectangular VOIs, and an optimized 3D VOI model with an adapted surface curvature, the Rectangle
with Adapted Surface (RAS) model. In addition, a 250-mm-thick Curved-RAS model analyzed bone at three
discrete subchondral levels. Simple geometric VOIs failed to analyze *17% of the tissue volume, mainly near the
top of the curved trochlear ridges. The RAS models revealed that after debridement and drilling, only 31% of the
original bone remained within the VOI and bone loss was mainly accounted for by surgical debridement.
Adapted surface VOIs are better than simple geometric VOI shapes for quantifying structural features of subchondral bone below a curved articular surface. Structural differences between the bone plate and cancellous
bone were best captured using the smaller, depth-dependent Curved-RAS model.
Introduction
M
icro-computed tomography (micro-CT) is increasingly being used in animal research models to observe
and quantify three-dimensional (3D) bone structures in bone
disease, surgery, and repair.1,2 Our laboratory is interested in
studying subchondral bone repair, as it is an important part
of the cartilage repair process,3 in cartilage repair strategies
involving subchondral bone damage. To this end, we previously developed a surgical model of marrow stimulation in
rabbit, where rectangular full-thickness cartilage defects are
created in the trochlea of the femur by debridement, and the
subchondral bone is pierced with multiple 1-mm-diameter
microdrill holes4,5 or microfracture holes.6
Once a given bone sample is scanned by micro-CT and a
3D reconstruction model is obtained, a 3D volume of interest
(VOI) must be selected to quantify bone parameters within a
specific region. The 3D VOI is obtained by interpolating the
volume between two or more two-dimensional (2D) regions
of interest (ROI) placed within the image dataset. Inter-
polation between two circular ROIs will generate, for example, a cylindrical VOI. As the trochlea has a concave
surface, subchondral bone repair below a trochlear defect
cannot be fully represented by simple 2D and 3D geometric
shapes with flat top surfaces. Further, a distinct VOI should
ideally be generated for different regions that is dependent
on the anatomical characteristics defining those regions.7 In
the case of femoral end, it is well known that below the
calcified cartilage lies a dense subchondral bone plate that is
connected to a more porous subchondral trabecular bone
with different morphologic and mechanical bone properties.8
To our knowledge, few animal studies involving marrow
stimulation have applied micro-CT to analyze subchondral
bone repair; furthermore, the development of VOIs for
curved articular mineralized surfaces has not been carefully
addressed.
Previous micro-CT studies analyzing rabbit subchondral
bone have used VOIs with simple geometric shapes, including cylinders9–11 and cubes.12 However, this approach11
can be limited since bone measurements either include void
1
Institute of Biomedical Engineering, École Polytechnique de Montreal, Québec, Canada.
Department of Chemical Engineering, École Polytechnique de Montreal, Québec, Canada.
Groupe de Recherche en Sciences et Technologies Biomédicales, École Polytechnique de Montréal, Québec, Canada.
2
3
475
476
MARCHAND ET AL.
volume above the trochlea or omit curved structures above
a flat VOI surface. Alternatively, hand-drawn polygonal
shapes have been used to delimit irregular structures such as
osteophytes,10 and to separately analyze trabecular and
cortical bone in the diaphysis.13 Subregional VOIs have been
used to evaluate bone mineral density (BMD) within vertebral subregions, instead of relying on a gross estimate made
on the entire vertebrae.7 Although anatomical landmarks or
reference points are necessary to ensure that VOIs are similarly positioned in different samples,10,14 this practice is
rarely described. In addition, to guarantee precision, creation
of VOIs should have defined sizes, shapes, and positions,
although current literature lacks appropriate information
describing the methods to achieve this, with corresponding
rationale, standardized terminology, and the minimal set of
outcome variables as recently described by Bouxsein et al.15
The aim of the present study was to develop robust subchondral customized VOIs with a standardized procedure,
that takes into account the curvature of the articulating
surfaces to quantify bone underlying the defect area in our
rabbit cartilage repair model.16
Materials and Methods
In vivo articular bone marrow stimulation model
and sample fixation
All animal studies were carried out with protocols approved by the University of Montreal animal division. Intact
(native) femur ends (N ¼ 8) from four skeletally mature
rabbits were dissected of all soft tissue and fixed (Table 1;
Fig. 1C, D, G). Bilateral 3.54.5 mm full-thickness articular
cartilage defects debrided into the calcified layer were created
in three skeletally mature New Zealand White rabbits, pierced
with microdrill holes (two distal 0.9 mm diameter and six
proximal 0.5 mm diameter holes), and treated with thrombin
and chitosan-glycerol phosphate/blood implant or thrombinonly as previously described16 (Fig. 1A, B, E, F, H). Rabbits
were euthanized after 1 day (*24 h) under anesthesia and the
femurs (N ¼ 6) dissected and fixed per Table 1. Distal femurs
were trimmed of the diaphysis using an IsoMet Low Speed
Saw Precision (Buehler, Markham, Canada), and the surgical
defect samples were further trimmed of the condyles.
Micro-CT scans
Femoral ends were moistened with several drops of
phosphate-buffered saline and wrapped in transparent
plastic wrap. To reduce metal interference from the metal
stage, a plastic cap from a 50 mL conical tube was placed
between the sample and the stage, where it was tightly secured with parafilm strips in a vertical or horizontal position
(see position, Table 1) and scanned (Skyscan X-ray Microtomography 1172, Kontich, Belgium). Images were acquired
at 80 kV, 100 mA with an Al þ Cu filter, and set to an image
size of 10241280 with a pixel size resolution of 10.09 mm.
For each specimen, a series of 419 projection images were
taken using an exposure time of 2000 ms with a rotation step
of 0.458 between each image and two frames were averaged.
Each scan lasted about 38 min. Several intact femur samples
were re-scanned one to three times during protocol development, or to reproduce scans free of motion artifacts. Final
post-alignment values were within a range of 13.5 and þ6.0
(no units).
Reconstruction, repositioning of the data set,
and BMD calculation method
Sample reconstruction was carried out with a Dell Alien
computer equipped with a GeForce GTX295 1.8GB video
card or ProSys computer with ATI Radeon HD 4670 512MB
video card, using NRecon software (version 1.6.2.0; Skyscan) with identical parameter setup (see Table 2). All data
were analyzed without resizing or condensing. The sample
images within the data sets were repositioned using DataViewer software (version 1.4.3; Skyscan) such that the axial
(equivalent to transverse) view was projected in the image
stack. The guidelines in all three orthogonal planes (axial,
sagittal, and coronal) at the surgical site in the trochlea (Fig. 2)
were made perpendicular to the bone surface (BS), and
the repositioned sample images saved as a new data set. All
quantitative 3D bone analyses were carried out using
CTAnalyser software (version 1.10.1.1; Skyscan), where
thresholding levels of gray values (95–255) were set to be
the same for all samples. The lower threshold value chosen
for the binary image was determined to be that which
mimicked bone as closely as possible compared to the raw
image. Our 3D bone analysis included parameters described in Table 3. To obtain true BS values within a given
subchondral VOI, we subtracted a value termed ‘‘intersection surface.’’ (IS). IS is calculated by CTAnalyser software,
and represents the bone interfacial surface of the VOI that
runs through the bone and is therefore not true trabecular
BS. 3D models were created using CTVol (version, 2.2.0.0;
Skyscan).
Table 1. Samples Analyzed in This Study
Rabbit
group
Intact
knee
Defect
knee
Animal
age
Animals
per group
Repair
period
Fixation solution
and storage time
9 months
Two females
—
80% ethanol, 48C, 110 days
12 months
Two females
—
8 months
Three males
1 day
(*24 h)
4% paraformaldehyde, 1% glutaraldehyde,
0.1 M sodium cacodylate (pH 7.3,
at room temperature overnight,
and then at 48C), 23 days
4% paraformaldehyde, 0.1 M sodium
cacodylate (pH 7.3, 48C), 7 days
Sample position
during the scan
Vertical:
facing
Vertical:
facing
Condyles
up
Condyles
up
Horizontal: Trochlea
facing up
ADAPTED BONE VOIS FOR CURVED SURFACES
477
FIG. 1. Defect created in trochlea
debrided into the calcified cartilage layer, with six proximal drill
holes of 0.5 mm and two distal
holes of 0.9 mm diameter. Panel
(A) shows a schematic of the
trochlear defect, and (B) a 3D
micro-computed tomography
model of the trochlear defect 1 day
postoperative. Safranin-O stained
histology section and corresponding 2D micro-computed tomography view are shown from of an
intact trochlea (C, D, G), and defect trochlea through the two 0.9mm distal holes (E, F, H) showing
complete removal of the calcified
layer. In (H) the implant is adhering to the bone surfaces (blue
arrows); the tear in the implant is
most probably due to a sectioning
artifact. Symbols are as follows: þ,
drill holes; AC, articular cartilage;
CC, calcified cartilage; bc, blood
clot; 3D, three-dimensional; 2D,
two-dimensional.
BMD values were obtained using the attenuation coefficient method where attenuation coefficient values of
two phantoms (8-mm-diameter hydroxyapatite cylinders
0.25 and 0.75 g/cm3; Skyscan), scanned and reconstructed
using identical parameters as the samples, were used for
calibration.
Four-step procedure to generate a precisely adapted
VOI in curved trochlear samples from intact femurs
Step 1: Re-orient the sample 3D image stack. This
method was developed to generate an standardized VOI
model that faithfully encompassed the whole curved trochlear
FIG. 2. Correct orientation of a
sample image stack. The three orthogonal views (i–iii) of rabbit knee
trochlea with optimal orientation are
shown for a defect trochlea (A, B) and
intact trochlea (C) and incorrect
position for intact trochlea (D). White
arrows (A–C) show the femoral
epiphyseal growth plate scar. The
dotted white box in Ciii shows the
approximate surgical defect site in the
intact trochlea. White bars in D (i–iii)
show that the trochlear surface is not
perpendicularly oriented in the three
orthogonal planes.
478
MARCHAND ET AL.
Table 2. Primary Reconstruction Parameters Used
Reconstruction parameter
Setting
Smoothing
Ring artifact correction
Beam hardening correction (%)
Object bigger than field of view (ON/OFF)
Minimum for cross-section to image conversion
Maximum for cross-section to image conversion
2
10
40
ON
0.0000
0.0400
defect area, as well as in matching areas of intact knees, specifically below a 3.5-mm-wide and 3.6-mm-long cartilage defect region. However, before generating the VOI measurement,
it was essential to re-orient the 3D image stack so that the three
orthogonal planes were perpendicular to the trochlear surface
(Fig. 2A–C vs. D). The image stack was oriented to toggle
through the axial (transverse) plane (Fig. 2A–Cii). Re-orienting
the 3D image stack permitted accurate subchondral bone plate
thickness measurements at 908 from the surface and reduced
the number of 2D ROIs needed to generate an appropriately
curved VOI model along the sagittal axis (proximo-distal axis)
(Fig. 2A–Ciii), as in step 3 below.
Step 2: Determine the shape and size of the 2D ROI in the
X-Y (axial) plane and adapt the surface. A standardized
simple geometric ROI consisting in a rectangle 3.5 mm wide
and 1 mm deep in the axial plane was created to cover the
bone region below the trochlear cartilage defect (Fig. 3A–D),
which was 3.5 mm wide and 4.5 mm long, with *3.5-mmdeep microdrill holes.4,5,16
To adapt the top surface, the ‘‘node function’’ (CTAnalyser
software) was used to insert intermovable nodes that were
placed along the trochlear surface (Fig. 3F). We named this
model the Rectangle with Adapted Surface (RAS) (Fig. 3E–H).
The RAS model was made to be only 1.0 mm deep to ensure
that the resulting VOI always remained above the femoral
epiphyseal growth plate bone scar and fatty marrow bone
void (see white arrows in Fig. 2A–Cii, Ciii) that, if included,
could give biased bone volume (BV) measurements (e.g., in
some animals the femoral growth plate scar was *1.3 mm
below the surface in the area of the defect).
The Curved RAS (C-RAS) is an adaptation of the RAS
model, but used for only analyzing specific 250-mm-thick bone
subregions. The ROI was also 3.5 mm wide, like the RAS
model, but only 250 mm thick instead of 1 mm, with edges
perpendicular to the surface of the trochlea (Fig. 3I–L). For the
C-RAS model, the same numbers of nodes were added to the
top and bottom surfaces, and they were fitted to the trochlear
surface while maintaining a standardized ROI (3.5 mm wide
250 mm thick) (Fig. 3J). The C-RAS was first created at the
surface to evaluate bone within the subchondral bone plate,
and then it was perpendicularly translated 500 and 1000 mm
below the mineralized surface to evaluate bone within the
subchondral trabecular bone region (Fig. 3, lower panel). Note
that, in this model, the 250 mm thickness was chosen as a dimension smaller than the measured average rabbit subchondral bone plate thickness (unpublished data—distal:
403 109 mm and proximal: 341 117 mm, N ¼ 14, measurements taken at the lowest point of the trochlear curvature [also
our reference point] outside the surgery site) and similar to
those found by others using rabbit trochlea histological sec-
tions which omitted calcified cartilage (CC).17 The subchondral
bone plate thickness measured in our study includes CC, as we
were not able to distinguish it from the subchondral bone
plate, even when using a range of thresholds in the reconstructed data sets (Fig. 1C, D, G).
Step 3: Paste 2D ROIs through the defect at the reference
point, to obtain a 3D VOI following the curvature of both the
top surface and the proximo-distal axis. The trochlear surface has two curvatures: one that follows the trochlear
groove (lateral-medial axis) and another that follows the
natural slope of the trochlea along the proximo-distal axis.
Further, the trochlear surface curvature is slightly different in
proximal and distal ends. Therefore, for each sample analyzed, two 2D ROIs were created, one at the proximal end
and one at the distal end. The distance between these ROIs
along the proximo-distal axis (i.e., the proximo-distal VOI
edges) was always 3.6 mm, so as to encompass all the drill
holes and avoid inclusion of bone outside the defect area. For
intact femurs, the VOI proximo-distal boundaries were selected by simultaneously viewing all three orthogonal
planes, and approximately covered the mid-trochlear area
above the femoral growth plate scar (Fig. 2Ciii).
Then, the top of these 2D ROIs was fixed to a reference
point that was set to be the center of the trochlea and the
lowest point of the trochlear curvature (see yellow triangle,
Fig. 3A, B, E, F, I, J). To faithfully cover the defect region
along the proximo-distal curvature of the trochlea, the
proximal and distal 2D ROIs were re-pasted from their
starting end continuing until the middle of the defect,
keeping the top aligned with the reference point. In our
study, depending on the proximo-distal curvature of the
trochlea, the 2D ROI was re-pasted within the VOI from 8 to
14 times. On intact trochlea, nodes at the curved surface were
also adjusted as necessary to fit the slight change in curvature to fit with trochlear ridges.
Step 4: Interpolate all ROIs and verify that the VOI-adapted
surface is congruent with the BS. Once the standardized 2D
ROI was placed at the reference point, adapted (using nodes)
to the surface, and positioned so as to ensure that the 3D VOI
fell precisely inside the correct defect area, we then verified
that the adapted surface of the new interpolated 3D VOI fit
precisely with the trochlear surface by scrolling through the
entire data set with interpolated ROIs (interpolation function
in CTAnalyser). If needed, re-adjustments of specific adapted
nodes can be made, as this method is a semi-automated sliceby-slice hand-contouring approach. Then, all the interpolated 2D ROIs forming a 3D VOI were analyzed using
CTAnalyser software and virtual solid 3D models were obtained (Fig. 4). These adapted surface VOIs can be easily
modified for use in other cartilage or bone repair models.
Fulfilling each requirement summarized in Table 4 would
lead to an appropriate surface-adapted 3D bone analysis.
Modified procedure to generate an adapted 3D VOI in
curved trochlear samples with a surgical defect. In femur
samples containing a debrided and drilled defect, the subchondral plate had been removed by debridement (Fig. 3D,
H, L), consequently removing the location of the reference
point indicated by a yellow triangle in Figure 3A, B, E, F, I,
and J. Therefore, to generate a VOI in the defect area, two
(NS)
(NS)
(NS)
(NS)
(NS)
(NS)
(NS)
(NS)
;
:
:
:
:
;
;
:
:
;
:
:
:
0.0398
0.0446
0.1359
0.0901
0.0661
0.1091
0.0685
0.0820
0.0017
0.1303
0.0397
0.0398
0.3432
(10.64)
(4.03)
(0.84)
(2.92)
(0.41)
(0.02)
(0.39)
(0.02)
(52.9)
(0.03)
(10.65)
(10.64)
(36.59)
51.34
17.44
8.61
2.12
1.09
0.15
3.50
0.17
154.0
0.05
48.63
48.66
91.51
0.0001
0.0003
0.0002
0.0213
0.0003
0.0034
0.0162
<0.0001
<0.0001
0.0486
0.0001
0.0001
0.0031
(15.1)
(4.67)
(1.15)
(6.92)
(1.7)
(0.02)
(0.71)
(0.03)
(11.66)
(0.26)
(15.28)
(15.10)
(19.62)
66.14
12.59
7.81
2.71
0.14
0.16
4.07
0.15
80.0
0.20
33.70
33.86
77.1
(3.90)
(2.45)
(2.10)
(5.36)a
(1.84)
(0.02)a
(0.48)a
(0.01)a
(6.25)
(0.22)
(3.91)
(3.90)
(30.8)
96.19
3.92
3.69
10.74
4.31
0.20
4.89
0.05
9.38
0.46
3.37
3.81
31.11
BV/TV (%)
(BS-IS)/BV (1/mm)b
(BS-IS)/TV (1/mm)c
Tb.Pf (1/mm)
SMI
Tb.Th (mm)
Tb.N (1/mm)
Tb.Sp (mm)
Obj.N
Po(cl) (%)
Po(op)%
Po(tot) (%)
Conn.Dn (1/mm3)
a
For trabecular pattern factor, trabecular thickness and trabecular separation values within the top Curved-Rectangle with Adapted Surface region analyzed are not representative of real
trabeculae as no trabeculae are found in the dense bone plate; therefore, the automated software measurements refer to bone structures and not trabeculae per se.
b
Note that a statistically significant difference was observed between (BS-IS)/BV and BS/BV (not shown) values (t test, p 0.001).
c
Note that a statistically significant difference was observed between (BS-IS)/TV and BS/TV (not shown) values (t test, p 0.0001)
AVG, average; STDV, standard deviation; NS, not significant.
Description
Percent bone volume
Bone surface/volume ratio
Bone surface density
Trabecular pattern factor
Structure model index
Trabecular thickness
Trabecular number
Trabecular separation
Number of objects
Closed porosity (percent)
Open porosity (percent)
Total porosity (percent)
Connectivity density
Overall tendency from top
to deep bone regions
p-Value between mid
and deep regions
Deep AVG
(STDV)
p-Value between top
and mid regions
Mid AVG
(STDV)
Topa AVG
(STDV)
Abreviation
(units)
Table 3. Three-Dimensional Bone Parameters Obtained from the 250 mM Curved-Rectangle with Adapted Surface Model Volume
of Interest at Three Different Depths (Top, Mid, Deep) in Intact Knees (N ¼ 8)
ADAPTED BONE VOIS FOR CURVED SURFACES
479
adapted-surface 2D ROIs were first created outside the defect
area in the flanking proximal and distal regions (to account
for slight proximal vs. distal differences). The two 2D ROIs
were first pasted at their starting end of the defect and both
re-pasted toward the middle to cover the whole defect area.
Each 2D ROI (RAS or C-RAS) was carefully aligned with the
top of the native flanking bone present on the edges of the
defect (Fig. 3H, L) to best approximate trochlear curvature.
In our sample, the 3D VOI was obtained after interpolation
of all of the 8 to 14 ROIs pasted in the defect area. The
resulting VOI gave rise to a virtual curved surface intended
to best mimic the original intact trochlea, and which also
excluded void volumes above the trochlea.
Statistical analysis
The Student t-test was used to analyze differences between bone parameters obtained using distinct 3D VOI
models. p-values <0.05 were considered significant.
Results
Simple rectangular geometric VOI shapes lacking
an adapted surface do not encompass all the bone
in a curved surface bone tissue
In intact femurs, subchondral bone below a 3.5-mm-wide
and 3.6-mm-long rectangular defect region was analyzed
using simple geometric VOIs (Flat and Curved 3D rectangles
[Fig. 4B, D], and a standardized 3D RAS VOI model [Fig.
4E]). Since standardization in the use of custom VOIs is
critical, we aimed to have the same tissue volume (TV) for
different samples, and for each VOI model. By following our
four-step procedure, the Flat and Curved 3D Rectangle
models generated the same TV of 12.71 mm3 (Fig. 5A). For
the customized RAS model, the same procedure was followed, except that the top surface of the 2D ROI was adapted
to fit the trochlear surface. This extra step permitted encompassing all the bone normally excluded in the Flat and
Curved 3D simple Rectangle models (as in Fig. 4A, C white
translucent structures), and led to a slightly larger TV of
15.25 mm3, a 17% increase in TV (Fig. 5A). The smaller TV for
the simple geometric shapes was paralleled by a 24% lower
BV (*8.21 mm3 vs. 10.87 mm3 for RAS, Fig. 5B). Importantly,
the Flat and Curved 3D Rectangle VOIs generated significantly lower BV/TVs ( p < 0.0001), compared to the 3D RAS
target tissue (Fig. 5C). These data showed that simple geometric shapes with flat top surfaces fitted to samples with
irregular anatomic borders were missing a significant portion of the subchondral bone plate (Fig. 4A, C), which resulted in subchondral VOIs with mis-representative bone
features.
Debridement and drilling surgical procedures
remove significant amounts of bone, mainly
within the subchondral bone plate
The 3D RAS model showed that a substantial amount of
bone was removed by debridement and drilling; only 31% of
the original subchondral bone 1 mm below the defect remained (22.3% 4.5% vs. 71.3% 10.6% BV/TV, Fig. 6A,
and 0.39 g/cm3 vs. 1.0 g/cm3 BMD Fig. 6B), but these data
did not indicate whether the bone plate was more damaged
480
MARCHAND ET AL.
FIG. 3. Two-dimensional
ROIs size, shape, and reference point. Simple geometric
ROI (A–D)), RAS (E–H), and
Curved-RAS ROI (C-RAS,
I–L). Panels B, F, and J show
the interpolation nodes (white
circles) used to adapt the ROI
top surface to the trochlear
grove and the reference point
(indicated by the yellow
triangle) used to identically
position all the ROIs. Panels
C, G, and K show the binary
image of the ROI placed in the
surgery site in the intact femur, or defect site (panels D,
H, and L). The lower panel
shows the three subchondral
bone regions analyzed with
the C-RAS model. ROI,
regions of interest; C-RAS,
curved-rectangle with adapted surface.
than the trabecular bone. Using the 3D C-RAS model with a
250 mm thickness, however, we observed that most bone removed from debridement and drilling came from removal of
the calcified cartilage and bone plate, since the percentage of
bone removed from the top, mid, and deep-bone regions was
99%, 50%, and 40%, respectively (Fig. 6C) with corresponding BMD loss (Fig. 6D).
Bone analysis at different depths below intact
knee trochlea reveals significant differences
in subchondral bone properties
Here, the 3D C-RAS model was used for specifically
analyzing smaller, 250-mm-thick bone subregions at three
levels in intact femurs. As expected, analyses revealed
FIG. 4. Four 3D subchondral VOI models (gray
structures) from the trochlea
of an intact rabbit knee, in a
region covering the cartilage
defect site. Structures A and C
(white translucent) show the
BV omitted from the VOI
created by interpolating only
two 2D rectangle ROIs
(straight blue arrows) (B, Flat
3D Rectangle) or by interpolating 8 to 14 2D rectangle
ROIs positioned at the reference point (yellow curved
line) to follow the bone surface curvature along the
proximo-distal axis (blue
curved arrows) (D, Curved
3D Rectangle). (E) Shows the
VOI of an RAS 3D model,
1 mm deep from the reference
point (yellow triangle). (F–H)
Show the 250-mm-thick CRAS 3D model, which encompasses the entire surgical
site surface area. The C-RAS
3D model was positioned at the reference point at the surface (top-bone region, F), at 500 mm (mid-bone region, G) and
1000 mm (deep-bone region, H) from reference point. The yellow curved line in (D–F) represents the interpolation of all the 2D
ROIs each positioned at the reference point through all the VOI. VOI, volumes of interest; BV, bone volume.
ADAPTED BONE VOIS FOR CURVED SURFACES
481
Table 4. Requirements Associated with Step-Wise Procedure to Generate
a Robust Subchondral Volume of Interest
Requirement
Aim
Re-orient the sample to have the trochlea
in a horizontal position
To easily and specifically adapt the ROI to curvature of the trochlear
groove surface and make accurate measurements 908 from surface
Determine VOI dimensions for standardization:
Maximum dimensions must be similar to the original defect
Size (length, width, depth) and Position
(x, y, z boundaries)
VOI must be positioned above the femoral epiphyseal growth plate
bone scar to avoid any overlapping or inclusion of fatty marrow space
Determine a reference point along the
axial plane
To ensure proper alignment and registration from one two-dimensional
ROI to another, to generate comparably oriented VOIs across all
study samples
Draw two-dimensional polygons on axial
(transverse) planes image data set
To specifically adapt the ROI to fit the curved trochlear surface and
precisely measure ROI depth 908 from surface
To avoid missing any curved subchondral bone at the surface along
the area of interest (width and length)
ROI, region of interest; VOI, volume of interest.
significant differences between the top (calcified cartilage
and dense subchondral bone plate) and mid-bone regions
(porous trabecular bone) (Table 3). Interestingly, bone parameters of the deep-bone region (1000 mm below the surface) also showed significant differences compared to the
mid-bone region (Table 3), even though both of these regions are generally considered to be within the same subchondral trabecular bone region. In general, 3D analyses
revealed a very dense bone structure at the surface, and a
trabecular structure becoming less thick and more separated as it radiated from the bone plate toward the marrow.
This was shown by a quantitative decrease in BV (BV/TV%)
and closed porosity, and an increase in BS area (BS-IS/TV) at
the deep versus mid-bone region (Fig. 6C, Table 3). A significant difference was observed between the value BS/TV and
(BS-IS)/TV ( p 0.0001), showing the importance to subtract
the IS value for accuracy of this parameter using Skyscan.
Structure model index (SMI) is a bone parameter that quantifies the plate versus the rod characteristics of trabecular bone;
typical values range from 0 (plate-like shape trabeculae) to 3
(rod-like shape trabeculae). Negative SMI values are also
possible when the surfaces become more concave, as in regions
of bone containing a prevalence of enclosed cavities with relative BV/TV% value being above 50% (Skyscan CTAn manual)18 such as high-density subchondral bone plate region
analyzed with top C-RAS model (SMI ¼ 4.31 1.84). The
deep-bone region (SMI ¼ 1.09 þ 0.41) was more plate-like in
shape as well as the mid-bone region (SMI ¼ 0.14 þ 1.7), but
in this case, SMI calculation in the mid-bone region, is again
influenced by the BV/TV% value of 66.1% (i.e., >50%). To
summarize, these C-RAS VOI data showed that the subchondral bone immediately below articular cartilage was
comprised of 96% BV/TV dense bone with low porosity and
connectivity density, which progressively increased toward
plate-like bone structure with 51% BV/TV, at 1 mm below the
surface.
The measured BMD values from 3D C-RAS model were
significantly different when comparing the bone at the top
versus mid-bone region ( p < 0.005) and top versus deepbone region ( p < 0.0001), but showed a slightly decreased
value between the bone regions analyzed at the mid versus
deep-bone region (Fig. 6D). As expected, BMD values were
correlated to BV fraction (BV/TV%) and diminished with the
deeper region analyzed (Fig. 7). Thus, subregional differences in the subchondral bone were observed where BV/
TV% and BMD diminished with increasing distance from the
trochlear surface.
Finally, the intact knee trabecular thickness (Tb.Th)
graph distribution clearly showed different profiles for the
top-bone region VOI (calcified cartilage and subchondral
bone plate) versus the mid- and deep-bone regions. The topbone region showed an eccentric curve with a mode at
242 mm versus similar and normal distributions for the two
other VOI (mid and deep) with a peak Tb.Th around 150 mm
FIG. 5. Bone measurements in subchondral trochlear defect zone using
simple geometric shape VOIs versus an
adapted surface VOI yield standardized
TVs with significantly different bone
features. TV (A), BV (B), and BV/TV% (C)
measured in intact femurs with simple 3D
geometric shapes: Flat 3D Rectangle
(white columns) and Curved 3D Rectangle (gray columns) (shapes B and D,
respectively, in Fig. 4), compared to 3D
RAS model (black columns, shape E in Fig. 4). Graphs (A–C) use the same legend. Data are shown as the average standard
deviation. Note the very small standard deviations for the TV in panel A being 0.05 mm3 for Flat and Curved 3D Rectangles
and 0.38 mm3 for 3D RAS model. TV, tissue volume; BV, bone volume, ***p < 0.0001.
482
MARCHAND ET AL.
FIG. 6. BV/TV% and BMD values of
intact knees (N ¼ 8) and defect knees
(N ¼ 6) were quantified using 3D RAS
model (A, B), and 3D 250 mm C-RAS
model at three depth levels below the
trochlear surface (C, D). Data are shown
as the average standard deviation.
*p < 0.05, **p < 0.005, ***p < 0.0001,
#
p < 0.00005. In panels C and D, note
that for each matching C-RAS level, intact and defect VOIs showed significant
differences p < 0.001. BMD, bone
mineral density.
(Fig. 8A). The trabecular separation (Tb.Sp) graph distribution also showed for the top-bone region VOI an eccentric curve for the top VOI (40 mm) versus a broad
nonspecific range for the mid- and deep-bone regions (Fig.
8B). Note that thickness and separation distribution profiles
for the top C-RAS VOI refer to the subchondral bone plate
region that includes the calcified layer and contains no
trabeculae.
Discussion
FIG. 7. BV/TV% versus BMD obtained from the C-RAS
model at different depths (top, mid, deep) below intact
trochlear surface. Data are represented as the average value
of N ¼ 8 samples.
In this study, we report on the use of customized VOIs to
more precisely analyze anatomical sites that have irregular
borders compared with two standardized but simple geometric shapes (Flat and Curved Rectangles). Our analysis
showed that VOIs with a flat top surface generated significant errors and thus emphasizes the importance of customizing the VOIs when analyzing sites that have irregular
anatomic borders. Further, structural differences were best
captured using the smaller, depth-dependent adapted VOIs
(C-RAS).
When applied to intact and surgically treated rabbit femurs, the RAS model was useful in determining the quantity
of bone removed by debridement and drilling up to 1 mm
below the surgical site, but generated average values for
clearly distinct bone subregions. To address this limitation
we developed the C-RAS model, which gave discrete 3D
bone values within precise subchondral bone regions. The
specific depths chosen (top, mid, and deep) were motivated
by our observation that depth-dependent differences in re-
ADAPTED BONE VOIS FOR CURVED SURFACES
483
FIG. 8. Averaged distribution
curves for trabecular thickness
(mm) (A) and trabecular separation
(mm) (B) in intact knee trochlea
with VOI positioned at surface (top,
thick black line), VOI 500 mm below
surface (mid, plain black line), and
VOI 1000 mm below surface (deep,
dotted black line). Note that the top
C-RAS VOI includes calcified cartilage layer and does not contain
trabeculae, and the automated software measurements refer to bone
structures, which resulted in atypical trabecular thickness and separation distributions.
pair are seen at these levels after 6.5 months of repair in a
similar animal model (Marchand et al., Trans ORS 2010,
manuscript in preparation). Interestingly, in femur trochlea
with surgical defects, the very low BV/TV% value obtained
from the top 3D C-RAS level clearly shows the complete
removal of the calcified cartilage layer from surgical debridement (see Fig. 1E, F, H), which is a recommended
aim of marrow stimulation procedures since it can lead to
more regenerated cartilage tissue19 and better integration of
marrow-derived cartilage repair tissue.4,20
Comparisons of measured parameters between different
rabbit bone micro-CT studies must be made carefully; in
general, the differences in rabbit femoral bone parameters
measured between our study and others can be explained
by differences in the region encompassed by the VOI, the
depths chosen, the position at which the VOI was placed,
and the rabbit age. Further, the chosen cross-section to
image conversion min/max values (values determined at
the reconstruction step, see Table 2) and thresholding values (determined after reconstruction) may differ within
most micro-CT studies, due to the subjective nature of their
determination, and this has to be kept in mind when
making comparisons. It has been shown that a variation of
0.5% in thresholds resulted in a 5% difference in BV/TV, for
regions of low-density bone.21 Nonetheless, some comparisons can be made, although using value ranges. The BMD
previously measured in mature male rabbit femur condyles
was also observed to decrease with increasing distance
from the articular surface,22 in agreement with our C-RAS
trochlear measures. Moreover, an in vivo micro-CT rabbit
study using a cube VOI12 obtained bone parameters of BV/
TV% (40.8%–50.1%), Tb.Th (158–186 mm), and Tb.Sp (179–
233 mm), which are comparable with our 1-mm-deep 3D
C-RAS model range values, BV/TV% (36%–64%), Tb.Th
(123–177 mm), and Tb.Sp (149–201 mm). In addition to VOI
size and position being different, the use of only 2- and
6-month-old female rabbits by others may partly explain their
higher measured values. Use of older rabbits should be preferred, because although 6-month-old rabbits are considered skeletally mature, physes closure occurs typically after
8 months.23
Our interest in developing adapted surface models was to
ensure that analyses were not biased by including voids
above the trochlea, or missing bone regions, or including
other bone structures (growth plate scar). The 3D C-RAS
model follows the natural bone curvature, thereby providing
very specific subchondral bone structural characteristics. 3D
RAS and C-RAS models’ thickness can easily be modified to
address other analytical aims; provided the new 3D model
answers each requirement determined in our study (Table 4),
it will lead to an appropriate adapted surface 3D bone
analysis. These two 3D models can also be applied to different types of defects with curved BSs. For example, a
common defect used to study large osteochondral lesions in
rabbits is the creation of a large, drilled, single cylindrical
hole through the condyle or the trochlear surface.11,24,25 To
analyze bone within this defect model, the 3D VOI could
then be a Cylinder with Adapted Surface to encompass the
entire defect or a Curved-Cylinder with Adapted Surface to
cover specific bone regions. Our interest in developing the
3D RAS and C-RAS-adapted surface models was to also
better understand the bone repair processes and for evaluating bone in repair tissue. Future work will apply the
models to evaluate mineralized repair tissue of repaired
microdrilled defects.
Conclusions
Micro-CT is a powerful tool for bone structure characterization, but the VOIs chosen for the analysis must be carefully
selected and standardized. Depending on the study objectives, VOIs can be designed to be all-encompassing with respect to the defect, or to cover anatomically specific regions.
Adapted surface 3D models are superior to simple geometric
VOI shapes for quantifying curved subchondral bone features, as they avoid including or excluding regions leading to
bias and follow natural curved surfaces. In this study the 3D
C-RAS model was better at capturing subchondral bone
structural differences since the VOI fits within a specific bone
region and has depth-dependent adapted VOIs. The adapted
surface VOIs developed here could be easily modified for use
in other cartilage or bone repair models.
Acknowledgments
We thank Jun Sun, Geneviève Picard, Gaoping Chen, and
Corine Martineau for valuable technical support. Financial
support: Canadian Institutes of Health Research (Operating
Grant no. 185810-BME), Canadian Arthritis Network, Fonds
de la Recherche en Santé du Quebec for salary support (C.M.
and C.D.H.).
484
MARCHAND ET AL.
Disclosure Statement
No competing financial interests exist.
15.
References
1. Holdsworth, D.W., and Thornton, M.M. Micro-CT in
small animal and specimen imaging. Trends Biotechnol 20,
S34, 2002.
2. Jiang, Y., Zhao, J., White, D.L., and Genant, H.K. Micro CT
and Micro MR imaging of 3D architecture of animal skeleton. J Musculoskelet Neuronal Interact 1, 45, 2000.
3. Hoemann, C.D., Chen, G., Marchand, C., Tran-Khanh, N.,
Thibault, M., Chevrier, A., et al. Scaffold-guided subchondral
bone repair: implication of neutrophils and alternatively
activated arginase-1þ macrophages. Am J Sports Med 38,
1845, 2010.
4. Hoemann, C.D., Sun, J., McKee, M.D., Chevrier, A., Rossomacha, E., Rivard, G.E., et al. Chitosan-glycerol phosphate/
blood implants elicit hyaline cartilage repair integrated with
porous subchondral bone in microdrilled rabbit defects.
Osteoarthritis Cartilage 15, 78, 2007.
5. Chevrier, A., Hoemann, C.D., Sun, J., and Buschmann, M.D.
Chitosan-glycerol phosphate/blood implants increase cell
recruitment, transient vascularization and subchondral bone
remodeling in drilled cartilage defects. Osteoarthritis Cartilage 15, 316, 2007.
6. Chen, H., Sun, J., Hoemann, C.D., Lascau-Coman, V.,
Ouyang, W., McKee, M.D., et al. Drilling and microfracture
lead to different bone structure and necrosis during bonemarrow stimulation for cartilage repair. J Orthop Res 27,
1432, 2009.
7. Briggs, A.M., Perilli, E., Parkinson, I.H., Wrigley, T.V., Fazzalari, N.L., Kantor, S., et al. Novel assessment of subregional bone mineral density using DXA and pQCT and
subregional microarchitecture using micro-CT in whole human vertebrae: applications, methods, and correspondence
between technologies. J Clin Densitom 13, 161, 2010.
8. Burr, D.B. Anatomy and physiology of the mineralized tissues: role in the pathogenesis of osteoarthrosis. Osteoarthritis Cartilage 12 Suppl A, S20, 2004.
9. Batiste, D.L., Kirkley, A., Laverty, S., Thain, L.M., Spouge,
A.R., and Holdsworth, D.W. Ex vivo characterization of
articular cartilage and bone lesions in a rabbit ACL transection model of osteoarthritis using MRI and micro-CT.
Osteoarthritis Cartilage 12, 986, 2004.
10. Batiste, D.L., Kirkley, A., Laverty, S., Thain, L.M., Spouge,
A.R., Gati, J.S., et al. High-resolution MRI and micro-CT in an
ex vivo rabbit anterior cruciate ligament transection model of
osteoarthritis. Osteoarthritis Cartilage 12, 614, 2004.
11. Maehara, H., Sotome, S., Yoshii, T., Torigoe, I., Kawasaki, Y.,
Sugata, Y., et al. Repair of large osteochondral defects in
rabbits using porous hydroxyapatite/collagen (HAp/Col)
and fibroblast growth factor-2 (FGF-2). J Orthop Res 28, 677,
2010.
12. Voor, M.J., Yang, S., Burden, R.L., and Waddell, S.W. In vivo
micro-CT scanning of a rabbit distal femur: repeatability and
reproducibility. J Biomech 41, 186, 2008.
13. Monfoulet, L., Rabier, B., Chassande, O., and Fricain, J.C.
Drilled hole defects in mouse femur as models of intramembranous cortical and cancellous bone regeneration.
Calcif Tissue Int 86, 72, 2010.
14. McErlain, D.D, Appleton, C.T., Litchfield, R.B., Pitelka, V.,
Henry, J.L., Bernier, S.M., et al. Study of subchondral bone
adaptations in a rodent surgical model of OA using in vivo
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
micro-computed tomography. Osteoarthritis Cartilage 16,
458, 2008.
Bouxsein, M.L., Boyd, S.K., Christiansen, B.A., Guldberg,
R.E., Jepsen, K.J., and Muller, R. Guidelines for assessment
of bone microstructure in rodents using micro-computed
tomography. J Bone Miner Res 25, 1468, 2010.
Marchand, C., Rivard, G.E., Sun, J., and Hoemann, C.D.
Solidification mechanisms of chitosan-glycerol phosphate/
blood implant for articular cartilage repair. Osteoarthritis
Cartilage 17, 953, 2009.
Frisbie, D.D., Cross, M.W., and Mcllwraith, C.W. A comparative study of articular cartilage thickness in the stifle of
animal species used in human pre-clinical studies compared
to articular cartilage thickness in the human knee. Vet Comp
Orthop Traumatol 19, 142, 2006.
Mocsai, A., Humphrey, M.B., Van Ziffle, J.A., Hu, Y., Burghardt, A., Spusta, S.C., et al. The immunomodulatory adapter
proteins DAP12 and Fc receptor gamma-chain (FcRgamma)
regulate development of functional osteoclasts through the
Syk tyrosine kinase. Proc Natl Acad Sci U S A 101, 6158, 2004.
Shamis, L.D., Bramlage, L.R., Gabel, A.A., and Weisbrode, S.
Effect of subchondral drilling on repair of partial-thickness
cartilage defects of third carpal bones in horses. Am J Vet
Res 50, 290, 1989.
Frisbie, D.D., Morisset, S., Ho, C.P., Rodkey, W.G., Steadman, J.R., and McIlwraith, C.W. Effects of calcified cartilage
on healing of chondral defects treated with microfracture in
horses. Am J Sports Med 34, 1824, 2006.
Hara, T., Tanck, E., Homminga, J., and Huiskes, R. The influence of microcomputed tomography threshold variations
on the assessment of structural and mechanical trabecular
bone properties. Bone 31, 107, 2002.
Bouchgua, M., Alexander, K., Carmel, E.N., d’Anjou, M.A.,
Beauchamp, G., Richard, H., et al. Use of routine clinical
multimodality imaging in a rabbit model of osteoarthritis—
part II: bone mineral density assessment. Osteoarthritis
Cartilage 17, 197, 2009.
Hunziker, E.B., Kapfinger, E., and Geiss, J. The structural
architecture of adult mammalian articular cartilage evolves
by a synchronized process of tissue resorption and neoformation during postnatal development. Osteoarthritis
Cartilage 15, 403, 2007.
Shapiro, F., Koide, S., and Glimcher, M.J. Cell origin and
differentiation in the repair of full-thickness defects of articular cartilage. J Bone Joint Surg Am 75, 532, 1993.
Qiu, Y.S., Shahgaldi, B.F., Revell, W.J., and Heatley, F.W.
Observations of subchondral plate advancement during osteochondral repair: a histomorphometric and mechanical
study in the rabbit femoral condyle. Osteoarthritis Cartilage
11, 810, 2003.
Address correspondence to:
Caroline D. Hoemann, Ph.D.
Department of Chemical Engineering
École Polytechnique de Montreal
2900 boul Edouard Montpetit
Montréal, Québec
Canada H3C 3A7
E-mail: [email protected]
Received: July 16, 2010
Accepted: December 06, 2010
Online Publication Date: January 14, 2011