Abstract
Purpose To characterise microstructural contributions to the magnetic susceptibility of carotid arteries.
Method Arterial vessels were scanned using high resolution quantitative susceptibility mapping (QSM) at 7T. Models of vessel degradation were generated using ex vivo porcine carotid arteries that were subjected to several different enzymatic digestion treatments that selectively removed microstructural components (smooth muscle cells, collagen and elastin). Magnetic susceptibilities measured in these tissue models were compared to those in untreated (native) porcine arteries. Magnetic susceptibility measured in native porcine carotid arteries was further compared to the susceptibility of cadaveric human carotid arteries to investigate their similarity.
Results The magnetic susceptibility of native porcine vessels was diamagnetic (𝒳native = −0.1820ppm), with higher susceptibilities in all models of vessel degradation (𝒳elastin degraded = −0.0163ppm; 𝒳collagen degraded = −0.1158ppm; 𝒳decellularised = −0.1379ppm; 𝒳fixed native = −0.2199ppm). Magnetic susceptibility was significantly higher in collagen degraded compared to native porcine vessels (Tukey-Kramer, p<0.01) and between elastin degraded and all other models (including native, Tukey-Kramer, p<0.001). The susceptibility of fixed healthy human arterial tissue was diamagnetic and no significant difference was found between fixed human and fixed porcine arterial tissue susceptibilities (ANOVA, p>0.05).
Conclusions Magnetic susceptibility measured using QSM is sensitive to the microstructural composition of arterial vessels – most notably to collagen. The similarity of human and porcine arterial tissue susceptibility values provides a solid basis for translational studies. As vessel microstructure becomes disrupted during the onset and progression of carotid atherosclerosis, QSM has the potential to provide a sensitive and specific marker of vessel disease.
1. Introduction
1.1 Carotid atherosclerosis and stroke
Atherosclerosis is a widespread form of cardiovascular disease that causes the formation of plaque inside arterial vessels. Plaque formation is caused by a build-up of fat, cholesterol, calcium, fibrous tissue and other substances that disrupt the microstructural composition of the vessel, causing narrowing of the luminal space and stiffening of the vessel. Carotid arteries are particularly prone to plaque formation, and stenosis or rupture here can have catastrophic consequences as these vessels carry the main supply of blood to the brain. An estimated 20% of ischaemic stroke is due to rupture of carotid plaques, with the majority of ischaemic stroke caused by stenosis1,2. Carotid artery disease is also responsible for nearly 50% of transient ischaemic attacks (TIA) and linked to an increased risk of heart attack3. Early identification and treatment is of utmost importance to avoid stroke related disability and death caused by carotid plaque. The current indicator for surgical intervention is to assess the degree of stenosis caused by atherosclerotic plaque in the vessel. Although the patency of the blood vessel can be identified using standard diagnostic imaging techniques (MRI, CT and ultrasound) it has been shown that improved specificity for stroke prediction can be gained from identifying plaque features, such as intraplaque haemorrhage (IPH), that are associated with increased stroke risk4,5.
1.2 Quantitative Susceptibility Mapping
Quantitative susceptibility mapping (QSM) is an MRI method capable of spatially mapping the magnetic susceptibility of biological tissue6–8. Recently, a number of studies have applied QSM to carotid plaques and demonstrated marked improvement in IPH and calcification depiction compared to conventional multi-contrast MRI9–13 based on the inherent differences between the magnetic susceptibilities of these structures: calcifications are diamagnetic, whereas haemorrhages are paramagnetic14.
1.3 Arterial microstructure and QSM
Although QSM has proven promising for evaluating carotid plaques, specific disease driven sources of susceptibility changes have yet to be investigated in arterial vessels and plaques. Arterial tissue is predominantly composed of smooth muscle cells, elastin and collagen that are helically arranged to form the vessel microstructure15. The quantity, quality and organisation of this microstructure is finely tuned to maintain proper physiological function in healthy blood vessels16. Furthermore, arterial microstructure has been shown to become disrupted during the onset and progression of atherosclerosis and this changing microstructural arrangement may be indicative of mechanical stability, and therefore rupture risk, of advanced plaques17. As the magnetic susceptibility of a tissue is governed by its molecular makeup, QSM can provide insight into the microstructural composition of biological tissue. To date, QSM has been demonstrated to be sensitive to various tissue microstructural components such as myelin in brain18,19, collagen in liver20,21 and articular cartilage22–24, tubules in kidney25 and myofibers in heart26. As such, investigation of the sensitivity of QSM to arterial microstructure is warranted to determine if this is a source of susceptibility contrast that may provide an important biomarker of disease onset and progression in arteries.
1.4 Aim and summary
In this study we hypothesize that the magnetic susceptibility of arterial vessels is sensitive to changes in the vessel wall microstructure. To test this, a high resolution QSM protocol was applied to excised porcine carotid arteries subjected to a range of enzymatic digestion treatments27. To relate the magnetic susceptibility of porcine tissue to human arterial tissue, to accelerate potential clinical translation, the same QSM protocol was applied in ex vivo human carotid arteries. In both porcine and human vessels, detailed histological analysis was used to understand the underlying tissue microstructural composition and interpret the magnetic susceptibility of arterial vessels measured using QSM.
2. Methods
2.1 Sample preparation
2.1.1 Porcine carotid artery (microstructural) models
Carotid arteries were harvested from 6-month-old, Large White pigs within three hours of slaughter. All vessels were cleaned of connective tissue and cryopreserved using a protocol to preserve the integrity of tissue microstructure during freezing28. In brief, cryopreservation was performed at a controlled rate of -1°C/min to -80°C in tissue freezing media composed of Gibco RPMI 1640 Medium (21875034, BioSciences), sucrose (S0389, Sigma) and the cryoprotectant dimethylsulfoxide (PIER20688, VWR International). The use of a cryoprotectant prevented microstructural damage caused by the formation of ice crystals29. In preparation for imaging, vessels were thawed at 37°C and rinsed in phosphate buffered saline (PBS) (P5493, Sigma).
To investigate the sensitivity of QSM to arterial tissue microstructure, four different vessel models were developed with distinct microstructural compositions using porcine carotid arteries. All vessels were cryopreserved prior to treatment and imaged directly after treatment and before fixation. The native vessel model refers to porcine carotid arteries that were not subjected to any treatment and acted as a control tissue for comparison with the following models.
The collagen degraded model was produced by incubating native porcine carotid arteries in 1000 U/ml purified collagenase (CLSPA, Worthington Biochemical Corporation) in magnesium chloride and calcium chloride supplemented PBS (D8662, Sigma) at 37°C for 28 hours on a rotator at 10 rotations per minute.
The decellularised model was produced using an established protocol as follows30: using a pressure of 100 mmHg, 0.1 M sodium hydroxide (S8045, Sigma) was perfused through native porcine carotid arteries via a peristaltic pump at 2 Hz for 15 hours, followed by 0.1 M sodium chloride (S3014, Sigma) for 32 hours. Vessels were then treated with 10 µl/ml DNAase (LS006343, Worthington Biochemical Corporation) and 2 µl/ml primicin (Ant-pm-2, InvivoGen) at 37°C for 19 hours.
The elastin degraded model was produced by incubating native porcine carotid arteries in 10 U/ml purified elastase (ESFF, Worthington Biochemical Corporation) with 0.35 mg/ml trypsin inhibitor (10109886001, Sigma) in Dulbecco’s Modified Eagle Medium, high glucose, GlutaMAX (61965026, BioSciences) at 37°C for 3.5 hours.
2.1.2 Human carotid artery
Carotid arteries (including the carotid bifurcation and proximal sections of the common, internal and external carotid artery) were excised from five embalmed cadavers. One artery was obtained from each subject. The subjects (3 females and 2 males) ranged from 70 to 103 years in age (mean 81.6 ± 12.7 years). Cardiovascular disease was not implicated as the cause of death in any subjects. Vessels were cleaned of connective tissue and stored in PBS. To facilitate comparison of susceptibility values between human and porcine arteries, fixed native porcine carotid arteries were produced by immersing native porcine vessels in 4% formalin (HT501128, Sigma) for 7 days at 4°C. Prior to MR imaging all samples were washed and placed in fresh PBS.
2.2 MR imaging
2.2.1 Vessel positioning
All vessels were positioned using 3D-printed holders composed of polylactic acid and placed in 50-ml falcon tubes (Supporting Information Figure S1) in which samples were immersed in fresh PBS prior to imaging at room temperature. PBS was chosen as the “embedding” material as it has previously been identified as providing a stable experimental set-up facilitating good image quality for QSM of post-mortem brain specimens31. For the porcine vessels (native, decellularised, collagen degraded, elastin degraded and fixed native), six vessels (n=6) of each model were produced and holders were designed to secure six vessels in a single 50 ml falcon tube. For human vessels, holders were designed to hold a single specimen.
2.2.2 Image acquisition
A small bore (30 cm) horizontal 7T Bruker BioSpec 70/30 USR system (Bruker, Ettlingen Germany) equipped with a receive-only 8-channel surface array coil, birdcage design transmit coil, shielded gradients (maximum strength 770 mT/m) and Paravision 6 software was used for all imaging. For each session, a test tube containing the vessels of interest was placed securely in the cradle of the 8-channel surface array coil. A total of ten scan sessions were performed (five sessions for porcine vessels and five sessions for human vessels).
For QSM, data were acquired using a 3D multi-echo gradient echo (ME-GRE) sequence with the following parameters: TEs = 5, 13.1, 21.2 and 29.3 ms with monopolar readout gradients, TR = 150 ms, flip angle = 30°, bandwidth = 34722 Hz and averages = 2. The readout direction was oriented along the long axis of the tube. An isotropic voxel resolution of 0.117 x 0.117 x 0.117 mm3 was achieved using a field-of-view of 30 mm x 30 mm x 30 mm and a 256 x 256 x 256 matrix size. Total scan time for this sequence was 5 hours 27 minutes.
2.3 Image Processing
For all samples, the multi-channel ME-GRE data was coil-combined. Coil-combined magnitude images were calculated using the root sum of squares of the channels32 and coil-combined phase images were produced using the phase difference approach33.
2.3.1 QSM pipeline in porcine vessels
The processing pipeline used to produce quantitative susceptibility and relaxometry maps from the imaging data acquired in porcine vessels is summarised in Figure 1. R2* maps were calculated using the Auto-Regression on Linear Operations algorithm (ARLO)34 applied to the coil-combined ME-GRE magnitude data. To aid masking, an echo-combined magnitude image was calculated using the root sum of squares of all echoes. For QSM, an initial mask was created by thresholding the echo-combined magnitude image with the threshold set to include all vessels and PBS but to exclude the 3D-printed vessel holder and air outside the tube. The mask was manually refined to exclude air bubbles using the echo-combined magnitude and R2* map as references. Non-linear field fitting35,36 was used to estimate field maps from the complex ME-GRE data and remaining wraps were removed using Laplacian phase unwrapping37. This approach (non-linear field fitting followed by Laplacian phase unwrapping) provided a computationally efficient and robust approach to calculating unwrapped total field maps and has previously been applied to investigate tissue susceptibility in ex vivo articular cartilage22. Alternative approaches, such as linear field fitting, require prior unwrapping of individual echoes, which would lead to long computation times particularly for the large matrix sizes acquired here (256 x 256 x 256) whereas the approach used in this study requires unwrapping of only a single image. The local field map was calculated using the projection onto dipole fields (PDF) method and the unwrapped field map and magnitude mask as input38,39. PDF was chosen as it has been shown to perform well in comparison to alternative methods40. A susceptibility map was calculated from the local field map using an iterative Tikhonov method7,41,42 with correction for susceptibility underestimation43. The same regularisation parameter (α) was used for all samples and was chosen by performing L-curve optimisation44 in all five porcine datasets and calculating the mean of the individually optimised parameters. Alternative susceptibility calculation methods (TKD45, direct Tikhonov7,41,46 and MEDI47) were tested and found not to influence the trends and final conclusions reported for porcine vessels (see Supporting Information Figures S9 and S10).
2.3.2 QSM pipeline in human vessels
All five human vessels scanned in this study had regions of advanced atherosclerotic disease close to the bifurcation (see Figure 7 yellow arrows). These heavily diseased regions contained structures with little or no signal that caused significant streaking in the final susceptibility maps likely attributable to the presence of calcification or haemorrhage. To investigate the impact of these low signal-to-noise-ratio (SNR) regions on the susceptibility measurements made in regions of the common carotid unaffected by disease/plaque, two different masking procedures were compared for QSM (Figure 2).
2.3.2.1 Comparison of masking procedures for human vessels
Firstly, a tube mask was generated for each human vessel that included heavily diseased low SNR regions. This mask was generated by thresholding the echo-combined magnitude image to exclude the 3D-printed vessel holder and air outside the tube and include all other contents of the tube, i.e. PBS and the whole vessel (including diseased regions). No manual refinement step was implemented here due to the difficulty in distinguishing diseased regions from air bubbles.
Secondly, a noise mask was generated for each human vessel that excluded heavily diseased low SNR regions. High noise regions in areas of advanced atherosclerotic disease were identified by thresholding the inverse noise map (calculated during non-linear field fitting36 using noise propagation46) at one third of the mean voxel value contained within the tube mask. To generate the final noise mask, the regions of high noise identified by the inverse noise map were removed from the tube mask (see Figures 2 and 8).
2.3.2.2 QSM calculation in human vessels
Masking procedures were compared for final susceptibility map calculation through comparison of tube mask and noise mask QSM pipelines. For both pipelines, field maps were generated using non-linear field fitting35,36 and unwrapped using Laplacian phase unwrapping37. At this point, the pipelines diverge as background field removal and susceptibility calculation steps require masks identifying the region of interest. Local field maps were calculated with the PDF method34 using the unwrapped field map and a region of interest mask as input i.e. tube mask for the tube mask pipeline and noise mask for the noise mask pipeline. Using the relevant local field map and mask as inputs, susceptibility maps were calculated for the tube mask and noise mask pipelines using the iterative Tikhonov approach as described for the porcine arteries. The same regularisation parameter (α) was used for all samples and was chosen by performing L-curve optimisation44 for both tube mask and noise mask pipelines in all five human datasets and calculating the mean of the individually optimised parameters. Similar to the porcine vessels, alternative susceptibility calculation methods (TKD, direct Tikhonov and MEDI) were tested and found not to influence the trends and final conclusions reported for human vessels (see Supporting Information Figure S10).
2.4 Histological Analysis
To allow histological validation, the model tissues (native, decellularised, collagen degraded and elastin degraded vessel models) were fixed immediately after MR scanning for histological processing by immersing vessels in 4% formalin for seven days at 4°C. This was unnecessary for fixed native porcine and human vessels as fixation was performed prior to scanning. Stepwise dehydration was performed on the fixed tissue in ethanol to xylene and samples were then embedded in paraffin wax and sectioned into 8 µm thick slices prior to staining.
For porcine and human vessels, Haematoxylin and Eosin (H&E), Picrosirius red, Verhoeff’s elastin and Alcian blue staining were performed to identify the presence of smooth muscle cells, collagen, elastin and glycosaminoglycans (GAGs) respectively27. Additionally, Alizarin Red staining was performed on human vessels to identify the presence of calcium. Histological imaging was performed using an Olympus BX41 microscope with Ocular V2.0 software for the porcine models and an Aperio CS2 microscope with ImageScope software V12.3 for the human arteries. Bright-field microscopy was performed for all stains with additional polarised light microscopy (PLM) on Picrosirius red to infer the orientation of collagen fibres.
For human vessels, alignment of histology to QSM was performed as follows; using anatomical features as landmarks, the first echo of the magnitude combined ME-GRE image was used to guide the identification of tissue sections in the common carotid for histological analysis. The location of the selected region on the ME-GRE image was noted and used to guide the manual alignment of the digitised whole-mount histology image to the relevant MRI slice. The final alignment of histology to MRI was refined via manual rotation and scaling of the histology image (see Figure 4 and Supporting Information Figures S2-S6)
2.5 Data Analysis
2.5.1 Regions of interest
For both human and porcine tissue, regions of interest (ROIs) were manually defined for each vessel using the echo-combined MRI magnitude image. Care was taken to position ROIs within arterial tissue and avoid regions of partial volume near the edges of each vessel. This was facilitated by the high resolution of the ME-GRE data. ROIs in human vessels were limited to normal appearing arterial tissue in the common carotid artery. This was achieved by cross referencing the MRI defined ROI with histology and excluding regions from the final ROI that showed abnormal smooth muscle cell, collagen, elastin, GAG or calcium content as identified by histology and defined by Stary et al.48 (see Figure 4 and Supporting Information Figures S2-S6). Using the echo-combined magnitude image, ROIs were manually defined for PBS in each sample. Following ROI definition, ROI mean susceptibility values were extracted for each vessel and for PBS in each sample. To compare across samples, vessel susceptibility was referenced to PBS49. This was achieved by subtracting the mean PBS susceptibility from the mean vessel susceptibility in the same sample.
2.5.2 Statistical methods
For measurements made in porcine vessels, the null hypothesis of no susceptibility difference between tissue models (native, decellularised, collagen degraded, elastin degraded and fixed native) was tested using a one-way ANOVA. If the null hypothesis was rejected, post hoc pairwise comparisons were performed using the Tukey-Kramer method. To compare the susceptibility measurements made using tube mask and noise mask pipelines in human common carotid arteries, the intraclass correlation coefficient (ICC)50 was calculated. For comparison of susceptibility measurements made in human common carotid arteries with those in fixed native porcine arteries, the null hypothesis of no susceptibility difference between groups (human commontube mask, human commonnoise mask and fixed native porcine) was tested using a one-way ANOVA. If the null hypothesis was rejected, post hoc pairwise comparisons were performed using the Tukey-Kramer method.
3. Results
3.1 Histology
3.1.1 Histological validation of tissue models
Figure 3 presents histological validation of the vessel models where enzymatic digestive treatments were used to selectively remove smooth muscle cells, elastin and collagen in selected groups of porcine carotid arteries. H&E staining confirms the absence of smooth muscle cells in the decellularised vessels (Figure 3 d), Verhoeff’s elastin stain confirms the degradation of elastin in the elastin degraded vessels (Figure 3 g) and Picrosirius red verifies the removal of collagen in the collagen degraded vessels (Figure 3 n and r). Furthermore, H&E staining (Figure 3 a-d) verifies that smooth muscle cell content remained intact in the non-decellularised vessels, Verhoeff’s elastin staining (Figure 3 e-h) verifies that elastin content was maintained in the non-elastin degraded vessels and Picrosirius red staining with bright field microscopy (Figure 3 m-p) and PLM (Figure 3 q-t) verifies the preservation of collagen content and orientation in the non-collagen degraded vessels. Alcian blue staining (Figure 3 i-l) showed a decrease in GAG content across all the degradation models when compared with native tissue. This may be explained by the leaching of GAGs out of the tissue in order to maintain osmotic balance due to the presence of PBS. Although this hasn’t been observed in arterial tissue it has been observed in intervertebral disc and articular cartilage51.
3.1.2 Histological validation of ROIs in cadaver vessels
Figure 4 shows example H&E in human common carotid artery for a representative cadaver specimen imaged in this study. Supporting Information Figures S2-S6 display histological analysis and ROI definition for each human vessel. Histology was used to guide the definition of MRI ROIs in “normal” appearing common carotid tissue by avoiding regions of obvious abnormality or disease defined as a disruption to cell, collagen, GAG, elastin or calcium content.
3.2 MRI
3.2.1 QSM of tissue models
Figure 5 presents susceptibility and R2* relaxometry maps for each of the vessel models. Susceptibility maps produced using the iterative Tikhonov method are presented in three different viewing planes (sagittal, coronal and axial in the context of the tube which was placed with its long axis parallel to the orientation of B0) for qualitative comparison of image quality. The regularisation parameter used for susceptibility map calculation in porcine vessels was α=0.02 as determined by taking the mean of the L-curve optimisation44 across the five samples (Supporting Information Figures S7).
Qualitative differences are apparent when comparing the susceptibility maps of different vessel models. The susceptibility of native (Figure 5 a) vessels appear the most diamagnetic with the susceptibility of decellularised (Figure 5 b) and collagen degraded (Figure 5 c) vessels appearing elevated in comparison while maintaining contrast with the background fluid. The susceptibility of the elastin degraded vessels (Figure 5 d) appear the most elevated with almost no susceptibility difference with the surrounding fluid. Qualitatively, R2* relaxometry maps demonstrated similar but inverse trends to the susceptibility maps with R2* of native vessels (Figure 5 e) appearing highest and R2* of elastin degraded vessels (Figure 5 h) appearing lowest and similar to that of surrounding background fluid.
As a quantitative comparison, Figure 6 shows boxplots of the susceptibility and R2* values measured in each vessel (n=6 per model) and grouped by tissue model. The mean (± standard deviation) susceptibility of PBS measured across the five samples was 0.0219 ± 0.0121 ppm. Susceptibility and R2* measurements in fixed native porcine carotid artery (n=6) are presented for later comparison with human vessels. The baseline magnetic susceptibility of native porcine arteries was found to be diamagnetic with a mean value of −0.1820 ppm and vessel susceptibilities ranged from −0.2346 ppm to -0.0092 ppm across all tissue models. The mean susceptibility of the tissue models (𝒳elastin degraded = −0.0163 ppm; 𝒳collagen degraded = −0.1158 ppm; 𝒳decellularised = −0.1379 ppm; 𝒳fixed native = −0.2199 ppm) exhibited the following trend: 𝒳elastin degraded > 𝒳collagen degraded > 𝒳decellularised > 𝒳native > 𝒳fixed native. Significant differences were detected between the susceptibility of the vessel models (ANOVA, p<0.001). Post hoc pairwise comparisons revealed the susceptibility of collagen degraded vessels to be significantly higher than native vessels (p<0.01), the susceptibility of elastin degraded vessels to be significantly higher than all other vessel groups (p<0.001) and the susceptibility of fixed native vessels to be significantly lower than both decellularised (p<0.01) and collagen degraded vessels (p<0.001). Comparing vessel-wise measurements of R2*, the null hypothesis was rejected (p<0.001) indicating differences between groups. Post hoc testing revealed significantly lower R2* in elastin degraded vessels compared to all other vessel models (p<0.001).
3.2.2 QSM of human common carotids
In the human carotid arteries, regions of advanced disease were seen close to the bifurcation in all five vessels scanned in this study (Figure 7 yellow arrows). These heavily diseased regions contained structures with little or no signal – likely attributable to the presence of calcification or haemorrhage. For a representative vessel, susceptibility maps are presented for tube mask and noise mask pipelines, with the region where the ROI was defined, indicated in red (Figure 8). It can be seen that human carotid vessels exhibited pronounced streaking artefacts attributed to the inclusion of low SNR regions in the QSM calculation (tube mask). Streaking was visibly reduced in the susceptibility maps produced using the noise mask pipeline (Figure 8). The optimal regularisation parameter was determined as α=0.02 using L-curve optimisation, remaining the same as that used for the porcine vessels (Supporting Information Figures S7).
Figure 9 shows susceptibility measurements in human common carotid arteries. The mean susceptibility of human common carotid was −0.1898 ± 0.0253 ppm using the noise mask pipeline and −0.2007 ± 0.0183 ppm using the tube mask pipeline. This is in comparison to the fixed native porcine carotid arteries which had a mean susceptibility of −0.2199 ± 0.0100 ppm. Figure 9 a directly compares common carotid susceptibility measurements between the noise mask and tube mask pipelines with an ICC value of 0.88 (p<0.05) being obtained between the pipelines. Figure 9 b displays box-plots of the mean susceptibility in each vessel using the noise mask and tube mask pipelines compared with those in fixed native porcine carotid arteries. The null hypothesis was not rejected (p>0.05), however the p-value of 0.050 was close to the significance level of 0.05.
4. Discussion
4.1 Main findings
In this study we demonstrate for the first time the sensitivity of magnetic susceptibility, measured using QSM, to the microstructural composition of arterial tissue. Vessels with different microstructural compositions were generated by applying different enzymatic digestion treatments to ex vivo porcine carotid arteries. For each model (decellularised, collagen degraded and elastin degraded), six vessels were imaged using a high resolution QSM protocol and compared to untreated (native) and fixed native porcine carotid arteries (Figure 5). The microstructural composition of each vessel model was validated using histology (Figure 3) and statistically significant differences were found between the group mean susceptibility of porcine vessels with different microstructural compositions (Figure 6). Post hoc statistical testing revealed significantly higher susceptibility in collagen degraded vessels compared to native vessels, while no such difference was present in equivalent R2* relaxometry measurements. Significantly lower susceptibility was measured in fixed native vessels compared to both decellularised and collagen degraded vessels with no such statistical differences existing in the equivalent R2* measurements. This suggests that QSM offers improved sensitivity to the microstructural composition of arterial vessels, in particular collagen, when compared to equivalent R2* measurements.
To provide a comparison with human tissue, susceptibility was measured in ex vivo human common carotid arteries using the same high resolution QSM pipeline (Figure 7). Regions of advanced disease were present in these vessels leading to streaking artefacts in the susceptibility maps. Exclusion of these low SNR regions resulted in a qualitative improvement in the susceptibility maps (Figure 8). However, using ICC “excellent” agreement52 was found between human common carotid measurements extracted from susceptibility maps that were calculated using pipelines which included and excluded the low SNR diseased regions (Figure 9 a). One explanation for this agreement is that the common carotid is positioned far enough away from the streaks (originating in the bifurcation) that they have very little effect on the mean susceptibility measured in that region. An example of this can be seen in Figure 8. Statistical testing revealed no significant differences between magnetic susceptibilities measured in fixed human common carotid arteries and fixed porcine vessels (Figure 9 b).
4.2 Susceptibility and arterial microstructure
Experimental results from porcine vessel models (Figure 6) suggest tissue susceptibility, measured using QSM, is sensitive to the microstructural composition of arterial vessels. The baseline magnetic susceptibility of native porcine arteries was found to be diamagnetic with a mean value of −0.1820 ppm and agrees well with the value of −0.25 ± 0.14 ppm reported for popliteal artery wall in vivo24.
4.2.1 Collagen
Compared to native vessels, a significantly higher susceptibility was measured in collagen degraded vessels (Δ𝒳 = 0.0662 ppm) while no such significant difference was apparent in the equivalent R2* relaxometry measurements. This agrees well with studies of collagen susceptibility in articular cartilage24,53 and liver fibrosis21 which observed collagen as strongly diamagnetic. Although articular cartilage degradation associated with collagen loss has been shown to cause changes in measured susceptibility in vivo54, ex vivo studies of articular cartilage failed to detect differences in the susceptibility of collagen degraded samples22. The differences in the results found in ex vivo collagen degraded articular cartilage and arterial vessels may well be explained by differences in experimental set-up, susceptibility reference material and enzymatic digestion treatments. It is important to note that collagen was completely removed from the samples imaged in this study (Figure 3 n & r) and PBS was used as a consistent susceptibility reference between samples.
4.2.2 Elastin
Significant differences were seen between the measured susceptibility of elastin degraded vessels and all other vessel models including native and fixed native vessels. The elastin degraded group also demonstrated the largest deviation in measured susceptibility from native vessels. However, as evidenced histologically by H&E staining (Figure 3 a & c) the removal of elastin results in a less compact microstructural arrangement of tissue, when compared to native vessel histology. This suggests that the removal of elastin results in an increased extracellular space, allowing the penetration of surrounding PBS into the tissue microstructure. This is supported by the visible increase in size of the vessels and the distinct lack of contrast seen between elastin degraded vessels and the surrounding PBS in QSM (Figure 5 d). Therefore, it is difficult to conclusively assess the contribution of elastin to vessel susceptibility from the tissue model presented here.
4.2.3 Smooth muscle cells
Considerable overlap is seen between the measured susceptibilities of decellularised and native vessels. The slightly higher group mean susceptibility of decellularised vessels was not significantly different from that of native vessels (Figure 6).
The results from all these porcine artery models suggest that QSM is most sensitive to detecting changes in arterial collagen. Further work is required to investigate the integrity of the elastin model and the resulting abolition, following elastin degradation, of the diamagnetic susceptibility found in native porcine arteries. An alternative approach to characterising the magnetic susceptibility of elastin is to use 3D printed tissue scaffolds where the density of elastin can be tightly controlled55.
4.2.4 Susceptibility anisotropy
Although this study has demonstrated the sensitivity of QSM to arterial microstructure, specific components, such as collagen, are known to possess B0 orientation-dependent magnetic susceptibility21–23. In this study, all vessels were imaged at the same orientation to the main magnetic field (long axis of the vessel parallel to B0). As arterial collagen and elastin fibres are circumferentially arranged, these fibres will be oriented at 90° to B0, facilitating consistent comparison of measured susceptibility within and between vessels. As QSM assumes isotropic susceptibility, further work is required to assess the susceptibility anisotropy of arterial tissue and the anisotropic contributions of its microstructural components, in particular collagen53.
4.3 Human artery
4.3.1 Comparison of human and porcine vessels
From Figure 9 b, variability of vessel susceptibility across the group is visibly lower for porcine vessels compared to the human samples. This may be partially attributed to the low variability in the age of the pigs compared to the human subjects, but also to differences between fixation and embalming, respectively. Porcine vessels were imaged directly after fixation whereas time from embalming was not controlled for in human vessels. Evia et al. reported no systematic change in magnetic susceptibility of fixed post-mortem brain over a period of 6 weeks56 but changes in susceptibility outside this timeframe may be possible. Although specific information regarding time from embalming to scanning is not available for the human vessels, imaging of all vessels included in this study occurred after this 6-week time frame. The variability in susceptibility seen between human vessels could also be due to biological variability (e.g. age, sex, body size, disease status), as well as its effect on the embalming protocol.
Despite this variability there was no statistically significant difference measured between the magnetic susceptibility of human and porcine arterial tissue. This suggests that the sensitivity of magnetic susceptibility to microstructural composition, demonstrated in porcine arterial tissue, is also likely to be found in human arterial tissue. This provides a promising springboard for studies seeking to translate QSM for use in characterising human carotid arteries and disease.
4.3.2 Fixed porcine tissue
The group mean susceptibility was slightly lower in fixed tissue (𝒳 = −0.2199 ppm) compared to native vessels (𝒳 = −0.1820 ppm), although this difference was not statistically significant (Figure 6). As noted by Wei et al.21, fixation alters tissue microstructure through crosslinking of proteins and differences in measured susceptibility are not surprising. Differences in susceptibility between in vivo and fixed ex vivo mouse brain have been reported57 and it has been noted that changes in tissue relaxation times that accompany fixation could impact QSM measurements of susceptibility26.
4.4 Conclusion, impact and future work
In diseased arterial tissue, key components of the tissue microstructure, such as collagen fibres, smooth muscle cells and elastic lamina become disrupted, impairing vessel function58. A number of recent studies have demonstrated the application of QSM for imaging carotid plaque in vivo, highlighting that many technical challenges associated with imaging carotid arteries using QSM in a clinical setting can be overcome9–13. These studies have demonstrated marked improvement in depiction of IPH and calcification in vivo but have largely ignored differences in regional plaque susceptibility that may be driven by compositional variations in the microstructure of fibrotic plaque tissue. Results from this study highlight that QSM is sensitive to the microstructural composition of arterial tissue and, with further development, has the potential to offer unique insight into the onset and progression of carotid atherosclerosis. Such characterisation of carotid plaques has the potential to improve the assessment of stroke risk using MRI59 and could complement existing MRI methods capable of detecting downstream haemodynamic alterations60,61. Future work will focus on QSM of diseased arterial tissue ex vivo, using the insights from this study as a basis to fully characterise the susceptibility contributions from IPH, calcifications, lipid and tissue microstructure in heterogenous atherosclerotic plaques.
Data Availability Statement
The imaging data for this study is available from the corresponding author on reasonable request. Tools used for QSM calculation are openly available through the MEDI toolbox (http://weill.cornell.edu/mri/pages/qsm.html) and UCL’s XIP repository (https://xip.uclb.com/i/software/mri_qsm_tkd.html)
Supporting Information
Acknowledgements
The authors would like to thank and acknowledge the Department of Anatomy, Royal College of Surgeons in Ireland (Professor Clive Lee and Bob Dalchan) for supporting this work