Keywords abdominal aortic aneurysm - imaging - ultrasound - modeling - growth
Introduction
An abdominal aortic aneurysm (AAA) is a progressive focal aortic dilation. In humans,
the maximal anteroposterior aortic diameter (APdmax ) on ultrasound (USS) is typically used to quantify aortic size.[1 ] Murine AAA models (e.g., periadventitial porcine pancreatic elastase [PPE][2 ]) can be used to gain biological insights and trial novel therapeutics in the preclinical
laboratory setting. However, while the normal human aortic APdmax may be up to 3 cm, the murine aortic APdmax maybe up to 60 times smaller.[3 ] Despite this, the APdmax is routinely utilized in preclinical research using murine AAA models.[4 ]
[5 ]
While such a measurement can detect the presence or absence of AAA, we hypothesize
that it may not be able to accurately detect subtle but important changes to aneurysm
geometry which occur in response to manipulation of pertinent genes relating to aneurysm
biology or following administration of novel disease-modifying substances. Furthermore,
single APdmax measurement also neglects the length of the aneurysmal segment and hence may not
be the optimal marker of disease severity in the preclinical setting and arguably
in the clinical setting.[6 ]
[7 ]
Measurement of three-dimensional lumen volume (3DLV) of murine AAA using USS has good
correlation with volumetric reconstruction of histological sections through the same
aorta.[8 ] However, this technique requires specialist equipment to capture high-frequency,
high-resolution data, and standardization of the region of interest studied and gating
for cardiac and respiratory motion. The operator must be able to identify and demarcate
the correct aortic segment and must have access to software available to facilitate
rapid semiautomated quantification of the aortic volume. The inter- and intraobserver
variability of the 3D USS technique remain unreported and the differences in the capability
to detect aneurysmal change compared with APdmax remains unclear.
The primary aim of this study was to present a standardized method for 3DLV ultrasound
of the infrarenal aorta in mice in the context of PPE-induced AAA and to investigate
observer reproducibility compared with APdmax . Second, we compared the differences in measurements using the two methods for detecting
the presence of an AAA using serial scanning following PPE application to the aorta.
We also assessed the magnitude of the changes in the AAA detectable with each method
of measurement.
Materials and Methods
Murine Porcine Pancreatic Elastase Abdominal Aortic Aneurysm Model
Male C57BL6/J mice were purchased from Charles River (
https://www.criver.com
, United Kingdom) and used for experiments at 12 weeks of age. Animals were housed
in GM500 cages (Techniplast, Italy) with a 12-hour light/dark cycle and free access
to standard chow diet and triple-filtered drinking water. Each cage contained a dome
home and chew sticks as environmental enrichment. The animals were housed at a maximum
of five per cage.
Midline laparotomy was performed under recovery isofluorane anesthesia and the aorta
exposed using blunt dissection and PPE or saline applied to the aorta for 5 minutes,
as described by Bhamidipati et al.[2 ] The peritoneal cavity was washed out three times with normal saline and the abdominal
wall closed in layers with Vicryl sutures. Two independent cohorts of mice (cohort
A and cohort B) were used in this study. Cohort A was used to assess observer variability
in USS measurements. Cohort B was used to assess single APdmax /3D USS measurement sensitivity in detection of aortic dilation.
Animal work was performed in accordance with the UK Animals, Scientific Procedures
Act 1986. This study was performed under existing institutional approval (the Home
Office Project License PPL: P606320FB). All investigators undertook additional training
to obtain Home Office personal animal licenses.
In Vivo High-Resolution Ultrasound
All mice were prepared by fur removal shaving (Contura, Weller) and deepilation (Veet)
of the anterior abdominal wall from the costal margins (superior margin) to the pubis
(inferior margin) and midaxillary line to midaxillary line (lateral margins). Ultrasound
was performed with a Vevo2100 high-frequency preclinical micro USS (Visualsonics,
the Netherlands). This included an automatic motor, an MS-550D probe (40 MHz frequency),
a heated platform with electrocardiogram and respiratory recording, and a thermometer.
Transverse imaging was performed on a layer of aquasonic gel (Parker Labs, the Netherlands).
All imaging was gated for respiration and triggered at 50 ms after the r wave, as
this corresponded to peak dilation in the infrarenal abdominal aorta. Mice were imaged
under recovery isoflurane anesthesia.
The APdmax (inner-to-inner) was calculated from a single electrocardiography-gated Kilohertz
Visualization (EKV) recording which was made in the brightness (“B”) mode at the largest
section of the aorta ([Fig. 1 ]). The 3DLV was calculated from serial transverse images recorded along an 11.96-mm
segment caudally from the left renal artery (157 frames at 0.076 mm intervals). These
images were reconstructed into a 3D image by the VevoLab (Visualsonics, the Netherlands)
software package ([Fig. 1 ]).
Fig. 1 Example images from the Vevo2100 demonstrating anterior–posterior diameter (APdmax ) and three-dimensional lumen volume (3DLV) measurements in sham control and periadventitial
porcine pancreatic elastase (PPE) treated mice. A, anterior, Ca, caudal; Cr, cranial;
L, left; P, posterior; R, right.
Image Segmentation
Image analysis for both the APdmax and 3DLV was performed in VevoLab v1.7.0 (Visualsonics, the Netherlands). The inner-to-inner
aortic diameter was measured in millimeters from the single slice cross-sectional
EKV recording to obtain the APdmax . However, to assess the 3DLV, the inner lumen was manually traced every forth slice
recorded below the left renal artery. The software subsequently rendered the manual
tracings to derive a volumetric output in cubic millimeters (mm3 ).
Assessment of Observer Variability
To assess for the observer variability, APdmax and 3DLV measurements were recorded from imaging performed at day 14 post surgery
(i.e., PPE or sham). Two independent trained observers (i.e., observer 1 [O1] and
observer 2 [O2]) recorded measurements blinded to the measurements of the other and
to the surgery performed on the animal (sham or PPE). Both observers were doctoral
students who were trained by an independent expert coinvestigator with at least 5
years' experience in murine and human USS. O1 had a basic science background and O2
was a resident in vascular surgery with prior experience in USS. O2 measured the images
twice (O2a and O2b) blinded to the results from his previous measurement to determine
intraobserver variability. Images were presented to each observer in a random order.
Detection of Aneurysm Growth
For comparative analysis, of APdmax and 3DLV measurements, imaging was performed preoperatively and then every 3 to 5
days postoperatively for 14 days in an independent cohort of mice. Single observer
(O2) measurements were utilized as this was likely to reflect routine laboratory practice.
At each time point, APdmax measurements were compared with 3DLV measurements of both sham and PPE mice. Percentage
change over time in aortic dilation (compared with baseline) detected by each measurement
technique was assessed.
Statistical Analysis
Differences in observer measurements were evaluated using Bland–Altman plots with
limits of agreement (LOA), Pearson's correlation, and two-sample t -tests. The relationship between APdmax and 3DLV measurements was assessed using Pearson's correlation and linear regression
analysis. Differences in measurements between the groups (i.e., sham and PPE) at time
points were evaluated using two-sample tests. Measurements were reported as a mean ± standard
deviation (SD). Percentage change detected with each technique was evaluated using
scatter plots and the lines of best fit.
Statistical analyses were performed using Minitab (Minitab, PA). A p -value < 0.05 was considered significant.
Results
In total, cohort A consisted of 10 mice (7 PPE and 3 sham) and cohort B of 14 mice
(9 PPE and 5 sham). There was no case of aneurysm rupture or death during the study.
Observer Variability
There were no significant interobserver differences (mean difference: −0.05, LOA:
0.14–0.24, p = 0.100; [Fig. 2A ]) or intraobserver differences (mean difference: −0.03 mm, LOA: 0.16 to −0.21, p = 0.36; [Fig. 2B ]) in the measurement of the APdmax . Despite a strong positive correlation (r = 0.98, p < 0.001), there was a small but significant difference in the interobserver variability
in 3DLV measurements (mean difference: −1.38, LOA: 1.58 to −4.34, p = 0.008; [Fig. 2C ]). There were also no significant intraobserver differences in 3DLV measurements
(mean difference: −0.55, LOA: 1.19 to −2.28, p = 0.053; [Fig. 2D ]). Measurements of both the APdmax and 3DLV were also well correlated between O1 and O2 ([Fig. 2 ]). Repeated measurements of the APdmax and 3DLV by O2 (i.e., O2a and O2b) were also well correlated ([Fig. 2 ]).
Fig. 2 Bland–Altman plots and correlation analysis of: (A ) interobserver variations in anterior–posterior diameter (APdmax ) measurements, (B ) intra-observer variations in APdmax measurements, (C ) interobserver variation in three-dimensional lumen volume (3DLV) measurements and
(D ) intraobserver variation in 3DLV measurements. Adj., adjusted; O1, observer 1; O1a,
first image by observer 1; O2, observer 2; O2a, first image by observer 2; O2b, second
image by observer 2.
Measurement Techniques
We observed significant positive correlations between APdmax and 3DLV measurements in cohort B overall (r = 0.83, p < 0.001). This trend remained evident in subanalysis of aortic measurements in mice
which had undergone PPE surgery (r = 0.70, p < 0.001; [Fig. 3 ]). However, this correlation was not apparent in measurements from mice which had
undergone sham surgery only (r = 0.24, p = 0.5; [Fig. 3 ]).
Fig. 3 Relationship between anterior–posterior diameter (APdmax ) and three-dimensional lumen volume (3DLV) measurements in mice treated with periadventitial
porcine pancreatic elastase (PPE; open squares) and sham controls (closed circles).
The solid lines represent the associations between the two measurements for the sham
(r = 0.24, adjusted R
2 = − 0.06, p = 0.5) and PPE groups (r = 0.70, adjusted R
2 = 0.46, p < 0.001).
Both measurements, APdmax and 3DLV, were able to detect the presence of AAA at the standard time point of the
model (i.e., 14-day postoperatively; [Fig. 4 ]). Absolute mean differences between the sham and PPE groups were 0.66 ± 0.11 mm
for APdmax and 5.72 ± 0.56 mm3 for 3DLV. However, the 3DLV technique was able to identify a statistically significant
aneurysmal dilatation at an earlier time point than the APdmax method ([Fig. 4A ] and [B ]). Furthermore, the percentage increase in sequential aortic measurements compared
with baseline over the course of the 14 days was approximately two-fold greater using
3DLV measurements compared with APdmax ([Fig. 4C ]). For example, a 50% increase in aortic measurements was seen after approximately
3 days using 3DLV measurements compared with nearly 7 days using APdmax measurements ([Fig. 4C ]). This was also reflected in the different rates of growth detected (i.e., 3DLV
slope 12.3 versus APdmax slope 5.7; [Fig. 4C ]).
Fig. 4 Differences in aortic measurements using anterior–posterior diameter (APdmax ) (A ) and three-dimensional lumen volume (3DLV) (B ) in sham controls (closed circles) and periadventitial porcine pancreatic elastase
(PPE) treated mice (open squares). Percentage change in aortic size detected with
time (C ) with APdmax (open circles) compared with 3DLV (closed squares). The time taken to detect a 50%
increase in abdominal aortic size is shown with the dotted lines.
Discussion
We describe a standardized method for APdmax and 3DLV assessment of the abdominal aorta in the PPE murine model of AAA using a
Vevo2100 high-resolution preclinical ultrasound system. 3DLV provides a simple, reproducible,
and comprehensive measurement of geometric changes in the infrarenal aorta in mice
comparable to traditional APdmax measurement.[9 ]
[10 ] However, 3DLV can detect aneurysmal changes earlier than APdmax, and the magnitude of the detectable difference is also larger. This likely relates
to the capture of the entire length of the aneurysmal segment when utilizing 3DLV
measurement, rather than a single slice which may fail to capture subtle changes.
This is supported by the data from contemporary studies, which demonstrated that two-dimensional
(2D) USS techniques overlook regional differences evident with volumetric USS quantification.[11 ] Furthermore, 3DLV measurement of murine aorta has been demonstrated to highly correlate
with more sensitive modalities, such as computer tomography and magnetic resonance
imaging, as well as histological assessment of aneurysmal aorta.[8 ]
[12 ]
[13 ] Despite these observations, 2D USS assessment of murine aorta remains the dominant
strategy.[9 ]
[10 ] The results of our study may help preclinical researcher interpret and compare the
results between studies utilizing the two differing measurement techniques.
Similar to our method, Soepriatna et al[11 ] utilized cardiac and respiratory gating capturing multiple images over the aneurysmal
segment of the aorta. However, they utilized the angiotensin II (AngII) AAA murine
model. However, direct comparisons between 2D and volumetric measurements of the murine
aortic were not reported as described in our study. Importantly, they highlighted
volumetric USS to take between two and four times longer than 2D measurements, which
may explain the reluctance for routine adoption in basic science research.
While strong positive correlations in measurements with each method between observers
were seen, there were significant absolute differences in the 3DLV measurements between
observers. To calculate the 3DLV, the observer is required to delineate the aortic
lumen from the surrounding anatomy. This method is, therefore, subject to the operator's
interpretation of the grayscale sonographic images as to exactly where this boundary
exists. Small differences in the positioning of the measuring calibers between observers
are then incurred across multiple frames of analysis. Despite this small potential
for absolute differences, it is reassuring that the observations are well correlated
between observers. This suggests that observers can reliably identify the same boundaries
consistently across different animals, but that there will still be subtle differences
between different observers. Therefore, we recommend single observer measurement,
blinded to treatment allocation, of all imaging in a single experiment.
We evaluated 3DLV measurements in only the periluminal PPE AAA model based on the
expertise of our group. However, the technique can also be applied to the intraluminal
PPE model or the CaCl2 models using the protocol described. For the AngII AAA model,[14 ] where the aneurysm has a propensity to form in the suprarenal segment of the abdominal
aorta, the same approach could be adapted by setting the region of interest to capture
images 12-mm cranially starting from the right renal artery.
Limitations of the Study
We solely utilized male mice, as they were more likely to develop an AAA and also
due to the high prevalence of AAA disease in the male population.[15 ]
[16 ] We did not feel it necessary to validate APdmax and 3DLV in the AngII AAA model or female mice, as the underlying mathematical relationships
between the measurement techniques should theoretically be the same. However, based
on our findings, we would advocate the use of 3DLV in AAA murine experimentation utilizing
female mice, as it may facilitate the detection of subtle changes within the aorta.
Conclusion
In conclusion, the ability to detect changes in the abdominal aorta with 3DLV measurements
makes this technique ideally suited to evaluate gene or treatment effects which modulate
AAA development and progression in the preclinical setting. 3DLV measurement may detect
clinically significant changes that would be plagued by type-I error when using APdmax .
