close

Вход

Забыли?

вход по аккаунту

?

Measurement of radiographic changes in adjuvant-induced arthritis in rats by quantitative image analysis.

код для вставкиСкачать
ARTHRITIS & RHEUMATISM Volume 38
Number 1 . January 1995, pp 129-138
0 1995, American College of Rheumatology
129
MEASUREMENT OF RADIOGRAPHIC CHANGES IN
ADJUVANT-INDUCED ARTHRITIS IN RATS BY
QUANTITATIVE IMAGE ANALYSIS
RONALD E. ESSER, ALAN R. HILDEBRAND, RICHARD A. ANGELO, LYNNETTA M. WATTS,
MARK D. MURPHEY, and LARRY E. BAUGH
Objective. To apply quantitative analytical methods to the evaluation of radiographic images in experimental arthritis.
Methods. Adjuvant was used to induce arthritis
in rats. Arthritis progression was followed by conventional methods. In addition, digitized images of radiographs of the calcaneus were examined for changes in
the mean and in the distribution pattern of gray values.
Periosteal new bone formation was measured as an
increase in image area of the calcaneus.
Results. Significant changes in the gray value
profile and increases in periosteal bone formation occurred in arthritic rats. More extensive changes occurred in Lewis rats than in Sprague-Dawley rats.
Analysis of serial radiographs revealed an initial decrease in the density of juxtaarticular bone, followed by
progressive increases in gray value variation due to
concurrent bone loss and bone formation. Eventually,
bone formation in arthritic rats resulted in increased
gray values above those in nonarthritic rats.
Conclusion. Image analysis represents a sensitive, quantitative method for detecting radiographic
changes in experimental arthritis.
The adjuvant- and collagen-induced models of
chronic arthritis are widely used for studies of the
underlying pathophysiology of joint disease and for the
evaluation of potential new therapeutic agents. In both
models, grossly apparent joint inflammation rapidly
Ronald E. Esser, PhD, Alan R. Hildebrand, BA, Richard A.
Angelo, PhD, Lynnetta M. Watts, BA, Larry E. Baugh, PhD: Marion
Merrell Dow Research Institute, Cincinnati, Ohio; Mark D. Murphey,
MD: Armed Forces Institute of Pathology, Washington, DC.
Address correspondence to Ronald E. Esser, PhD, 2317
Life Sciences Building, Ciba-Geigy Corporation, 556 Morris Avenue, Summit, NJ 07901.
Submitted for publication January 18, 1994; accepted in
revised form August 2, 1994.
increases in severity after onset, and then plateaus or
declines gradually with time. Permanent changes in
joint architecture and loss of joint function are common during chronic disease (1,2). Histologically, joint
disease is characterized by proliferation of synovial
lining cells and connective tissue elements, and by the
destruction of articular cartilage and bone during latestage disease (1-3). Periosteal new bone formation
also occurs in both models, and during severe chronic
disease, joints may become fused by fibrous and/or
osseous bridges (1-3).
The osseous changes that occur during the
course of experimental arthritis have not been thoroughly characterized, partly due to the lack of adequate methods. Jamieson et a1 (4),using serial magnification radiography, described the appearance and
course of periarticular soft tissue and osseous changes
in collagen-induced arthritis in rats. Soft tissue swelling and osteopenia were reported to be early findings,
with subsequent development of bone erosions and
periostitis. Sloboda et a1 ( 5 ) described similar changes,
along with joint space narrowing, in a large series of
rats with adjuvant- and collagen-induced arthritis.
Those investigators reported that bone erosions, joint
space narrowing, and periostitis were prominent features of both models, but that bone demineralization
and involvement of the distal phalanges were more
prominent in adjuvant disease. Other investigators
(6-13) have described similar radiographic changes in
adjuvant- and collagen-induced disease and have used
subjective scoring systems to evaluate the effectiveness of various treatments. However, with few exceptions (9,10,14), quantitative radiographic methods
have not been applied in these models.
In this report we describe 5 different quantitative image analysis measures for following osseous
changes that occur during adjuvant-induced arthritis.
ESSER ET AL
130
These methods clearly differentiate between normal
and arthritic tissues and correlate well with conventional histologic and radiographic measures of disease
severity. In addition, the sequence of radiographic
changes that occur during the progression of experimental arthritis is described and the changes are
quantified.
MATERIALS AND METHODS
Adjuvant-induced arthritis. Female inbred Lewis rats
and outbred Sprague-Dawley rats (Charles River, Wilmington, MA) were injected intradermally at the base of the tail
with heat-killed Mycobacterium tuberculosis cells (Difco,
Detroit, MI) suspended in light mineral oil. Periodically
following adjuvant injection, hindpaw volumes (in ml) were
measured using a Ugo B a d e plethysomometer (Stoelting,
Wood Dale, IL). In addition, inflammation in each hindpaw
was scored grossly on a scale of 0 (normal) to 4 (severe
inflammation), based on the number of joints involved, the
severity and extent of periarticular erythema and edema, and
the degree of distortion and ankylosis of joints (15).
At the end of each experiment, the rats were killed,
the hind limbs were removed just distal to the knee, placed
on film carriers with the medial aspect of the limb down, and
then radiographed (lateromedial view). Radiographs were
taken in a Faxitron x-ray cabinet (Hewlett-Packard, McMinnville, OR) using Kodak XK-1 film (except where noted)
exposed for 10 seconds at 60 kV. The film-to-source distance
was 56 cm. For serial radiographs, the rats were anesthetized by intraperitoneal injection of pentobarbital (25 mg/kg)
and placed in lateral recumbency on film carriers. The hind
limbs were then extended away from the torso, taped into
position, and radiographed as described above.
Radiographs of each hindpaw were scored visually,
on a scale of 0 (normal) to 3 (marked changes), for soft tissue
swelling, bone mineralization, joint space narrowing, bone
erosion, periostitis, and the presence of heterotopic bone.
Total radiographic scores were calculated from the sum of
the scores for each abnormality, giving a maximum possible
score of 18. Radiographs were also examined by quantitative
image analysis as described below.
Preparation of tissues for histologic evaluation. At the
termination of some experiments, hindpaws were removed
and placed in 10% buffered formalin. The fixed tissues were
then decalcified in formic acid, embedded in paraffin, cut
into 5p sections, and stained with hematoxylin and eosin.
Each section was scored on a scale of 0 (normal) to 3
(extensive changes) for destruction of the calcaneus and
replacement of bone by pannus. Periosteal new bone formation around the calcaneus was similarly scored on a scale of
0 to 3. Histologic scores for destruction and new bone
formation were combined to give a maximum possible histologic score of 6 for each hindpaw.
Image analysis of radiographs. Radiographs were
illuminated with a Variquest 100 transilluminator (Fotodyne,
New Berlin, WI) and captured with a solid-state video
camera equipped with a MOS image sensor (MicroImage
Video Systems, Boyertown, PA). The camera was fitted
with a Nikkor 85-mm lens equipped with no. 4 and no. 1
close-up lenses, resulting in an 18-fold magnification of the
image on the video monitor. Video images were digitized
using a Targa M8 framegrabber (Truevision, Indianapolis,
IN) and then processed and measured using Optimas image
analysis software (Bioscan, Edmonds, WA). For each specimen, 2 digitized images were averaged, and a 3 x 3
smoothing filter was applied, to generate the final digitized
image for measurement.
Since the amount of soft tissue about the ankle varied
significantly in normal and arthritic rats, the input lookup
table was used to subtract the gray value contribution of soft
tissue prior to measurement of the mean and the distribution
of gray values within the calcaneus. This was accomplished
by measuring the gray value of soft tissue immediately
adjacent to the calcaneus and then subtracting that value
from the digitized image of the calcaneus. After processing,
the images were either measured directly or saved as TIFF
files for later measurement.
Changes in the calcaneus were evaluated by determining the gray value of each picture element (pixel) within
a standard template positioned over the image of the calcaneus, but not extending beyond its normal borders (10,14).
The template size varied from 5,372 to 8,364 pixels, depending on the age and strain of rats under study, but was held
constant within experiments. The resulting data points were
then displayed as a histogram, and the mean gray value, gray
value variance (a measure of the range over which gray
values extended), and the histogram peak height-to-variance
ratio were obtained from the histogram. The percentage of
data points lying outside the normal range was determined
by using an algorithm to calculate the mean gray value and
standard deviation from pooled data collected from normal
rats, and then calculating the percentage of data points from
individual images which fell more than two standard deviations above or below the range defined by the normal
population. Periosteal new bone formation around the calcaneus was measured by first differentiating bone from soft
tissue through binary thresholding, manually outlining the
calcaneus, and then determining the image area of bone (in
mm2) based on the threshold values within the outlined area.
The reproducibility and accuracy of the radiographic
and imaging procedures were monitored by measuring the
gray values of individual steps of a 13-step aluminum wedge
radiographed along with all specimens. As a measure of
variability introduced by film and processing conditions,
gray values of low-, mid-, and high-density steps were
measured on 20 randomly selected films taken over a 15month period. The coefficient of variance in this series of 20
measurements was 8.4%, 10.3%, and 9.6% for the low-,
mid-, and high-density steps, respectively. As a measure of
imaging system variation, low-, mid-, and high-density steps
from a single film were measured once daily on 10 different
days. The coefficient of variance for these measurements
was 2.0%, 1.6%, and 2.1% for the low-, mid-, and highdensity steps, respectively.
Statistical analysis. All statistical analyses were done
using Data Desk Professional software (Odesta, Northbrook, IL). Comparisons between groups were made using a
2-tailed t-test for independent means with pooled estimate of
variance. Since right-to-left hindpaw variation in individual
RADIOGRAPHIC CHANGES IN EXPERIMENTAL ARTHRITIS
35 r S P R A G U E - D A W L E Y 7 3 . 0
-
3.51-LEWIS73.0
PAWVOLUME
--C PAWVOLUME
rats was less than the variation between animals, values for
the 2 hindpaws were averaged for each animal and the
combined value was used to calculate descriptive statistics
and for t-tests. The relationships between different measures
of disease severity were described on individual right and
left values using Pearson's product-moment correlation.
Experimental protocols. Five different experiments
were performed to identify image analysis measures that
reflected changes present in arthritic joints and to relate
these measures to conventional methods of analysis. Experiment 1 was designed to describe the image analysis changes
that occur in chronic arthritis. Sprague-Dawley (n = 15) and
Lewis (n = 11) rats were injected with adjuvant and the
clinical course of arthritis followed for 32 days. The rats
were then killed and the hindpaws radiographed. Radiographs were examined both visually and by image analysis as
described above.
Experiments 2 and 3 were designed to determine the
relationship between image analysis findings and the results
of histologic evaluation of arthritic hindpaws. In Experiment
2, Sprague-Dawley rats (n = 20) were injected with adjuvant,
and groups of 5 randomly selected rats were killed at 14, 21,
28, or 35 days after adjuvant injection. The hindpaws of each
rat were radiographed, and the radiographs evaluated by
image analysis. The hindpaws were also evaluated histologically for periosteal new bone formation around the calcaneus and for destruction of the calcaneus and replacement
by pannus. Each abnormality was scored on a scale of 0 to 3.
131
The temporal relationship between the appearance and progression of changes detected by the two methods and the
relative degree of change was compared. In Experiment 3,
Sprague-Dawley rats (n = 25) were injected with adjuvant,
killed 32 days later, and then the osseous changes in the
ankle joint were examined by image analysis and by histology. The statistical relationship between the results obtained
by the two methods was then determined.
Experiment 4 was designed to determine the relationship between image analysis parameters and the results of
conventional radiographic assessment. In this experiment,
30 Sprague-Dawley rats were injected with adjuvant and
then killed 32 days later. The hindpaws were radiographed
and radiographs were examined by image analysis and by
conventional subjective scoring. The correlation between
the total radiograph score and each of the individual image
analysis results was then determined.
Experiment 5 examined the sequence of changes in
the image analysis findings as they appeared during the
course of disease. Fourteen Sprague-Dawley rats were injected with adjuvant and then radiographed 13, 21, 27, 35,
and 49 days after adjuvant injection. Four normal agematched control rats were also radiographed at these time
points. Radiographs were examined both visually and by
image analysis.
RESULTS
Experiment 1: clinical course of arthritis and
image analysis changes observed in chronic disease. The
clinical course of arthritis in Lewis and SpragueDawley rats injected with adjuvant is shown in Figure
1. In both strains, gross clinical disease appeared 8-12
days postinjection, reached maximum severity 10
days after onset, and then decreased gradually over
the remainder of the observation period. In both
strains, disease was still grossly apparent at the termination of the experiment.
Radiographs of the hindpaws taken at the termination of experiment 1 were evaluated visually. All
of the radiographic changes measured were more
severe in Lewis than in Sprague-Dawley rats (Table I).
Radiographs from this study were also evaluated by
image analysis. For technical reasons, image analysis
measurements focused on the calcaneus. The calcaneus is spatially isolated from other bones of the ankle
-
Table 1. Assessment of radiographic changes in adjuvant-induced arthritis*
Strain of rats
Sprague-Dawle y
Lewis
Soft tissue
swelling
1.1
1.8
* 0.2
* 0.2
Osseous
mineralization
Joint space
narrowing
Bone
erosion
Periosteal
proliferation
Heterotopic bone
formation
1.1 2 0.2
1.7 2 0.2
1.0 f 0.2
2.0 2 0.3
0.8 t 0.2
1.3 2 0.2
1.8 2 0.2
2.5
0.4 0.1
1.4 2 0.2
f
*
0.2
~
~~
* Rats were killed at the end of experiment 1, the hindpaws radiographed, and then scored on a scale of 0 (normal) to 3 (extensive changes)
for each of the 6 radiographic features listed. Values are the mean t SEM.
132
ESSER ET AL
SPRAGUE-DAWLEY
LEWIS
600 7
600-
1
600
k
400t
l
400
' O 0"C, 4 -
600
0
'!
400
from normal rats (Figures 2A and D) were characterized by symmetric peaks extending over a narrow
range of gray values. Histograms from SpragueDawley rats with adjuvant-induced arthritis (Figures
2B and C) had irregularly shaped peaks with an
increased range of gray values which could extend
either above or below the normal range of gray values.
Histograms of Lewis rats injected with adjuvant (Figures 2E and F) also had irregularly shaped peaks with
an increased range of gray values, but typically, the
majority of abnormal values in Lewis rats were below
the range of normal.
Four different histogram parameters were used
to describe the changes in the calcaneus that occurred
during chronic arthritis: 1) the mean gray value of all
pixels within the defined analysis region, to represent
overall bone density of the calcaneus; 2) the gray value
variance, to reflect the spread of gray values present;
3) the histogram peak height-to-variance ratio, as an
additional measure of the extent of gray value changes; and 4) the percentage of gray values that fell
outside of the normal range, to reflect overall bone loss
or bone formation.
The values for each of the 4 histogram parameters obtained from groups of normal control rats
and rats with chronic adjuvant-induced arthritis are
shown in Table 2. Statistically significant changes in
gray value variance, histogram peak height-to-variance
ratio, and the percentage of values outside the normal
range were present during late-stage disease. The
mean gray values decreased substantially only in
Lewis rats, where the percentage of data points falling
outside the normal range was significantly greater (P<
0.05) than that in arthritic Sprague-Dawley rats. Overall, the pattern of histogram changes was thought to
reflect the combined effects of juxtaarticular osteopenia, focal erosion of the calcaneus, and periosteal
new bone formation (see below).
Periosteal new bone formation, as measured by
2ook
00
100
GRAY VALUE
Figure 2. Gray value profiles of the calcaneus of normal and
arthritic rats. Radiographs were taken at the termination of experiment I , the gray value of each pixel within a standard analysis region
of the calcaneus was measured, and the results from individual
hindpaws displayed as a histogram. A, Normal Sprague-Dawley
rats. B and C, Sprague-Dawley rats with adjuvant-induced arthritis.
D, Normal Lewis rats. E and F, Lewis rats with adjuvant-induced
arthritis.
joint, is of sufficient size to permit large numbers of
individual bone density measurements, is relatively
uniform in density in normal rats, and is consistently
involved in the arthritic process.
Figure 2 shows the changes in the distribution
of gray values of images of the calcaneus that occur
during chronic arthritis. Histograms from nonarthritic
control rats are shown for comparison. Histograms
Table 2.
Changes in image analysis measures during chronic adjuvant-induced arthritis*
Strain of rats,
group
Sprague-Dawley
Normal
Arthritic
Lewis
Normal
Arthritic
Mean gray
value
Gray value
variance
Peak height
to variance
Percent outside
normal range
Calcaneus image
area (mrn')
9923
95 t 2
46 2 6
146 t 211
12 f 2
3 f It
11 * 6
26 4
*
20.8 2 0.5
27.5 t 1.2"
98 f 4
85 5 5
51 -t 10
125 C 1st
14 f 2
4 2 It
5 5 3
45 f 7 t
16.8 5 0.2
29.3 5 1.6t
* Rats were killed at the end of experiment 1, the hindpaws radiographed, and then examined by image
analysis. Values are the mean SEM.
t P < 0.05 versus normal controls.
*
133
RADIOGRAPHIC CHANGES IN EXPERIMENTAL ARTHRITIS
Table 3. Histologic evaluation of bone destruction and periosteal new bone formation during the
course of adjuvant-induced arthritis*
Measurement
Bone degradation
New bone formation
Total score
Day 14
0.5
5
0.2
0
0.5 f 0.2
Day 21
Day 28
Day 35
1.4 2 0.4
1.8 2 0.4
3.2 f 0.8
2.4 f 0.4
2.9 2 0.1
5.3 2 0.4
1.9 2 0.1
2.4 f 0.2
4.3 f 0.5
* Twenty Sprague-Dawley rats were injected with adjuvant, and groups of 5 randomly selected rats
were killed at the times indicated. The hindpaws were radiographed, and histologic sections were
scored on a scale of 0 (normal) to 3 (extensive changes) for bone degradation and periosteal new bone
formation around the calcaneus. Values are the mean f SEM.
an increase in the image area of the calcaneus on
radiographs, is also shown in Table 2. Although statistically significant increases occurred in both strains,
the increases were larger in the Lewis rats.
Experiments 2 and 3: relationship between
changes in image analysis parameters and histologically
graded bone destruction. In Experiment 2, rats were
killed at various times during the course of arthritis,
and joint changes were assessed by image analysis of
radiographs and by histology. Periosteal new bone
formation and bone destruction were first observed in
tissues collected on day 21 (Table 3), with generally
increasing amounts of bone destruction and new bone
formation observed at subsequent time points.
The mean values for the image analysis parameters at each time point are shown in Figure 3. The
mean gray value parameter changed little over the
course of disease. In contrast, histogram variance,
histogram peak height-to-variance ratio, and the percentage of data points outside the normal range all
changed progressively through day 28 and then stabilized. This pattern closely mimicked the pattern of
histologic changes described in Table 3.
As an additional comparison, combined histologic scores for bone destruction and periosteal proliferation from this series of rats were plotted against
values for each of the image analysis parameters.
Significant linear relationships were obtained for gray
value variance, histogram peak height-to-variance ratio, and percentage of data points outside the normal
range (r = 0.59, r = -0.66, and r = 0.53, respectively).
The mean gray values did not show a strong linear
relationship with the histologic scores (r = 0.22).
However, the mean gray value for hindpaws with
maximum histologic scores for bone destruction was
lower than that for hindpaws with only minimal bone
destruction (mean k SEM 60 +- 5 versus 74 4 2; P <
0.02). Similarly, the mean percentage of gray values
falling below the normal range was significantly
greater in hindpaws with maximum bone destruction
scores relative to hindpaws with minimal destruction
(47 +- 12 versus 3 k 1; P < 0.01).
In Experiment 3, Sprague-Dawley rats (n = 25)
were injected with adjuvant, and the relationship between histologic and image analysis parameters at a
single time point during chronic arthritis was examined. Consistent with the results obtained when these
measures were compared at multiple points over the
course of disease, statistically significant linear relationships were observed between histologic scores and
80 r
I
w
30
TIME OF HARVEST
TIME OF HARVEST
Figure 3. Bone density profile changes during the course of adjuvant.
induced arthritis. Sprague-Dawley rats were injected with adjuvant,
killed at various times after adjuvant injection, radiographed, and
then the bone density profile of the calcaneus determined by image
analysis. Kodak GPB film was used for radiographs in this experiment, and this change accounts for the differences in absolute values
for gray value, gray value variance, and peak height-to-variance
ratio from those shown elsewhere in this report. Bars show the mean
and SEM.
134
ESSER ET AL
Table 4. Bone density profile of the calcaneus in groups of rats with mild, moderate, or severe
arthritis, as judged by histologic grading of the calcaneus for bone destruction and periosteal new bone
formation*
Disease
severity
No. of
samples
Mean gray
value
Gray value
variance
Peak height to
variance
Data points outside
normal range
Normal control
Histologic grade
Mild
Moderate
Severe
22
102 k 1
50 2 3
10.6 2 0.9
4 2 1
11
13
100 t 3
102 t 3
9052
73 t 14
110 t 21
179 t 18
10.7 2.9
4.9 -+ 1.1
2.0 0.2
*
*
9*3
19 5
34 4
25
*
*
* Arthritic rats were killed and hindpaws were harvested and then radiographed and processed for
histologic evaluation. The bone density profile of the calcaneus was determined by image analysis, and
histologic sections were scored for bone degradation and periosteal new bone formation. Individual
hindpaws were classified as having mild (scores 0-2), moderate ( 3 4 or severe (5-6) histologic
changes. Values are the mean t SEM.
histogram variance and histogram peak height-tovariance ratio (r = 0.56 and r = -0.65, respectively).
Linear relationships between histologic data and histogram mean gray value and data points outside the
normal range were less strong (r = 0.37 and r = 0.48,
respectively).
Samples from this series were also grouped
according to the extent of histologically measured
bone changes, and image analysis parameters were
calculated for each group. As shown in Table 4,changes
in the image analysis parameters corresponded well
with disease severity measured histologically. With
increasing disease severity, the gray value variance
and peak height-to-variance ratio increased progressively as a result of concurrent bone destruction and
new bone formation. Similarly, the percentage of data
points falling outside of the normal range also increased with disease severity. Because of the opposing
effects of bone destruction and bone formation,
changes in the histogram mean gray value were the
least sensitive indicator of disease severity. The combined results of these two experiments strongly suggest that changes in image analysis bone density
profiles that occur over the course of disease are due
to bone destruction and new bone formation.
Experiment 4: relationship of image analysis
parameters to conventional radiologic measures of disease severity. A preliminary comparison of image analysis results with the results of conventional radiographic scoring suggested that there was good overall
agreement between the 2 methods (see Tables 1 and 2).
To more fully characterize this relationship, the correlation coefficient between results obtained from the
2 methods in a series of 60 arthritic hindpaws was
determined (Figure 4). There were strong relationships
between total radiographic scores and gray value
variance, histogram peak height-to-variance ratio, and
periosteal proliferation. Values for mean gray value
and the percentage of data points outside the normal
range were less strongly correlated with total radiographic score, due to the opposing effects of bone
degradation and bone formation. Similar strong relationships with gray value variance (r = 0.6&0.65),
histogram peak height-to-variance ratio (r = 0.540.77), and periosteal new bone formation (r = 0.7%
0.92) were also obtained when image analysis parameters were plotted against other conventional measures of disease severity, such as clinical score or paw
volumes. These results suggest that several of the
individual image analysis measurements are excellent
indicators of overall disease severity.
Experiment 5: time course of bone changes in
adjuvant-induced arthritis. As described above, changes
in the bone density profile of the calcaneus are due to
the combined effects of bone destruction and periosteal new bone formation. At any one time, these
changes represent the net, cumulative result of a
number of processes which occur during the acute and
chronic phases of disease. To follow the sequence of
radiographic changes, serial radiographs were scored
visually and evaluated by image analysis. Visual assessment indicated that arthritis progressed methodically from initial soft tissue swelling and osteopenia, to
subsequent development of erosions, joint space narrowing, and extensive periosteal new bone formation.
The appearance and course of the different manifestations varied however, with soft tissue swelling and
osteopenia reaching maximum severity early and then
declining somewhat, while joint space narrowing,
bone erosion, and periosteal bone formation developed later and then continued to increase throughout
the observation period (data not shown).
RADIOGRAPHIC CHANGES IN EXPERIMENTAL ARTHRITIS
.
Bml
I
rr.67
. .. . I
135
sively, reflecting continued bone degradation and coincident periosteal new bone formation. By late-stage
disease, the mean gray values in individual rats could
be either higher, lower, or comparable to those in
normal control rats, depending on whether bone degradation or deposition predominated.
I
DISCUSSION
t
-[
4
-1
5
10
IS
TOTAL RADIOGRAPH SCORE
Figure 4. Linear relationship between image analysis parameters
and conventional radiographic scores. Sprague-Dawley rats were
injected with adjuvant to induce arthritis, and radiographs of the
right and left hindpaws were taken when the rats were killed.
Radiographs were evaluated by conventional scoring methods and
by image analysis. Total radiographic scores were then plotted
against the values for each of the image analysis parameters and
Pearson's correlation coefficient (r) was calculated for each comparison. A, Mean gray value. B, Gray value variance. C, Histogram
peak height-to-variance ratio. D, Percentage of data points outside
the normal range. E, Periosteal proliferation.
Table 5 shows the mean values for each of the
image analysis parameters for this series of radiographs. In normal rats, only minor changes associated
with growth occurred in any of the parameters over
the course of the study. In adjuvant-injected rats, the
initial changes were a decrease in mean gray values
(relative to normal controls) along with an accompanying increase in data points outside the normal range.
With time, however, decreases in mean gray values
were at least partially reversed in most rats, as periosteal new bone formation became widespread. After
day 21, changes in gray value variance and the histogram peak height-to-variance ratio increased progres-
A number of changes in bone and periarticular
soft tissue that occur during the course of adjuvantand collagen-induced arthritis have been described
using conventional radiologic analysis (4-13). These
changes include soft tissue swelling, joint space narrowing, bone demineralization, erosion of bone, periostitis, and the presence of heterotopic bone. Using
image analysis of radiographs, we have devised sensitive, quantitative measures for several of these manifestations, and used these parameters to follow the
course and severity of adjuvant disease in 2 different
strains of rats. Four different image analysis parameters were used to measure changes in juxtaarticular
bone: 1) mean gray value of all pixels within a defined
analysis region of the calcaneus; 2) variance of gray
values within the analysis region; 3) histogram peak
height-to-variance ratio; and 4) percentage of histogram data points falling outside the normal range.
These measures clearly differentiated between normal
and arthritic rats (Table 2). The pattern of changes
observed in the density profiles suggested that these
changes were due to the combined effects of focal
bone degradation and periosteal new bone formation.
This possibility was verified by comparing image analysis results with the results of conventional radiologic
and histologic evaluation.
Evaluation of tissues collected at multiple time
points during the course of disease indicated that the
appearance of changes in image analysis parameters
coincided with the appearance of histologic changes in
joint tissues. In addition, there was a strong correlation between the values for image analysis parameters
and histologically scored changes in bone from tissues
collected over the course of disease, and in a large
series of tissues collected at a single time point during
chronic disease. Taken together, these results strongly
suggest that the bone density profile changes described
for the calcaneus are due to simultaneous degradation
and formation of bone.
Several of the image analysis measures employed in this study revealed differences in the osseous
changes that occur in Lewis and Sprague-Dawley rats
136
ESSER ET AL
Table 5. Progression of changes in image analysis parameters during the course of adjuvant-induced
arthritis*
Group
~~
Mean gray
value
Gray value
variance
Peak height to
variance
Percent outside
normal range
Calcaneus
image area
85 2 4
85 t 2
79 t 27
70 2 9
11.2 2 5.9
8.1 2 0.7
4 f 3
4 f 2
19.4 2 0.6
19.6 f 0.3
95 f 1
87 4 4
53 2 16
96 2 12
13.0 2 3.5
5.8 2 0.8
4 2 2
33 f 7
19.8 5 0.5
22.1 + 0.5
6
4
47 t 6
169 t 26
12.0 + 3.0
3.5 f 0.7
3 f 2
33 f 7
20.4 + 0.4
27.0 f 1.2
106 2 7
104 4 3
65 2 16
173 t 23
10.0 2 3.1
2.5 f 0.4
3 2 3
12 + 3
20.7 f 0.4
31.8 t 1.9
105
127
47 t 3
244 t 39
11.1
2.2
f 1.3
f 0.7
4+1
70 t 7
21.6 t 0.6
38.3 + 2.7
~
Day 13
Normal
Arthritic
Day 21
Normal
Arthritic
Day 27
Normal
Arthritic
Day 35
Normal
Arthritic
Day 49
Normal
Arthritic
102
85
4
4
4
-t
2
8
* Fourteen Sprague-Dawley rats were injected with adjuvant and then radiographed at various times
after injection, Normal age-matched rats (n = 4) were radiographed as controls. Radiographic images
of the calcaneus were analyzed as described in Materials and Methods. Values are the mean 2 SEM.
with adjuvant-induced arthritis. Mean changes in gray
value, percentage of data points outside the normal
range, and periosteal new bone formation were all
greater in Lewis rats, although adjuvant disease was
qualitatively similar in the 2 strains. Others have
previously reported that bone changes are more severe
in Lewis rats with adjuvant-induced arthritis than in
Sprague-Dawley rats (16).
Examination of serial films taken over the
course of adjuvant disease indicated that the initial
radiographic change was a decrease in the mean gray
value of the calcaneus. In some rats, this initial decrease in gray value occurred without a significant
change in the gray value variance or in the histogram
peak height-to-variance ratio. This suggests a uniform
loss of bone and is thought to reflect inflammationmediated osteopenia. As disease progressed, mean
gray values continued to decrease while the gray value
variance increased and the histogram peak height-tovariance ratio decreased. These changes coincided
temporarily with the progressive erosion and replacement of bone by connective tissue and with periosteal
new bone formation as identified by conventional
histologic methods.
Changes in bone density continued with time,
but the pattern varied in individual rats and with
disease duration. Typically, bone degradation predominated early in the course of disease, resulting in mean
gray values lower than those of normal rats. With
progression, mean gray values increased due to new
bone formation and eventually exceeded normal values in many animals. This progression was also apparent when histogram data points outside the normal
range were separately tabulated as data points above
or below the normal range. During early disease, up to
87% of all data points lying outside the normal range
fell below the normal values. In contrast, during
late-stage disease only 28% of the total data points
outside the normal range were below and 72% were
above. However, the relative degree of bone degradation and formation could vary in individual rats (particularly Sprague-Dawley rats), leading to gray values
which were either predominantly above or below the
normal range. This difference may be due to variation
in the pattern of disease that develops in individual
rats, similar to the variation in the clinical pattern of
disease described by Muir and Dumonde (17). Alternatively, all rats may progress through a uniform
series of changes but do so at different rates.
Two different types of bone loss were detected
by image analysis during the course of arthritis. Focal
erosion and replacement of bone by connective tissue
was evidenced by a decrease in gray value along with
an increase in gray value variance. Osteopenia appeared as a decrease in histogram mean gray value
without an increase in gray value variance. Other
investigators have also described juxtaarticular or
generalized osteopenia in experimental arthritis using
conventional radiographic methods (6-1 3), quantitative histomorphometry (18,19), or radiographic photo-
RADIOGRAPHIC CHANGES IN EXPERIMENTAL ARTHRITIS
densitometry (14,18--20). Subcutaneous implantation
of nonspecific irritants has also been shown to cause
the loss of cortical and cancellous bone in rats (21,22).
Taken together, these observations suggest that generalized bone loss may be a general sequela of chronic
inflammatory disease. Osteopenia also occurs in rheumatoid arthritis (23), and the clinical relevance of
osteopenia is apparent from the increased occurrence
of stress fractures in patients with longstanding disease (24-26).
Other investigators have previously used image
analysis of radiographs to measure the severity of
experimental osteoarthritis (27-29). In these studies,
changes in the joint space width, the area and density
of the menisci, and the area and density of the femoral
and tibia1 epiphyses were measured by image analysis
as an index of the severity of spontaneously occurring
disease in mice. This system, like that described in the
present report, differs from most radiographic systems
for measuring the severity of joint disease by focusing
on the measurement of changes in the physical characteristics of predetermined areas of bone and adjacent tissues, rather than by counting or measuring the
size of individual pathologic lesions.
The image analysis methods described here
offer several significant advantages over conventional
methods of assessment. First, conventional methods
rely on subjective visual assessment of films, with the
resulting potential for observer error in the assignment
of scores. In contrast, image analysis generates objective data derived from precisely defined analysis regions. Second, conventional methods lack sensitivity,
since only unequivocal lesions are scored and the
entire spectrum of joint damage from minimal to severe
is represented by a small number of incremental steps.
Image analysis methods, on the other hand, are capable of detecting 256 levels of gray and a large number
of gray value distribution patterns which reflect osseous change. This greatly enhances sensitivity and
provides a more powerful method for detecting change
over the course of disease. Finally, the data sets
generated by image analysis are continuously distributed over a large range and are more amenable to
mathematical treatment than are data sets generated
by conventionaI scoring, which contain only a small
number of possible scores.
In conclusion, the methods described here offer
a sensitive and quantitative alternative to conventional
histologic and radiologic methods for assessment of
experimental joint disease. Image analysis methods
are capable of detecting focal bone erosion, periosteal
137
new bone formation, osteopenia, and periarticular soft
tissue changes, and the severity of arthritis, as measured by image analysis, correlates well with the
results of conventional histologic and radiologic measures. These new methods should be of significant
value for demanding applications such as evaluation of
new therapeutic approaches, where precise measurements of change and tissue preservation are important.
REFERENCES
1. Pearson CM: Experimental joint disease: observations on
adjuvant-induced arthritis. J Chronic Dis 16:863-874, 1963
2. Trentham DE, Townes AS, Kang AH: Autoimmunity to type I1
collagen: an experimental model of arthritis. J Exp Med 146:
857-868, 1977
3. Pearson CM, Wood FD: Studies of arthritis and other lesions
induced in rats by the injection of mycobacterial adjuvant. VII.
Pathologic details of the arthritis and spondylitis. Am J Pathol
42:73-95, 1963
4. Jamieson TW, DeSmet AA, Cremer MA, Kage KL, Lindsley
HB: Collagen-induced arthritis in rats: assessment by serial
magnification radiography. Invest Radio1 20:324-330, 1985
5. Sloboda AE, Oronsky AL, Kerwar SS: Radiographic analysis of
experimental rat arthritis, Handbook of Animal Models for the
Rheumatic Diseases 11. Edited by RA Greenwald, HS Diamond.
Boca Raton, FL, CRC Press, 1988
6. Ackerman NR, Rooks WH 11, Shott L, Genant H, Maloney P,
West E: Effects of naproxen on connective tissue changes in the
adjuvant arthritic rat. Arthritis Rheum 22: 1365-1374, 1979
7. Benslay DN, Nickander R Comparative effects of benoxaprofen and other anti-inflammatory drugs on bone damage in
the adjuvant arthritic rat. Agents Actions 12:313-319, 1982
8. Blackham A, Burns JW, Farmer JB, Radziwonik H, Westwick
J: An x-ray analysis of adjuvant arthritis in the rat: the effect of
prednisolone and indomethacin. Agents Actions 7:145-151, 1977
9. Esser RE, Watts LM, Angelo RA, Thornburg LP, Prior JJ,
Palmer JT: The effects of fluoromethyl ketone inhibitors of
cathepsin B on adjuvant induced arthritis. J Rheumatol20: 1176
1183, 1993
10. Esser RE, Angelo RA, Murphey MD, Watts LM, Thornburg
LP, Palmer JT, Talhouk JW, Smith RE: Cysteine proteinase
inhibitors decrease articular cartilage and bone destruction in
chronic inflammatory arthritis. Arthritis Rheum 37:236247,
1994
11. Otterness IG, Larson DL, Lombardino JG: An analysis of
piroxicam in rodent models of arthritis. Agents Actions 12:308312, 1982
12. Sloboda AE, Birnbaum JE, Oronsky AL, Kerwar SS: Studies
on type I1 collagen-induced polyarthritis in rats. Arthritis
Rheum 24:616-624, 1981
13. Wolos JA, Frondorf KA, Esser RE: Immunosuppression mediated by an inhibitor of S-adenosyl-L-homocysteine hydrolase:
prevention and treatment of collagen-induced arthritis. J Immuno1 151526534, 1993
14. Esser RE, Angelo RA, Hildebrand AR: Measurement of bony
changes in experimental arthritis by quantitative image analysis,
Proceedings of the Electronic Imaging International Conference. Dallas, TX, Miller Freeman, 1992
15. Cromartie WJ, Craddock JG, Schwab JH, Anderle SK, Yang
C-H: Arthritis in rats after systemic injection of streptococcal
cells or cell walls. J Exp Med 146:1585-1602, 1977
16. Francis MD, Hovancik K, Boyce RW: NE-58095: a diphosphonate which prevents bone erosion and preserves joint architec-
ESSER ET AL
138
17.
18.
19.
20.
21.
22.
ture in experimental arthritis. Int J Tissue React 11:239-252,
1989
Muir VY, Dumonde DC: Different strains of rats develop
different clinical forms of adjuvant disease. Ann Rheum Dis
41538-543, 1982
Bogoch E, Gschwend N, Bogoch B, Rahn B, Perren S: Juxtaarticular bone loss in experimental inflammatory arthritis. J
Orthop Res 6:648-656, 1988
Bogoch E, Gschwend N, Bogoch B, Rahn B, Perren S: Changes
in the metaphysis and diaphysis of the femur proximal to the
knee in rabbits with experimentally induced inflammatory arthritis. Arthritis Rheum 32:617-624, 1989
Del Pozo E, Gubler HU, Perrelet R, Hager T, Wiederhold KH:
Non-invasive quantitative estimation of bone density in rats
throughout the life cycle and in arthritic osteopenia: preliminary
results. Horm Metab Res 20:630-632, 1988
Pfeilschifter J, Minne HW, Enzmann E, Krempien B, Ziegler R:
Inflammation-mediated osteopenia in the rat: the effects of
artificial granuloma and sham operation on cortical and trabecular bone. Bone 6:461465, 1985
Minne HW, Pfeilschifter J, Scharla S, Mutschelknauss S,
Schwarz A, Krempien B, Ziegler R: Inflammation-mediated
23.
24.
25.
26.
27.
28.
29.
osteopenia in the rat: a new animal model for pathological loss
of bone mass. Endocrinology 11550-54, 1984
Sambrook PN, Reeve J: Bone disease in rheumatoid arthritis.
Clin Sci 74:225-230, 1988
Fam AG, Shuckett R, McGillivray DC, Little AH: Stress
fractures in rheumatoid arthritis. J Rheumatol 10:722-726, 1983
Michel BA, Bloch DA, Fries JF: Predictors of fractures in early
rheumatoid arthritis. J Rheumatol 18:8044?08, 1991
Isdale AH: Stress fractures of the pubic rami in rheumatoid
arthritis. Ann Rheum Dis 52:681484, 1993
Pataki A, Reife R, Witzemann E, Graf HP, Schweizer A:
Quantitative radiographic diagnosis of osteo-arthritis of the
knee-joint in the C57BL mouse. Tech Diagn Pathol 3:109-123,
1992
Pataki A, Fries R, Ochsner K, Witzemann E: Qualitative
radiographic diagnosis of osteo-arthritis of the knee joint in the
C57BL mouse. Agents Actions 22: 123-130, 1987
Pataki A, Graf HP, Witzemann E: Spontaneous osteo-arthritis
of the knee-joint in C57BL mice receiving chronic oral treatment with NSAIDs or prednisone. Agents Actions 29:210-217,
1990
Документ
Категория
Без категории
Просмотров
5
Размер файла
952 Кб
Теги
measurements, adjuvant, induced, image, change, radiographic, arthritis, analysis, rats, quantitative
1/--страниц
Пожаловаться на содержимое документа