Quantitative susceptibility mapping of the human carotid artery: Assessing sensitivity to elastin and collagen ex vivo

Abstract

Purpose

The aim is to establish the relationship between carotid susceptibility and microstructural components in diseased carotid arteries.

Methods

Excised cadaveric carotid arteries (n = 5) were scanned using high‐resolution QSM at 7 Tesla. After ex vivo imaging, all samples were brought to histology and stained for elastin, collagen, cells, and calcium. An image registration pipeline was used in combination with semi‐quantitative, regional histology analysis to evaluate relationships between MRI and microstructural components.

Results

Weak, non‐significant (p > 0.05) correlations were found between all components and regional magnitude and R₂* measurements. A significant, moderate negative correlation between the elastin fraction and regional magnetic susceptibility, r _elastin = −0.63 (p < 0.0001) was found, as well as a significant, moderate negative correlation between collagen and regional magnetic susceptibility, r _collagen = −0.59 (p < 0.0001).

Conclusion

Tissue magnetic susceptibility in diseased human carotid arteries was shown to be significantly correlated with the dominant microstructural components of pathological human cadaver samples—elastin and collagen. Knowing that elastin and collagen are disrupted in vascular disease progression, QSM offers clinically translatable potential for novel disease biomarkers.

1. INTRODUCTION

Although predominantly used in the brain, applying MRI QSM ¹ , ² to the rest of the body is gaining increasing interest because of its ability to elucidate differences in magnetic susceptibility among/between heterogeneous tissues. ³ In particular, carotid QSM can distinguish between calcification and intra‐plaque hemorrhage (IPH) in atherosclerotic carotid plaques in vivo. ⁴ , ⁵ Current best practice for the management of atherosclerotic plaques “at risk of rupture” is carotid endarterectomy surgery, where the plaque is excised to avoid the risk of stroke. ⁶ However, this requires the identification of plaques “at risk of rupture,” which are typically identified by measuring the percentage of carotid stenosis. ⁷ , ⁸ , ⁹ This percentage is not specific to any rupture‐related properties of the plaque, and it is known that many non‐stenotic plaques are symptomatic. ⁹ , ¹⁰ Plaque composition, organization, and microstructure have all been shown to link to the mechanics and stability of plaques, and therefore, may provide improved clinical indicators of rupture risk. ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ Although calcifications and IPH have been linked to plaque vulnerability, ¹⁶ so too have changes in arterial microstructure. ¹⁷ The ability of QSM to differentiate between gross features such as calcification and IPH has been demonstrated, ⁴ , ⁵ , ¹⁸ but the link to arterial microstructure is still under investigation. ¹⁹ Observing changes in the amount and organization of key mechanically sensitive microstructural components—elastin, collagen—has the potential to offer early insight into plaque vulnerability. ²⁰ , ²¹

Ex vivo MRI offers several advantages for investigating arterial tissue microstructure: high resolution, no motion, and few acquisition time limitations. Previous work has shown that the magnetic susceptibility of arterial tissue is significantly influenced by the presence of both collagen and elastin in porcine carotid arteries ex vivo. ¹⁹ The removal of entire microstructural components allowed for direct insight into which components contribute toward the magnetic susceptibility measured by QSM in arterial tissue. In particular, it was shown how, within the arterial wall, an absence of collagen contributed to a significant increase in susceptibility compared to the more diamagnetic susceptibility in healthy arteries. ¹⁹ However, the relationship between elastin and arterial tissue susceptibility is still poorly understood. Elastin is responsible for maintaining the integrity of the arterial wall and with its removal there is a significant increase in extracellular space. ¹⁹ , ²² This increase in extracellular space previously confounded QSM measurements and makes it difficult to establish the true influence of elastin on carotid susceptibility. ¹⁹ In this work, we aim to address this outstanding question by using QSM to image ex vivo human cadaveric carotid arteries, which all presented varying degrees of atherosclerosis, rather than the previously used animal tissue models. The use of pathological human tissue provides a broad range of microstructural tissue compositions in true disease morphologies.

To do this, we implement a previously established pipeline based on image registration, regional analysis, and semi‐quantitative histology. ²³ Regional analyses can take many forms, for example: manual segmentation, thresholding, or classifications. Previous work on atherosclerotic plaques has used k‐means clustering and found this method to be a robust unsupervised classification method. ²⁴ Clustering is simple to implement, efficient, and offers an objective way to define regions of interest (ROI). Through co‐registration of MRI and histology images, clustering allows for comparisons between regions with similar attributes. Quantification of histopathology has been transforming tissue biomarker and diagnostic discovery ²⁵ as the desire for more quantitative disease measures increases. Previous studies have investigated MRI derived metrics in the context of semi‐quantitative histology, which allows for a better understanding of how the underlying tissue microstructure drives signal changes. ²³ , ²⁶ , ²⁷ This type of fundamental knowledge can provide improved clinical insight into pathological changes in tissue and ultimately offer novel imaging biomarkers.

Therefore, our aim was to establish the relationship between carotid magnetic susceptibility, measured using QSM, and the microstructural composition of diseased human carotid arteries. This was achieved by imaging excised carotid arteries from human cadavers using a high‐resolution QSM protocol and comparing to a detailed histological work‐up to identify their microstructural composition. Together, this allowed clarification of the key drivers of magnetic susceptibility in arterial tissue and yielded novel insights for further applications of carotid QSM.

2. METHODS

2.1. Human cadaveric carotids

Carotid arteries (n = 5) were excised from five individual embalmed cadavers. Each carotid sample included the common, external, and internal carotid arteries and the bifurcation, see Figure 1. Cadaveric tissue was provided by The Royal College of Surgeons in Ireland and was used with approval from the Department of Anatomy, Royal College of Surgeons in Ireland institutional review board. Of the five cadavers, three were female and two male. They ranged from 70 to 103 years in age (mean, 81.6 ± 12.6). The cause of death for all was unrelated to cardiovascular disease; however, no information on risk factors, past medical history, or other cardiovascular related diseases was available. On excision from the embalmed cadavers, the fixed carotid arteries were cleaned of excess connective tissue and stored in phosphate buffered saline (PBS) at 4°C until imaging (<1 year).

FIGURE 1.

Cadaveric carotid arteries used in this study. (A) Imaging set‐up overview. (B–F) Five carotids were excised from cadavers and whole carotids were imaged. Sections of the common carotid (inset images) were histologically processed for registration to MRI data. Scale bars of whole carotids are 10 mm, scale bars in inset images are 5 mm.

2.2. MR imaging

All carotids were secured, individually, in 50‐mL falcon tubes using PLA 3D‐printed holders to position them. Samples were immersed in fresh PBS before imaging at room temperature. All samples were positioned such that the longitudinal axis of the artery was positioned parallel to B₀ (Figure 1A). A small‐bore (30 cm) horizontal 7 Tesla Bruker BioSpec 70/30 USR system (Bruker) equipped with a receive‐only 8‐channel surface array coil, birdcage transmit coil, and shielded gradients (maximum strength, 770 mT/m) was used. Paravision 6 (Bruker) software was used for all acquisitions. A total of five scan sessions were performed (one per sample). For QSM, data were acquired using a high‐resolution 3D multi‐echo gradient‐echo (ME‐GRE) sequence that was previously used to investigate arterial tissue. ¹⁹ Acquisition parameters were: TEs = 5, 13.1, 21.2, 29.3 ms with monopolar readout gradients, TR = 150 ms, flip angle = 30°, bandwidth = 34 722 Hz, and 2 averages. The readout direction was oriented along the longitudinal axis of the tube and artery. Using a FOV of 30 × 30  × 30 mm and a matrix size of 256 × 256 × 256, an isotropic voxel resolution of 0.117 × 0.117 × 0.117 mm³ was achieved. Total scan time for this sequence was 5 h and 27 min.

2.3. QSM analysis

QSM maps were calculated for each of the carotid arteries using a previously optimized pipeline for porcine carotid arteries. ¹⁹ For each sample, coil‐combined magnitude images were calculated using the RMS of the channels ²⁸ and coil‐combined phase images were produced using the phase difference approach. ²⁹ R₂* maps were calculated using the auto‐regression on linear operations algorithm on the coil‐combined magnitude data. ³⁰ For masking, an echo‐combined magnitude image was calculated using the RMS of all echoes and then thresholded to include all tissue and PBS and exclude the 3D‐printed holder and air outside the tube. This mask was manually refined to exclude air bubbles. Field maps were generated using nonlinear field fitting ³¹ and unwrapped with Laplacian phase unwrapping. ³² The projection‐onto‐dipole‐fields method ³³ was used to calculate local field maps using the unwrapped field maps and previously created mask as inputs. Susceptibility maps were calculated using an iterative Tikhonov approach ³³ , ³⁴ , ³⁵ where the regularization parameter (α = 0.02) was determined via L‐curve optimization as previously described in Stone et al. ¹⁹ Susceptibility values were referenced to the average susceptibility value in the masked region and could also have been referenced to the smaller PBS region, although this had no effect on the results (not shown). The mean and SD of susceptibility values in the masked region and in the PBS are shown in Table S2 in accordance with the QSM consensus recommendations. ³⁶ The computed echo‐combined magnitude images were used for all correlations and figures in the study. Magnitude images and R₂* and susceptibility maps are presented with tissue masks, please refer to Figure S1 for the unmasked images and maps.

2.4. Histological analysis

2.4.1. Histology processing

After MR scanning, the carotids were transferred to histological processing. For each carotid, a 2–3 mm long section of the common artery was selected and processed for histological staining as previously reported. ²³ Briefly, all samples were stepwise dehydrated and embedded in paraffin wax. Samples were sectioned into 7‐μm thick slices and individual, but consecutive, slices were then stained with hematoxylin and eosin (H&E) (cell content), Verhoeff's elastin (elastin), Picrosirius red (PSR) (collagen), and Alizarin red (calcium).

2.4.2. Semi‐quantitative histology

A previously established ²³ semi‐quantitative histological process was carried out to determine the amount of each microstructural component (cells, collagen, elastin, and calcium) in the tissue. QuPath software ²⁵ (version 0.2.3) was used to quantify elastin and calcium. To do this, a stain vector was set for positive staining of elastin and calcium, ³⁷ a tissue area was delineated by thresholding it from the background, and then a second threshold was used to identify the positively stained areas. The brightfield (BF) and the stained area images were exported from QuPath to MATLAB, where the stained area fraction (SAF) was calculated. Collagen content quantification (collagen fraction) followed a similar pipeline using an in‐house developed MATLAB code, which uses polarized light microscopy (PLM) to define the collagen areas and PSR‐stained BF images to define the tissue area. ³⁸ For cell density analysis, ROIs (see 2.5.2) were manually drawn within QuPath, where the cell detection function was used to yield 2D cell densities.

2.5. Data analysis

2.5.1. Image registration

For each sample, one histology slice per stain from the common carotid was registered to one MRI slice using a previously established method. ²³ Therefore, four consecutive histology slices were registered per sample. This was performed as follows: using anatomical landmarks visible in the MRI data (the apex of the bifurcation, the base of the common carotid attached to the 3D printed holder and features such as calcifications or branches), histology slices were localized to a single MRI slice in each sample. MRI tissue masks were defined by manually segmenting the vessel from the embedding media (i.e., background PBS) using the echo‐combined magnitude images. For registration, the BF and SAF images exported from QuPath were imported to MATLAB and masked. The BF histology images were then manually aligned to the 2D MRI image by rotating and scaling to provide an approximate alignment (pre‐registered images). The manually aligned BF histology images served as the starting point for subsequent transformations. Elastix (version 5.0.1) was used to perform nonrigid registration of the manually aligned BF histology images to the 2D MRI image. ³⁹ , ⁴⁰ , ⁴¹ B‐spline registration was used with an adaptive gradient descent optimizer and a maximum of 4000 iterations. The maximum step length was 0.1 and the B‐spline interpolation order was three. Finally, these registrations were applied to the SAF images to align them with the selected MRI slices.

2.5.2. Regional analysis

To create objective ROIs, k‐means clustering of the MRI data was performed. K‐means clustering was implemented in MATLAB using the echo‐combined magnitude images, R₂* maps and the susceptibility maps as inputs. As such, the resulting ROIs were evenly weighted toward the three MRI contrasts (magnitude, R₂*, and QSM). The elbow method was used to determine the appropriate number of clusters, ⁴² with 10 being determined as appropriate (k = 10). Nine of the clusters were in vessel tissue with the remaining cluster being the embedding media (PBS). Once the regions were defined, the mean magnitude, R₂* value, and susceptibility were calculated for each of the nine clusters (ROIs) located in the tissue. The same ROIs were applied to the SAF images that were registered to MRI space. This yielded mean values for elastin fraction, collagen fraction, cell density, and calcium fraction for each ROI.

2.5.3. Statistical method

To investigate the relationship between the MRI data and the microstructural components of arterial tissue, Pearson's correlation coefficients (r) were calculated, in line with previous studies, ⁴³ , ⁴⁴ between MRI (echo‐combined magnitude image, R₂*, and susceptibility) and the semi‐quantitative histology (elastin fraction, collagen fraction, cell density, and calcium fraction). Correlation coefficients below 0.3 were considered weak, 0.3 to 0.7 moderate, and above 0.7 strong, ⁴⁵ and α was defined at 0.05.

3. RESULTS

3.1. Image registration pipeline

Image registration and k‐means clustering allowed for direct correlations between MRI‐derived measures and microstructural component quantifications from histology (Figure 2).

FIGURE 2.

Example of image registration and cluster analysis from an example cadaveric carotid sample. Left to right: Verhoeff's elastin‐stained brightfield (BF) and stained area fraction (SAF) histology slice before non‐linear image registration; MRI echo‐combined magnitude, R₂*, and QSM in the same slice (masked to tissue compartment only); regions of interest (ROIs) determined by k‐means clustering; Verhoeff's BF and SAF post image registration.

3.2. QSM of cadaver carotids

Sagittal, coronal, and axial views of magnitude, R₂*, and susceptibility maps are shown to highlight different regions of the carotid artery (Figure 3). Although magnitude images show varying contrast throughout the vessel wall and plaque area, it is of interest that only the susceptibility maps differentiate low signal areas further into more diamagnetic or paramagnetic regions. In the axial views of the samples, sample 1 displays the most advanced regions of disease in the artery; however, heterogeneous contrast can be seen across cross‐sections of samples 2, 3, and 5 as well (Figure 4). The regional variations in contrast differs across magnitude, R₂*, and susceptibility maps.

FIGURE 3.

MRI maps of an example excised human carotid artery in three planes. Tissue‐masked magnitude images, R₂* maps, and susceptibility maps. FOV is 30 × 30 × 30 mm. The common carotid is marked by a red asterisk, the internal by a yellow asterisk, and the external by a green asterisk.

FIGURE 4.

A central, axial slice of magnitude images, R₂* maps, and susceptibility maps within the common carotid for each sample. Images and maps are masked to the tissue compartment.

3.3. Semi‐quantitative histology

Variations in PSR (collagen) intensity as well as Verhoeff's (elastin) density and intensity can be seen spatially in most samples (Figure 5). H&E highlights a clear lipid‐core in sample 1 as well as a thickened intima in the rest of the samples. Alizarin red (calcium) staining is overall very faint, although small high‐intensity spots can be seen in sample 1 and 3.

FIGURE 5.

Pre‐registered histological slices; one slice per stain. Scale bars are 1 mm.

Correlations between the different microstructural components (elastin, collagen, cell density, and calcium) and MRI (magnitude, R₂*, and susceptibility) are presented (Figure 6). Weak, non‐significant (p > 0.05), correlations were found between all microstructural components and regional R₂* (r _elastin = −0.14, r _collagen = 0.069, r _cell = −0.034, and r _calcium = 0.19) and magnitude measurements (r _elastin = 0.13, r _collagen = 0.28, r _cell = −0.26, and r _calcium = −0.19). A significant, moderate negative correlation between elastin fraction and susceptibility, r _elastin = −0.63 (p < 0.0001), is presented. All samples have a significant correlation, with four of the five having strong negative correlations (Table 1). There is also a significant, moderate negative correlation between collagen and susceptibility, r _collagen = −0.59 (p < 0.0001). It is worth noting that four of the five samples (2–5) individually have significant, strong correlations with susceptibility (Table 1). There is no significant correlation between tissue susceptibility and cell density (r _cells = 0.0063). Overall, there is no correlation between susceptibility and calcium (r _calcium = −0.061). Although, individually, sample 4 has a significant correlation with susceptibility and samples 2 and 5 have significant correlations with magnitude.

FIGURE 6.

Pearson's correlations between MRI data and microstructural tissue components. Susceptibility values are presented in the top row, R₂* in the middle, and magnitude values and their correlation with components are given in the bottom row. The first column is elastin, the second is collagen, the third is cell density, and the last is calcium. Different colored dots correspond to different samples (n = 5) and each individual dot corresponds to a singular region of interest (ROI) as defined by k‐means clustering (n = 9 per sample). Pearson's correlations were performed and presented for the entire dataset. Pearson's coefficients (r) and p values are presented within each plot; yellow boxes represent significant correlations.

TABLE 1.

Individual correlations between MRI data and microstructure components.

Sample	Magnitude		R2*		$\mathrm{\chi }$
	r	p	r	p	r	p
Elastin
1	0.32	0.41	−0.12	0.76	−0.76	0.017
2	0.33	0.39	−0.094	0.81	−0.90	0.0011
3	0.49	0.19	−0.32	0.40	−0.87	0.0025
4	0.66	0.054	0.18	0.65	−0.98	<0.0001
5	0.14	0.72	0.82	0.0072	−0.65	0.059
All	0.13	0.41	−0.14	0.37	−0.63	<0.0001
Collagen
1	0.78	0.012	−0.70	0.035	−0.33	0.38
2	−0.56	0.12	0.72	0.028	−0.75	0.020
3	0.46	0.22	−0.31	0.41	−0.85	0.0038
4	0.34	0.37	0.56	0.12	−0.90	0.0009
5	0.46	0.21	0.47	0.20	−0.88	0.0018
All	0.28	0.065	0.069	0.65	−0.59	<0.0001
Cell density
1	−0.49	0.18	0.63	0.069	−0.39	0.30
2	−0.020	0.96	−0.39	0.30	0.43	0.25
3	−0.46	0.21	0.20	0.60	0.46	0.21
4	−0.59	0.10	0.47	0.20	0.49	0.18
5	0.37	0.32	0.07	0.86	−0.18	0.65
All	−0.26	0.083	−0.034	0.83	0.0063	0.97
Calcium
1	−0.33	0.39	0.062	0.87	0.57	0.11
2	0.69	0.038	−0.43	0.25	0.27	0.48
3	−0.20	0.60	−0.12	0.75	0.28	0.47
4	−0.29	0.45	−0.40	0.28	0.73	0.025
5	−0.69	0.041	0.56	0.11	0.21	0.59
All	−0.19	0.20	0.19	0.21	−0.061	0.69

Pearson's r values and p values are presented for each sample between elastin, collagen, cell density, calcium, and MRI data. Significant correlations (p < 0.05) are italicized.

4. DISCUSSION

4.1. Main findings

In this study, we further demonstrate and clarify the sensitivity of magnetic susceptibility, as measured with QSM, to microstructural components in arterial tissue. With a high‐resolution 3D ME‐GRE sequence optimized for QSM, semi‐quantitative histology, and image co‐registration, we investigated this sensitivity in excised human carotid arteries. Although the excised carotids had been fixed and then stored in PBS from excision until imaging, previous work has shown no impact of formalin fixation (with rehydration in PBS) on R₂* or susceptibility values compared to fresh in arterial tissue. ¹⁹ Moderate, statistically significant correlations were found between both the collagen and elastin fraction and tissue susceptibility, measured by QSM. With increasing collagen or elastin content, the susceptibility decreased becoming more diamagnetic. However, no such trends were seen when correlating tissue microstructural components with the magnitude images and the R₂* relaxometry measurements. Previous work investigating bulk degradation of healthy porcine arteries highlighted a general sensitivity to the presence of collagen but the influence of elastin was unclear as results were confounded by severe partial volume effects ¹⁹ stemming from its removal. It is worth noting that two of the five samples displayed a correlation between collagen and elastin (Table S1), which agrees with studies showing that these two proteins co‐align alongside the smooth muscle cells, which produce them. ⁴⁶ However, in more diseased vessels, where elastin and collagen turnover is happening at different timescales, ⁴⁷ , ⁴⁸ elastin/collagen no longer co‐align or elastin/collagen may not even be present. Compared with previous studies, ⁵ , ¹⁸ , ⁴⁹ where a strong relationship between the presence of calcification was associated with highly diamagnetic regions, there was no evidence of calcification in the sections of samples analyzed in this study and, therefore, the susceptibility was dominated by the presence of elastin and collagen. The calcium fraction measured in these samples was very low and the overall correlation of all MRI measures with calcium fraction was weak and non‐significant. There was a lack of variation in calcium content within individual samples, leading to the biggest calcium differences being between samples. There was less than 10% calcium in each ROI, which can be clearly visualized in Figure 5. This indicates that the presence of calcium was not a confounding factor in establishing the relationship between tissue microstructure and MRI measures.

4.2. Clinical relevance

At present, QSM has been applied to carotid arteries in vivo in only a handful of studies. ⁴ , ⁵ , ¹⁸ , ⁵⁰ These have focused specifically on the differentiation of IPH and calcification, showing the ability to identify both morphologies in patients scheduled for carotid endarterectomy surgeries. Ikebe et al. ¹⁸ showed, for the first time, that calcification susceptibility was significantly lower than in lipid rich necrosis. Wang et al. ⁵ furthered this, highlighting QSM's capability to differentiate IPH and calcification. Different QSM processing toolboxes were compared on simulated and in vivo carotids by Nguyen et al., ⁴ finding that MEDInpt provided reliable QSM maps of IPH and calcification. These studies show the promise of QSM for identifying and discriminating advanced, vulnerable ⁵¹ , ⁵² , ⁵³ atherosclerotic morphologies in vivo. These findings are supported by ex vivo QSM, which has been linked to quantitative histology for IPH with and without iron and calcifications. ⁴⁹ Clinically, vulnerable plaques are currently only diagnosed by the percent stenosis, which are patients showing luminal blockage greater than 50% being referred for a carotid endarterectomy. However, there is evidence showing that patients with mild stenosis can also have a high risk of stroke. ¹⁷ Moreover, IPH can be present in low‐grade stenosed vessels. ⁵⁴ Although a better clinical indicator of plaque rupture risk is needed, it is not necessarily clear at what stage in atherosclerosis this indicator is needed. Early changes in the vessel wall are characterized by changes in smooth muscle cell density, collagen content and organization, and decreased elastin. ⁵⁵ , ⁵⁶ In this work, we show that QSM is sensitive and significantly correlated to changes in both collagen and elastin, showing, for the first time in aged human arterial tissue, an imaging biomarker for early pathological changes in the arterial wall. Outside of atherosclerosis, an imaging biomarker for collagen and elastin changes has real potential in other pathologies such as aneurysm or dysfunctional elastin production, such as Williams syndrome ⁵⁷ or cutis laxa. ⁵⁸

4.3. Future work

Although studies have been performed in vivo using 3D ME‐GRE QSM at the carotid bifurcation, there are multiple aspects to consider for the translation of the findings from this ex vivo study for clinical decision‐making. First, from an ex vivo standpoint, the sensitivity of QSM to these components needs to be examined from a mechanical perspective to truly gauge the usefulness of the technique in identifying microstructures, which may be more at risk of rupture. Different studies have shown particular promise in elucidating this relationship, pointing to a microstructural dependence on plaque mechanical failure properties. ¹³ , ³⁸ , ⁵⁹ Additionally, the relationship between susceptibility and the tissue microstructure, such as its regional elastin and collagen fiber fraction and their orientation, will need to be extensively outlined. Specifically, in addition to collagen content sensitivity, other collagenous tissues have shown a susceptibility anisotropy behavior, ⁶⁰ which remain to be explored in arterial tissue, especially when considering the mechanical implications of altered collagen and elastin alignment. ¹³ Extensive outlining of this relationship will be extremely challenging and will require improvements on two main aspects: the accuracy of the QSM‐derived susceptibility maps that can be obtained for the tissue and more representative microstructural evaluation techniques. Regarding the accuracy of the QSM‐derived susceptibility maps, Stone et al. ¹⁹ showed that the presence of disease regions within vascular tissue drastically decreases the quality of the susceptibility maps, which yields more severe streaking artifacts and residual background fields when a conventional QSM reconstruction pipeline is used. To overcome this issue, a tissue masking technique was used to exclude those regions. ¹⁹ The pipeline used in this study is based on a pipeline previously optimized in ex vivo arterial tissue ¹⁹ because the tissue samples in this study show aging and very mild disease. However, for more diseased arterial tissue, to the authors' knowledge, no formal pipeline optimization has been carried out to obtain accurate susceptibility maps, although work has been done for in vivo carotid imaging. ⁵⁰ Although we tried to adapt in‐phase echo times optimized for head‐and‐neck QSM at 3 T for the increased fat–water frequency shift at 7 T, future work will refine the acquisition for 7 T, particularly for increased numbers of echoes where phase shifts can become large. ⁶¹ Future work could also evaluate multi‐peak fat model‐based effective in‐phase echo times ⁶² or deep learning‐based fat–water QSM. ⁶³ Note that we did not observe any fat–water chemical shift artifacts, likely because of negligible amounts of lipid present. Regarding microstructural evaluation, histology currently represents the gold‐standard technique for visualization. However, its 2D and destructive nature renders the technique suboptimal for the assessment of atherosclerotic tissue as this presents structural variation in more than two directions. Although other techniques, such as 3D histology and second harmonic generation imaging allow for the evaluation of microstructure in a 3D (2D stack) fashion, these are either labor intensive or limited to a small FOV and remain unable to capture meaningful microstructural variations directly to correlate with MRI. Contrast‐enhanced X‐ray micro‐focus computed tomography could offer a lab‐based technique with the potential to overcome such limitations. ⁶⁴ However, the binding specificities of different contrast‐enhancing staining agents to specific microstructural components still remains under exploration, with some agents, such as phosphotungstic acid and 1:2 hafnium‐substituted Wells Dawson polyoxometalate, having been specifically suggested to show a binding affinity to collagen and elastin, respectively, in vascular tissue. ⁶⁵ , ⁶⁶ From an in vivo standpoint, although there are a handful of in vivo studies existing, all have used anisotropic voxel‐sizes and it has been shown that isotropic resolution is preferable for accurate QSM. ⁶⁷ Additionally, a study by Karsa et al. ³⁵ established an optimized and repeatable head and neck QSM sequence with a suitable processing pipeline, which should be used as a basis for translating these results, incorporating relevant elements of the recent consensus recommendations for clinical QSM in the brain. ³⁶

5. CONCLUSION

Tissue susceptibility, measured by QSM, has previously been linked to key components of arterial tissue microstructure. ¹⁹ This study demonstrates significant direct negative correlations between tissue magnetic susceptibility and collagen and elastin content in diseased human carotid arteries ex vivo. It highlights the potential of QSM to offer unique insights into microstructural changes that could provide a useful clinical indicator of the onset and progression of carotid atherosclerosis.

FUNDING INFORMATION

European Research Council (ERC) under the European Union's Horizon 2020 research innovation programme, Grant Agreement Number: 637674; ERC 2022 Proof of Concept (VASCOLL); ERC Consolidator Grant: DiSCo MRI SFN 770939 to K.S.

Supporting information

Table S1. Correlations of individual samples, and across all samples, between different microstructural components. Pearson's r values and p values are presented for each sample. Significant correlations (p < 0.05) are italicized.

Table S2. Susceptibility values (mean and standard deviation) of the masked region and the background medium (PBS) for each sample.

Table S3. Pearson's r values between component and MRI metrics ($\mathrm{\chi }$, R ₂*, Magnitude) across all samples as a function of differing k‐means.

Table S4. R ₂* values (mean and standard deviation) of the background medium, PBS, for each sample.

Figure S1. A central, axial slice of magnitude images, R ₂* maps, and susceptibility maps within the common carotid for each sample. Unmasked images and maps are presented.

Stone AJ, Tornifoglio B, Digeronimo F, Shmueli K, Lally C. Quantitative susceptibility mapping of the human carotid artery: Assessing sensitivity to elastin and collagen ex vivo. Magn Reson Med. 2025;94:771‐784. doi: 10.1002/mrm.30500