Feasibility of longitudinal monitoring of atherosclerosis with Pulse Wave Imaging in a swine model

Abstract

Objective.

Atherosclerosis is a vascular disease characterized by compositional and mechanical changes in the arterial walls that lead to a plaque buildup. Depending on its geometry and composition, a plaque can ruptured and cause stroke, ischemia or infarction. Pulse Wave Imaging (PWI) is an ultrasound-based technique developed to locally quantify the stiffness of arteries. This technique has shown promising results when applied to patients. The objective of this study is to assess the capability of PWI to monitor the disease progression in a swine model that mimics human pathology.

Approach.

The left common carotid of three hypercholesterolemic Wisconsin miniature swines, fed an atherogenic diet, was ligated. Ligated and contralateral carotids were imaged once a month over 9 months, at a high-frame-rate, with a 5-plane wave compounding sequence and a 5MHz-linear array. Each acquisition was repeated after probe repositioning to evaluate the reproducibility. Wall displacements were estimated from the beamformed RF-data and were arranged as spatiotemporal maps depicting the wave propagation. The PWV estimated by tracking the 50% upstroke of the wave was converted in compliance using the Bramwell-Hill model. At the termination of the experiment, the carotids were extracted for histology analysis.

Main results.

PWI was able to monitor the evolution of compliance in both carotids of the animals. Reproducibility was demonstrated as the difference of PWV between cardiac cycles was similar to the difference between acquisitions (9.04% vs. 9.91%). The plaque components were similar to the ones usually observed in patients. Each animal presented a unique pattern of compliance progression, which was confirmed by the plaque composition observed histologically.

Significance.

This study provides important insights on the vascular wall stiffness progression in an atherosclerotic swine model. It therefore paves the way for a thorough longitudinal study that examines the role of stiffness in both the plaque formation and plaque progression.

1. Introduction

Atherosclerosis can be described as an inflammatory disease for which a thick pocket of a mix of various components is covered by an endothelial membrane and a thin fibrous cap [1]. Atherosclerotic plaques might lead to stenosis of the vessel and are prone to rupture, potentially blocking the artery when releasing the plaque material. It is thus not surprising that carotid atherosclerosis is correlated with a higher risk of cardiovascular diseases such as myocardial infarction and stroke [2]. Neither the exact nature of the disease nor the atherosclerosis formation and progression are fully understood [3]. However, the composition of the plaque is recognized to be correlated with the risk of plaque rupture [4, 5]. A vulnerable plaque, prone to rupture, presents a thin fibrous cap enclosing a large lipid core and shows intraplaque hemorrhage and inflammation. On the opposite, a plaque with a thicker cap of smooth muscle cells and collagen matrix surrounding a necrotic calcified core is considered stable. Therefore, in recent years, several non invasive imaging techniques based on ultrasound, ultrasound elastography or magnetic resonance imaging (MRI), have been developed to study the plaque composition and/or biomechanics [6]. However, only ultrasound is used routinely in the clinic: to confirm the presence of plaque, to evaluate the degree of stenosis and the intima-media thickness. Ultrasound has also been proposed to identify the components based on the appeareance of the plaque in the B-mode image [7]. MRI can characterize the plaque morphology with high-resolution and identify the plaque components [8] but the clinical application is limited by the cost and the availability of equipment. Ultrasound elastography, consisting in quantifying the deformation of the imaged tissue in response to a certain constraint, has also been shown to correctly assess the vulnerability of plaque. Several elastography methods have been proposed and validated using different types of constraints: internal such as the pulse pressure wave generated by the heart [9–11] or external such as the push generated by acoustic radiation force (analyzed directly with Acoustic Radiation Force Imaging [12] or indirectly with Shear Wave Imaging [13, 14]).

Pulse Wave Imaging, developed by our team [15–22], is an ultrasound elasticity imaging method that quantifies locally the displacement of the arterial wall induced by the pulse pressure wave and estimates the Pulse Wave Velocity (PWV). At a large scale, from femoral to carotid artery, PWV has been associated with cardiovascular morbidity and mortality. At the local scale of an arterial segment, PWV can be converted in compliance via the Bramwell-Hill model [23]. Compliance derived from PWI has been found to reflect the plaque composition observed histologically [22, 24]. A first longitudinal monitoring study was performed in murine carotids to evaluate the capability of PWI to capture the early signs of atherosclerosis progression [25]. Despite being widely used as an atherosclerotic model, mice are known to present major disadvantages: e.g., the generated atherosclerotic lesions do not necessarily reproduce the human lesions, the small artery size requires different imaging methods preventing a direct translation to human scanning. Hence, the question remains whether or not the finding that PWI is able to monitor the progression of atherosclerosis in mice can be translated to humans. Ideally, a prospective study using a large cohort of subjects could be performed to answer this question. However, this would require dedicated studies with a large number of patients in a clinical trial. Swine represent a good intermediate animal model as atherosclerosis in swine closely mimicking that of humans but develops more rapidly (only within a few months). If PWI is able to predict development of vulnerable plaques in swine, it would provide additional confidence that this finding would be translatable to humans and potentially justify a large-scale prospective study involving humans. [26,27]. However, the specific anatomy, the size of the animal and the presence of aberrating fat layers might prove challenging for ultrasound imaging. Ge et al. proved feasibility of estimating displacement and strain in a swine model [28] but the animals were not monitored longitudinally.

This study aims to assess the feasibility of imaging the pulse wave propagation as previously developed and optimized by our group in an atherosclerotic swine model at different time points of the disease progression. The reproducibility of the method, in these specific conditions, is first evaluated. The PWV-based compliance of the atherosclerotic and normal carotids of three animals is monitored for ~ 9 months. At the termination of the experiment, histology of the lesions was performed and the plaque compositions were compared with the estimated compliances.

2. Materials and Methods

2.1. Animal model and study design

The procedures performed in this study were approved by the Institutional Animal Care and Use Committee (IACUC) of Columbia University (protocol AC-AAAU6460). Three 3-month old female Wisconsin Mini Swine-Familial Hypercholesterolemic (WMS-FH^™) were acquired from the University of Wisconsin Swine Research Farm (Madison, WI, USA). These animals would spontaneously develop atherosclerosis over the course of their life. However, to accelerate the process and make it more practical to monitor [29], the animals were fed a high fat diet (15% lard, 1.2% cholesterol) (UW Swine Research Farm, Madison, WI) and hemodynamic instability was induced by partially ligating the left common carotid of each animal. The right carotid, remaining intact, is expected to present less advanced atherosclerotic lesions, if any.

On the day of the surgery, anesthesia of the animals was induced with intravenous propofol. The animals were intubated and maintained under anesthesia for the procedure using 1 to 2% isoflurane. The throat area was incised on a 5-cm length, in the middle region to access and dissect the left common carotid. A 1.7 mm spacer (a 5F feeding tube/urethral catheter) was placed along the artery before tying it off with 5-0 Prolene (Ethicon, Cornelia, GA, USA). The spacer was then pulled out, creating a 80% stenosis in the carotid. Finally, the neck wound was surgically sutured layer by layer and the location of the ligation was indicated externally by a different suture.

The common carotids of the animals were imaged following the ultrasound protocol described below, immediately prior and following the surgery. The measurements were repeated two weeks after surgery and then once a month for 8 months. The animals were fed the atherogenic diet for the whole duration of the experiment.

One of the animals (animal #3) expired at 8 months. A necropsy was conducted by the Institute of Comparative Medicine veterinary staff: the cause of death was determined to be a stroke, due to a ruptured plaque. The common carotids were extracted during the necropsy. For the two other animals, following the last ultrasound acquisition, at 9 months, the animals underwent carotid angiography to document the presence of plaques and the degree of stenosis. They were finally euthanized with an intravenous bolus (100 mg per kg) of euthanasia solution. After euthanasia, an incision was performed in the neck of the animals and the common carotids were extracted, ensuring the whole region imaged in the ultrasound acquisitions was included.

2.2. Ultrasound data acquisition

The carotids of the animal were imaged with a 128-element, 5 MHz linear array (ATL L7-4, ATL Ultrasound, Bothell, WA, USA) controlled by a research Ultrasound system (Vantage 256, Verasonics Inc., Redmond, WA, USA). The array covers a field of view of ~ 38 mm. In order to fully image the common carotids of the animals, three different acquisitions were performed on the left, ligated carotid. Only one acquisition was performed on the right, intact carotid since this vessel is expected to be fairly homogeneous. To ensure the same arterial position was maintained over the course of the study, the suture that represents the location of the ligation was used as reference. Furthermore, previously acquired B-modes were visualized such that the sonographer was able to use landmarks on the ultrasound scan to approximately get the same FOV for each acquisition.

A high frame rate is required to capture accurately the pulse wave propagation [19]. A 5-angle plane wave compounding sequence was used to achieve this high frame rate while maintaining the data quality [30]. The plane waves covered an angular region of 2° and transmitted with a pulse repetition frequency of 8333 Hz. The angular region of 2 degrees and the compounded frame rate used was based on a previously performed optimization study [19]. The RF signals backscattered after each plane wave excitation were recorded during 1.2 s

For visualization and segmentation purposes, a single B-mode image of better quality was acquired using a 32-angle compounding sequence. The plane waves covered an angular region of 40°. This sequence was also used to generate the live B-mode guiding the sonographer.

The RF data corresponding to each excitation was subsequently beamformed using a delay-and-sum algorithm implemented in parallel with CUDA (CUDA 8.0, NVIDIA Corporation, Santa Clara, CA, USA). The beamformed RF frames were then coherently combined to create the sequence of beamformed compounded RF frames, at 1667 Hz, from which the propagation of the pulse wave will be analyzed.

2.3. Pulse Wave Imaging

The beamformed data were processed offline with the Pulse Wave Imaging technique developed previously by our group [15, 31–33]. The incremental axial displacements of the region of interest were estimated using a GPU-accelerated 1-D cross-correlation algorithm [15]. These axial displacements were converted into axial wall velocities by multiplying them with the frame rate. With the help of the reference Bmode, the walls of the carotid artery were manually segmented on the first frame. They were then automatically tracked across the successive frames. Similar to seismology or non-destructive testing, the propagation of the pulse wave was observed in a spatio-temporal map: the axial wall velocity waveforms observed at each lateral position were stacked, each row corresponding to one velocity waveform across time. Two spatio-temporal maps were created: one for the anterior wall, one for the posterior wall. In order to filter out any rigid motion, these two maps were subtracted and yielded a map corresponding to the distension of the artery [34].

The spatio-temporal map was then analyzed to estimate locally the Pulse Wave Velocity. A distinctive feature of the pulse wave, the 50% upstroke [33], was tracked across the whole field of view, at each lateral position. A linear fit was performed between the time of arrival of this feature and the traveled distance. The slope of this fit yielded a PWV estimate, while the coefficient of determination, r², and the confidence interval indicated the quality of the fit. Our team recently proposed more sophisticated PWV estimation methods to minimize the influence of reflections [18, 21] or to tackle the problem in heterogeneous arteries [35]. Because of the specifics of the animal model (namely different anatomy and different atherogenesis process), these techniques were not deemed necessary to obtain an accurate PWV estimate.

When multiple pulse wave velocities can be calculated for a specific location, the PWV that corresponds to the highest R2 is selected in order to reduce the probability of bias due to erroneous measurements. The PWV was converted into arterial compliance, k_p, using the definition proposed by Bramwell and Hill [23]:

$$k_p=\frac{A}{\rho .PW{V}^2},$$ (1)

A denoting the vessel area at the start of the cycle and ρ the blood density (1060 kg·m⁻³). The area is approximated assuming a circular cross-section whose diameter is measured on the reference B-mode by manual segmentation of the lumen. The end-diastolic diameter was derived using this high quality segmentation and the estimated displacements of the carotid wall, so that all the diameters used in this study correspond to the end-diastolic diameter. The end-diastolic diameter was selected since that aligns with the time-point in the cardiac cycle in which the PWV was measured throughout the entire study. The average diameter across the field of view was considered as the reference value.

2.4. Histology

The extracted carotids were fixed in 10% formalin according to the following procedure. The distal end of the carotid was closed with a suture, a syringe filled with formalin was introduced and tied with a suture, at the proximal end. Formalin was injected and the pressure was maintained for 5 min. The carotids were then stored in a solution of 10% formalin for at least 48 h to allow complete fixation of the tissue.

The left, atherosclerotic, carotid was cut in a series of 5 mm segments, using the site of initial ligation as a reference. Nine to fifteen segments were obtained, depending on the animal. For identification purposes, the pre-ligated area is labeled as region 1, the ligated area as region 2 and the post-ligated area as region 3. The contralateral carotid being more homogeneous, only three 5 mm segments were cut for histology analysis. These segments were placed in a solution of 70% ethanol before being embedded in paraffin. Two 5 μm slices were cut from each segment: one was stained with hematoxylineosin to determine the presence of atherosclerotic lesions and one was stained with Mason’s trichrome to quantify the degree of medial fibrosis. A color-based thresholding method was used to assess the fibrosis degree [36].

2.5. Reproducibility study

To ensure that the reproducibility of the PWI method in swine is suited to perform a longitudinal monitoring study, the difference in PWV and compliance estimates between two consecutive cardiac cycles was compared against the difference between two acquisitions following repositioning of the probe. The beat-to-beat difference between two consecutive cardiac cycles, recorded in one acquisition, indicates the level of variation in the PWV and compliance estimate that can be expected physiologically or due to measurement accuracy. On the other hand, the difference between two successive acquisitions indicates the level of variation in the PWV and compliance estimate that can be attributed to the repositioning of the probe. At each time-point of the 9-month experiment, one to two cardiac cycles were recorded during each of the two successive acquisitions for each animal. Among the estimates obtained from these data, the one corresponding to a linear fitting with the highest correlation coefficient and the smallest confidence interval was considered the reference PWV value. The compliance value based on this reference PWV was considered the reference compliance value.

The beat-to-beat difference was calculated as the between the PWV and compliance estimates obtained for the first and second cardiac cycles normalized by the reference PWV and compliance value. The differences obtained for all available acquisitions were averaged.

The inter-acquisition difference was calculated as the normalized average of the differences between the PWV and compliance estimates obtained in the first and the second acquisition; this average being normalized by the reference PWV and compliance value. Ultrasound imaging of the carotid of a healthy, young, small swine is easier than the equivalent of a fatty, atherosclerotic swine whose carotid has been ligated. For that reason, we quantified the reproducibility at two time-points: on the first day of the study, before performing the ligation surgery and on the final day of the study.

3. Results

3.1. Feasibility of ultrasound imaging in a growing swine model

On the day of surgery, at 3 months of age, the animals weighted 21.43 ± 3.77 kg. The growth of the animals followed an exponential curve with a sudden increase in weight during the 5 months following surgery and a slower growth for the remaining 4 months (Fig. 1). Over the 9 months of the study, the animals gained an average of 35.37 kg. The animal weight gain is attributed to the normal age-related growth of the animal but also to the accumulation of fat due to the atherogenic diet. The increasing number of fat layers can be seen, on the top part of the ultrasound images, i.e. under the skin (Fig.2a). As the fat layers accumulate over time, the carotids appear deeper and the sonographer has to push harder to compress the tissues and keep the arteries in the field of view. In addition, the echography of the ligated carotid is rendered difficult by the scarring of tissue around the ligation (Fig.2b). The ligation surgery also made the vessel more tortuous and thus difficult to capture in its entirety in the field of view of the US probe. However, despite those obstacles, it was possible to consistently and repeatedly image the same region of the carotid over the course of the experiment, as shown Fig. 3.

Figure 1.

Evolution of the weight of each animal across the 9 months of the experiment. The growth was exponential for the first 5 months and then slowed down for the remaining 4 months. Due to a problem with the weighting scale, a data point is missing for animal 2.

Figure 2.

Ultrasound B-mode images obtained on animal 2 on the first and last days of the experiment depicting the normal, contralateral carotid (a) and the ligated carotid (b). To help the reader, the region around the carotid is indicated by the dotted red lines. For both arteries, an increasing number of fat layers can be seen (red brackets). For the ligated artery, we can also notice the presence of additional scar tissues around the vessel.

Figure 3.

Ultrasound B-mode images obtained on animal 3 at each timepoint of the experiment. To help the reader, the region around the carotid is indicated with dotted red lines. After day 0, the ligation did not appear in the field of view of our 2-D acquisition. The surrounding structures on the B-mode as well as landmarks on the neck of the animal were used to locate the ligated area. The specific surrounding structures seen in all of these images indicate that the same area could be consistently imaged throughout the 9-month experiment.

3.2. Reproducibility study

Qualitatively, both on the first and last days of the study, the spatio-temporal maps acquired consecutively, for two consecutive cardiac cycles and after repositioning of the probe looks similar: the same waves are observed, across time, with the same spatial evolution of the axial wall velocities (see example of animal 3, Fig. 4).

Figure 4.

Spatio-temporal maps depicting the reproducibility of the pulse wave propagation characterization on the first and last days of the experiment, for animal 2. Comparing the two top rows indicates the degree of beat-to-beat, physiological variation. Comparing the two top rows and the bottom one indicates the inter-acquisition difference and thus the reproducibility. Qualitatively, the PWI technique is reproducible: the same waves are observed, across time, with the same spatial evolution of the axial wall velocities. (Note: the blank part on one of the maps is due to the fact the acquisitions were not triggered but manually aligned a posteriori.)

Quantitatively, we estimated the pulse wave velocity of the main wave for each available acquisition and each animal. The data were compared following the method described in section 2.5. On the first day of the experiment, the average beat-to-beat difference across animals was 6.80% while the inter-acquisition difference was 11.29%. On the final day, the beat-to-beat difference was 14.16% and the inter-acquisition difference was 8.10%. Pooling the data at the starting and end points of the experiment indicated a general beat-to-beat difference of 9.04% and a general inter-acquisition difference of 9.91%. The pooled difference of the radius was equal to 0.28% and 2.77% for beat-to-beat and inter-acquisition, respectively. The compliance is a combination of these measured parameters and therefore have a larger pooled difference of 10.52% and 19.44% for beat-to-beat and inter-acquisition respectively (Table 1).

Table 1.

The beat-to-beat and inter-acquisition differences of the PWV, radius and the compliance at day 0, the last day and pooled.

	PWV		Radius		Compliance
	Inter-acquisition [%]	Beat-to-Beat [%]	Inter-acquisition [%]	Beat-to-Beat [%]	Inter-acquisition [%]	Beat-to-Beat [%]
Day 0	11.29	6.80	2.54	0.21	22.97	8.68
Last Day	8.10	14.16	3.00	0.34	15.92	12.35
Pooled	9.91	9.04	2.77	0.28	19.44	10.52

3.3. Temporal evolution of compliance

The ligated carotid of each animal evolved with a unique pattern, despite having followed the same protocol (Fig.5a). The artery of the animal 1 was the softest and became noticeably softer over time. The carotid compliance of animal 2 followed a similar but more moderate increasing trend while the compliance of the ligated carotid of animal 3 was the stiffest and remained constant. For all three animals, we note an initial decrease in compliance at the two weeks timepoint, before observing the trends mentioned above. The same initial decrease in compliance was observed for the contralateral carotids (Fig.5b). The compliance of the carotids of animals 1 and 2 then followed the same trend with a slow increase up to 180 days. The carotid of animal 2 then exhibited a similar stiffness until the final day while the carotid of animal 1 presented an increase in stiffness, back to the initially observed value. The carotid of animal 3 showed an opposite evolution with a global decrease in compliance over time. As observed for the ligated carotids, the contralateral carotid of animal 3 was significantly stiffer than the carotids of the two other animals.

Figure 5.

Evolution of the compliance (ab), diameter (cd) and the PWV (ef) of the ligated (left) and the contralateral carotid (right) based on the PWV value with the highest r². The 95% CI is based on the uncertainty of the linear regression coefficient. Note that this does not include the uncertainty caused by the measurement variation in the radius. The evolution of the compliance of the ligated carotid is different for each animal, reflecting the individual progression of atherosclerosis. The compliance of the contralateral carotids, for animals 1 and 2, follow the weight gain pattern (indicating the influence of geometry in the compliance). The carotids of animal 3 are stiffer than the carotids of the two other animals.

3.4. Spatial evolution of compliance

Spatially, the propagation of the pulse wave was homogeneous over region 1 (before ligation) and over region 3 (after ligation) (Fig.6). In region 2 (around the ligation), we note the existence of a reflected wave arising from the initial ligation site. Since this reflection affects the PWV estimation and since histology does not provide insight into to the plaque components in this region, we decided to focus this initial study on the regions 1 and 3 of each animal.

Figure 6.

Spatial evolution of the ultrasound images (top row) and spatio-temporal maps (bottom row) for the ligated carotid of animal 2. Region 1 is located before the ligation, region 2 is centered around the ligation and region 3 is located after the ligation. PWV is slightly higher in region 1 than in region 3 but the carotid diameter is estimated to be smaller in region 3. We can thus conclude that for this animal, the carotid is stiffer, less compliant in region 3 than in region 1 (Table 2).

Because the propagation appeared homogeneous, we calculated PWV and compliance over the whole field of view instead of using our previously developed adaptive piecewise PWI method [22]. As shown Table 2, at the end of experiment, for animals 1 and 2, the arterial compliance was higher in region 1 than in region 3. While for animal 3, the carotid artery was stiffer in region 1 than in region 3.

Table 2.

Spatial evolution of compliance for each animal, region 1 being proximal and region 3 distal

Compliance (m2 · Pa−1)	Region 1	Region 3
Animal 1	2.18e−9	0.29e−9
Animal 2	0.55e−9	0.41e−9
Animal 3	0.03e−9	0.25e−9

In average through the whole experiment, the r² coefficient was bigger and the confidence interval was smaller for the contralateral side than for the ligated one 3. This confirms that the propagation is significantly more homogeneous on the contralateral side.

3.5. Histology results

For both animals 1 and 2, histology revealed the presence of typical atherosclerosis features throughout the whole ligated carotid. Representative slides for each animal, Fig. 7 a) and d), showed reduced lumen with respect to histology slices without plaque further away from the ligation, an accumulation of foam cells, a broken elastic lamina allowing smooth muscle cell infiltration and proliferation, some neoangiogenesis and a disorganized media layer. These features were more pronounced in the distal part of the carotid, post-ligation. The plaques in this area were in a more advanced state as we can notice calcifications surrounded by a necrotic area as well as a fibrous cap around the lumen (Fig. 7 b), c) and e), f)). The plaques observed in these two animals can thus be qualified as active fibroatheroma lesions, type V according to the AHA classification [1]. The H&E images of the ligated carotid of animal 3 also depicted a highly disorganized media layer and accumulation of foam cells along with smooth muscle cell proliferation in the media. The unique features observed Fig. 8 a), multiple lumen and presence of hemosiderin, indicated the existence of a revascularized thrombus. Therefore, histology confirms the diagnosis of the veterinarians (i.e. the cause of that animal death is likely a stroke). For that animal, the distal region presented a less active plaque than the proximal region, with only presence of foam cells in the intima layer.

Figure 7.

Representatives H&E stained sections of the ligated carotids of animal 1 (a), b) and c)) and animal 2 (d), e) and f)). Lesions from the proximal region 1 (a) and d)) are less advanced than the ones from the distal region 3 (b), c) and e), f)). SMC: smooth muscle cells proliferation, F: foam cells, C: calcification, N: necrosis, H: hemorrhage.

Figure 8.

Representatives H&E stained sections of the ligated carotids of animal 3. Lesions from the proximal region 1 (a) and b)) are more advanced than the ones from the distal region 3 (c)). The presence of multiple lumen along with hemosiderin indicates the presence of a revascularized thrombus. SMC: smooth muscle cells proliferation, F: foam cells, C: calcification, N: necrosis, H: hemorrhage.

On the contralateral side, the three sections cut for each animal all show a similar intact vessel indicating that the right carotid was homogeneously affected by the high fat diet and ageing. Therefore, we only focused on the central section for each animal, Fig. 9. Qualitatively, we note differences in the organization of the media layer between the three animals: animal 3 presenting a thicker media layer with a looser organization than animals 1 and 2. Quantitatively, the color-based segmentation evidenced a higher degree of medial fibrosis for animal 3 (32.1%) compared to animal 1 (26.1%) and animal 2 (16.6%).

Figure 9.

Representatives Mason’s trichrome stained sections of the contralateral carotids of the three animals (a), c) and e)). Fibrosis appearing blue in these sections, a color-based thresholding method is applied to quantify the medial fibrosis. The thresholded images (b), d) and f)) reflect the compliance observations on the last day of the experiment.

4. Discussion

As evidenced by the B-modes (Fig. 3), ultrasound-based PWI can image the carotids of hypercholesterolemic pigs throughout the progression of atherosclerosis. PWI yielded compliance measurements for each animal, every month for 9 months with the exception of two timepoints (day 29 and day 146) for animal 2. These outliers can be attributed to the difficulty in obtaining ultrasound images of good quality over the full field of view of the probe. In the corresponding images, the poor echogeneicity of the walls indicated a misalignment of the probe. This difficulty depends on the position of the animal on the table, on its specific anatomy and can be, in most cases, overcome by the sonographer after some training.

The reproducibility study conducted for the three animals on the first and last day of the experiment indicated that the inter-acquisition difference in PWV (9.91%) is similar to the beat-to-beat difference (9.04%). The variation in the PWV estimate due to the positioning of the probe is therefore similar to the physiological variation from one cardiac cycle to the other. The difference in compliance is higher, up to 20%, which is caused by variations in the initial radius segmentation between acquisitions. In future studies, automated segmentation, together with additional information such as power Doppler, will be pursued to reduce the variation caused by the segmentation and thereby improve the repeatability of monitoring of compliance. Nonetheless, the observed differences between the pigs in this study is larger than 20% and thereby was still capable of differentiating the stiffness between the three different swine. We can note an improvement in the reproducibility between the first day (11.29%) and the last day of the experiment (8.10%). This can be interpreted as an improvement in the skills of the sonographer who scanned pig carotids for the first time on the first day of the experiment. On the other hand, a decrease in the beat-to-beat consistency was observed (from 6.80% to 14.16%) and can likely be explained by the progression of the disease. One potential source of variation is the change in pressure. In future studies, the pressure will be measured invasively using telemetry to investigate the potential impact of pressure on the PWV-measurements and its beat-to-beat consistency. Another potential cause of this variation is the accuracy of PWI. The 95% CI indicates the uncertainty of the regression coefficient, and shows the uncertainty of the linear regression coefficient on which the PWV is based. In the case of disease progression, the pulse wave propagation becomes less linear. This increases the uncertainty of the linear regression coefficient and thereby also increases the beat-to-beat variation.

The compliance evolution of the ligated carotid (Fig. 5 a) reflects the individual progression of atherosclerosis in each animal. This observation is consistent with the results of the studies using a similar model: Ishii et al. found that 4 out of 6 carotids presented advanced lesions and the cross-sectional area stenosis presented a high standard deviation (53.4 ± 34.8%) [29]; Shi et al. identified advanced plaques in only 8 out of 17 carotids. More generally, conflicting findings have been reported regarding the association between PWV (or arterial stiffness) and the presence of plaque. Farrar et al. used a monkey model to assess this association: they observed an initial decrease in PWV before a linear increase with the progression of the disease [37]. However, this study was not performed longitudinally but with different groups of animals and their diet-induced atherosclerosis model differs from the progression of the disease in humans. In a rabbit model, Zhang et al. observed an increase in arterial stiffness as atherosclerosis progresses [38], even though neither intimal thickening, elastic membrane disruption nor foam cells were observed. Paini et al. studied the mechanical characteristics of atherosclerotic plaques in 25 patients and identified two different groups: 16 patients presented a higher vessel distensibility at the plaque site compared to the adjacent wall while 8 patients presented a lower distensibility at the plaque site [39].

The trend in compliance reflects the stiffness of the plaque components but also the remodeling capability of each animal: the carotid of animal 1 became more compliant over time and we observed a more preserved media layer (Fig. 7) and a faster increase in lumen diameter than for the other animals while the carotid of animal 3 became less compliant over time and we observed a more damaged media layer (Fig. 8) and a slower increase in lumen diameter.

The initial decrease in compliance observed for each animal is opposite to the trend reported by Farrar et al. in their primate model [37]. However, the time-frame was different: their initial observation date was 6 months and the final one at 30 months. In our case, the decrease in compliance might be attributed to a bias due to the reflection artifact presented by the initial ligation rather than to a physiological change in the properties in the first days after surgery. Another explanation is a systemic effect induced by either the ligation or the high fat diet. This could explain why both the ligated and the contralateral carotid show this initial decrease in compliance. Invasive telemetry measurements will be performed in subsequent studies to investigate the impact of the ligation and high fat diet on the blood pressure.

The compliance of the contralateral carotids of animals 1 and 2 appears to be related to the growth of the animals: from day 0 to day 150, the animals were rapidly growing and the compliances increased while after day 150, the growth was happening at a more moderate rate (Fig. 1) and the compliance appeared to decrease (animal 1) or appeared to plateau (animal 2). Future studies with a larger cohort size will provide additional insights. The decrease in compliance observed for animal 3 suggest that the mechanical properties of the artery were degrading faster than the vessel was remodeling its geometry. The Bruneck study group previously reported an association between the lack of vascular remodeling and plaque thrombosis such as the one observed for animal 3 [40].

While no gold standard is available to confirm the accuracy of the compliance evolution, the histological analyses can validate the observations made on the last day of the experiment. Indeed, mechanical testing has been performed by several teams to quantify the properties and the respective influence of the different vessel components (smooth muscle cells, elastin and collagen). For example, Kochova et al. evidenced that a decrease in smooth muscle cells or collagen would cause a stiffening of the artery [41]. Dinardo et al. have shown that a decrease in elastin was associated with an increase in smooth muscle cell rigidity [42] and Qiu et al. that arterial stiffness was related to smooth muscle cell rigidity [43].

Histology revealed that the media layer for animals 2 and 3 were more disorganized than for animal 1 and the elastic lamina was noticeably more disrupted for animal 3 than for the two other animals (Fig. 7 and 8). The lesion components also differed between the three animals: mostly smooth muscle cells for animal 1, a mixture of foam cells and smooth muscle cells for animal 2, mostly foam cells for animal 3. These observations are consistent with the fact that the ligated carotid of animal 1 was found to be softer than the one of animal 2, which was itself softer than the one of animal 3 (Fig. 5). Indeed, Tracqui et al. measured local Young modulus of different plaque components with acoustic force microscopy and observed that cell-rich fibrosis regions (such as seen in animals 1 and 2) are, on average, 5–6 times softer than hypocellular fibrosis regions (such as seen in animal 3) [44].

The image analysis of the histological sections on the contralateral carotid indicated a lower degree of medial fibrosis for animal 2 and a higher degree of medial fibrosis for animal 3 (Fig. 9). These different degrees of fibrosis reflect the compliance observed on the final day of the experiment, Fig. 5. We note that the relationship between the medial fibrosis and the compliance does not seem to be linear. However, the number of animals is too small to draw conclusions.

In future studies with a larger cohort, more extensive comparison within the ligated carotid and non-ligated cohort could be performed to quantify the ability to resolve the differences in histology between the ligated and non-ligated carotid. Due to the difference in the wall composition between the ligated carotid with plaque, and non-ligated carotid without plaque, a more extensive comparison would not lead to additional insights with the sample size used in this feasibility study. Similarly, PWI estimated that, in animals 1 and 2, the ligated carotids were softer in the region 1 (pre-ligation) than in region 3 (post-ligation) (Table 2) while histology showed that the more advanced lesions were located in region 3 (Fig. 7). Thim et al. reported the same regional arrangement of lesions in their hypercholesterolemic pigs with partially occluded carotids [45]. The opposite organization was observed for animal 3 with more advanced lesions in the region 1 (Fig. 8) and PWI correctly estimated the carotid to be stiffer in that region.

While these initial feasibility findings provide confidence in the observed variability in compliance values being dominated by the observed individual differences of the animals, further studies with a larger cohort of swine would aid in quantifying the ability to distinguish between different types of plaques. While these initial feasibility findings provide confidence in the observed variability in compliance values being dominated by the observed individual differences of the animals, further studies with a larger cohort of swines would aid in quantifying the ability to distinguish between different types of plaques.

While this initial study proves feasibility of imaging the pulse wave propagation in this atherosclerotic swine model with promising results, a number of limitations are also entailed.

The ligation model used to speed up the atherosclerotic process created advanced plaques with human-like features. However, the lesions appeared to be diffuse rather than focal: histology showed similar lesions in all the sections along the carotids and the ultrasound images presented a diffuse speckle pattern in the whole lumen. Such a type of plaques has been reported in patients presenting risk factors such as hyperlipidemia and/or hypertension and is referred to as nonstenotic atherosclerosis [46]. Our previous reports [22, 35] focused on the stenotic atherosclerosis characterized by a focal lesion, usually located around the carotid bulb. In order to produce this second type of lesions, alternatives to the ligation model could be considered for future studies: for example, allowing more time to let the atherosclerotic lesions develop after stopping the high-fat diet or creating an injury in the carotid wall using an angioplasty balloon [29].

The main source of bias in our PWI measurements comes from the model we used. The Bramwell-Hill equation has various underlying assumptions such as negligible wall viscosity, negligible radial fluid flow and a plug flow velocity profile. Furthermore, it assumes an infinitely long straight tube [4]. While the carotid is not an infinitely long straight tube, the observed reflection caused by the bifurcation is distant enough to not impact the 50% upstroke markers. A plaque can also cause reflections impacting the results, however, the observed plaques in this study are diffuse in nature and do not present a sudden change in radius and/or material composition, reducing the impact of the reflections on the compliance value. Correcting for assumptions such as negligible radial flow and velocity profile requires measuring additional variables, each of them with their own measurement uncertainties. While it is true that not satisfying all assumptions could induce a bias in the compliance value, the progression of the estimate over time can still provide important insights into how the plaque develops. When using the Bramwell-Hill model, we assumed circular symmetry for the artery (and the plaque) to use the lumen diameter as a proxy for the lumen area. 3-D PWI imaging could overcome this limitation [20]. The B-mode quality of the research ultrasound scanner used in this study is not optimal for image segmentation, introducing a potential bias in the diameter measurements. The luminal area in the ligated carotid increases based on the US-images (Fig. 5cd), whereas a reduced lumen can be observed in the histology results. This could be due to the echolucent nature of the plaque components. Another potential cause for this discrepancy is that the pressure in-vivo compresses the plaque and make it look more obstructed in the histology than it would be in-vivo. Furthermore, maintaining or increasing luminal area in plaque progression can be explained by positive remodeling [36]. In future studies, clinical ultrasound systems that provide better B-mode quality, power Doppler imaging and ex-vivo scanning of the extracted carotid samples will be pursued to better understand this observation.

This study in 3 animals enabled us to prove initial feasibility but further studies with a higher sample number are warranted in order to assess the correlation between plaque components and compliance as well as to evaluate the capability of PWI to diagnose atherosclerosis in its early stages.

5. Conclusion

In this initial study of the ligated and contralateral carotids of three hypercholesterolemic swines, we have demonstrated the feasibility of using PWI to monitor longitudinally, over ~ 9 months, the compliance of arteries affected by advanced atherosclerotic lesions. The PWI technique was found to be reproducible as the beat-to-beat difference in PWV was similar to the inter-acquisition difference. Histologically, the components of the created lesions were similar to the ones observed in patients. The compliance estimates provided by our technique reflected the histology findings. Therefore, this study represents a significant first step in using PWI to monitor atherosclerosis progression.

Table 3.

Comparison of homogeneity indices (r² and confidence intervals) between ligated and contralateral carotids

	r 2		Confidence interval (m · s−1)
	Ligated	Contralat.	Ligated	Contralat.
Animal 1	0.81	0.95	2.15	0.41
Animal 2	0.91	0.93	0.49	0.54
Animal 3	0.65	0.73	5.21	4.85