Abstract
The qMATCH technique enables high-resolution 3D multi-contrast evaluation of carotid atherosclerosis, but the optimal acquisition orientation for reproducible image quality remains unclear. This prospective study compared image quality, motion robustness, and reproducibility between transverse and coronal qMATCH acquisitions. Thirty participants with moderate-to-severe carotid stenosis each underwent two scans, totaling 60 imaging sessions. Coronal imaging demonstrated significantly fewer motion artifacts compared with transverse imaging (34.7% vs 41.5%, p < 0.05) and yielded higher image quality scores (2.78 vs 2.72, p < 0.05) and arterial wall aSNR (5.10 vs 4.36, p < 0.05). In motion-corrupted images, coronal imaging also showed superior image quality (2.75 vs 2.65, p < 0.05) and arterial wall aSNR (4.89 vs 4.03, p < 0.05). Reproducibility for T1/T2 quantification was better for coronal acquisitions, particularly in motion-affected cases. These findings suggest that coronal qMATCH is more robust for motion-prone patients and better suited for longitudinal carotid imaging studies.
INTRODUCTION
Carotid atherosclerosis is a leading cause of cerebral ischemic stroke, underscoring the importance of accurate evaluation of the carotid vessel wall for risk assessment and disease management.1 MR vessel wall imaging is particularly well-suited for this purpose due to its superior soft-tissue contrast and ability to noninvasively characterize plaque composition. A standardized classification system for carotid atherosclerotic plaques, known as the Carotid Plaque–Reporting and Data System (RADS),2 highlights the importance of MRI in identifying high-risk plaque features: lipid-rich necrotic core, fibrous cap status and rupture, and intraplaque hemorrhage. These features improve diagnostic accuracy and stroke prevention but require high-quality, co-registered multi-contrast imaging robust to observer variability.
We designed the quantitative multi-contrast atherosclerosis characterization (qMATCH) technique based on the low-rank tensor framework3 which exploits the partial separability of space and contrast dimensions in the multi-contrast images to achieve vast acceleration. This technique enables simultaneous acquisition of bright-blood MRA, dark-blood wall images, multiple T1/T2 weightings, and quantitative T1/T2 mapping in under eight minutes. As a 3D imaging technique, qMATCH allows flexible orientation for imaging, high spatial resolution and reformatting capabilities, offering a tailored solution for carotid plaque assessment.
Despite these advancements, motion artifacts remain challenging due to the thinness of the vessel wall and its sensitivity to abrupt bulk motion. These artifacts exacerbate partial volume effects, leading to incomplete vessel wall visualization and errors in quantification. Selecting an optimal imaging orientation may provide a universally applicable solution to mitigate motion artifacts. The objective of this study is to determine the optimal imaging orientation and support integration of qMATCH into risk-stratification frameworks such as Plaque–RADS.
METHODS
Cohort
This prospective study was approved by the Institutional Review Board at all participating institutions and followed the STROBE guidelines (Supplemental Data). Written informed consent was obtained from all participants. Inclusion criteria were age >50 years with >30% stenosis or >4 mm plaque thickness by ultrasound. Exclusion criteria included MRI intolerance, contraindications, contrast allergy, incomplete scans, or severe motion artifacts in both sessions.
Sequence Implementation
We designed the qMATCH technique based on the low-rank tensor framework which exploits the partial separability of space and contrast dimensions in the multi-contrast images to achieve vast acceleration. This sequence employed 3D flow-compensated spoiled gradient echo readout with variable-duration T2-inversion recovery preparations to generate T1 and T2 contrast. Cartesian acquisition with randomized reordering in ky and kz directions was implemented according to a variable-density Gaussian distribution. The center k-space line was collected every eighth readout to serve as low-rank tensor training data.
MR Imaging Protocol
Each participant underwent both transverse and coronal imaging, with two scans performed within two weeks. Before each scan, participants received verbal instructions to minimize head and jaw motion during data acquisition. All imaging was performed on 3.0-T scanners (Skyra, Vida and PrismaFit, Siemens Healthineers, Erlangen, Germany) using a Neck-Shape-Specific coil. The transverse acquisition had a spatial resolution of 0.63 × 0.63 × 2 mm and an FOV of 160 × 160 × 60 mm, while the coronal acquisition had a spatial resolution of 1 mm isotropic with an FOV of 160 × 160 × 36 mm. The following parameters were consistent across both orientations: flip angle = 5°, TR = 2.08 s, and TEs = 20/30/40/60/70/80 ms. The spatial resolution and FOV differed between transverse and coronal acquisitions to ensure adequate coverage and vessel wall visualization while maintaining the same total scan times, approximately 8 minutes under free-breathing conditions.
Image Quality Review
To standardize comparisons, images acquired in the coronal plane were reformatted into transverse planes. Transverse images of bilateral carotid arteries from both imaging orientations were matched by distance from the bifurcation and evaluated at five locations: the bifurcation (slice 0), below (slice −4, −2), and above (slice 2, 4). Image quality was rated by two radiologists (5-year and 1-year experience) using a four-point standardized scale4 (1 = poor; 2 = adequate; 3 = good; 4 = excellent) with scores averaged. Motion corruption at each location was determined by consensus.
Quantitative Metrics
For images with a quality score of 2 or greater, a radiologist manually delineated ROIs on dark-blood vessel wall images, outlining the arterial wall and lumen to measure signal intensities (Sw and Sl, respectively). Noise (σn) was measured as the standard deviation of signal intensity from an ROI (~12.5-mm2) placed in the sternocleidomastoid muscle region near the target artery. Due to the use of parallel imaging, absolute signal-to-noise ratio and contrast-to-noise ratio were difficult to quantify. Instead, the values were calculated as apparent signal-to-noise ratio (aSNR) and apparent contrast-to-noise ratio (aCNR), respectively. The aSNR of the wall and lumen was determined by using the following equation: Sl(w) / σn. Wall-lumen aCNR was defined as the difference between wall and lumen aSNRs. The ROIs prescribed on the dark-blood vessel wall images were transferred across all multi-contrast image sets, and corresponding T1 and T2 values were extracted from these regions on quantitative maps.
Reproducibility Assessment
To evaluate the impact of motion on reproducibility, each image pair of the same location from the same subject was classified as either ‘with motion interference’ or ‘without motion interference.’ Motion interference was defined as the presence of motion artifacts in at least one image of the pair. Intraclass correlation coefficient (ICC) was calculated for T1 and T2 measurements within each category. Bland-Altman plots with limits of agreement (LoA) were generated for transverse and coronal acquisitions.
Statistical Analysis
A chi-square test was applied to assess intergroup differences in motion artifact rates for both imaging orientations. Paired t-tests were used to compare intergroup differences in image quality scores, aSNR, and wall-lumen aCNR. Cohen’s kappa coefficient was calculated to determine the interrater agreement between the two experts. Statistical significance was defined as p <0.05. All statistical tests were performed using Pingouin (version 0.5.5), an open-source statistical package in Python.
RESULTS
Participant Characteristics
Thirty participants (3 women, 27 men; mean age, 74.6 years; range, 58–87 years) were included in the final analysis, yielding 60 datasets. One additional participant was excluded due to severe motion artifacts in both imaging sessions, preventing region-based analysis.
Qualitative Image Quality
A total of 1,200 images were analyzed, including 600 coronal and 600 transverse images (5 slices × 2 carotid arteries × 60 imaging sessions). Coronal imaging demonstrated significantly fewer cases with motion artifacts compared to transverse imaging (34.7% vs. 41.5%, p < 0.05) (Figure 1). Among the 264 motion-corrupted images, 131 (49.6%) exhibited artifacts only on transverse imaging, 49 (18.6%) exhibited artifacts only on coronal imaging, and 84 (31.8%) exhibited artifacts on both acquisitions. Figure 2 shows plots visualizing the distributions of image quality, wall aSNR, lumen aSNR, and wall-lumen aCNR for both imaging orientations. In overall comparison, coronal imaging demonstrated significantly higher image quality scores (2.78 vs. 2.72, p < 0.05) and arterial wall aSNR (5.10 vs. 4.36, p < 0.005). In non-corrupted images, coronal acquisition showed significantly higher wall aSNR (5.42 vs. 4.84, p < 0.005). In motion-corrupted images, coronal acquisition yielded significantly higher image quality scores (2.75 vs. 2.65, p < 0.005) and wall aSNR (4.89 vs. 4.03, p < 0.005) (Table 1). Interrater agreement for image quality was substantial (κ = 0.78; 95% CI, 0.74–0.82). Dark-blood vessel wall images acquired in the coronal plane demonstrated superior robustness to motion (Figure 3).
FIG 1.

Pie charts comparing the proportion of motion-corrupted and non-corrupted images for transverse and coronal acquisitions.
FIG 2.

Violin plots depicting distributions of image quality score, wall aSNR, lumen aSNR, and wall-lumen aCNR for transverse and coronal acquisitions without motion corruption and with motion corruption.
Table 1:
Comparison of image quality score, wall apparent SNR, lumen apparent SNR, and wall-lumen apparent CNR between two imaging orientations.
| Parameter, orientation | Without motion corruption (N=743) | With motion corruption (N=457) | Overall (N=1200) |
|---|---|---|---|
|
| |||
| Image quality score | |||
| Transverse | 2.84 ± 0.06 | 2.65 ± 0.06 | 2.72 ± 0.04 |
| Coronal | 2.83 ± 0.06 | 2.75 ± 0.05 | 2.78 ± 0.04 |
| P value | 0.91 | < 0.005 | 0.009 |
| Mean difference | 0.00 (−0.07, 0.07) | 0.10 (0.05, 0.16) | 0.06 (0.01, 0.10) |
| Wall apparent SNR | |||
| Transverse | 4.84 ± 0.30 | 4.03 ± 0.26 | 4.36 ± 0.20 |
| Coronal | 5.42 ± 0.45 | 4.89 ± 0.36 | 5.10± 0.28 |
| P value | < 0.005 | < 0.005 | < 0.005 |
| Mean difference | 0.58 (0.20, 0.97) | 0.86 (0.53, 1.19) | 0.75 (0.50, 1.00) |
| Lumen apparent SNR | |||
| Transverse | 2.42 ± 0.30 | 1.73 ± 0.10 | 2.01 ± 0.13 |
| Coronal | 3.14 ± 0.36 | 2.51 ± 0.18 | 2.76 ± 0.18 |
| P value | < 0.005 | < 0.005 | < 0.005 |
| Mean difference† | 0.72 (0.46, 0.97) | 0.77 (0.59, 0.95) | 0.75 (0.60, 0.90) |
| Wall-lumen apparent CNR | |||
| Transverse | 2.42 ± 0.30 | 2.30 ± 0.22 | 2.35 ± 0.18 |
| Coronal | 2.29± 0.34 | 2.39 ± 0.29 | 2.35± 0.22 |
| P value | 0.43 | 0.53 | 0.99 |
| Mean difference† | −0.13 (−0.46, 0.20) | 0.09 (−0.19, 0.36) | 0.00 (−0.21, 0.21) |
FIG 3.

Representative qMATCH image sets acquired by transverse and coronal imaging without motion corruption (A and B) and with motion corruption (C and D). Without motion corruption, both imaging orientations exhibit good-to-excellent image quality, characterized by high spatial resolution and tissue contrast. With motion corruption, coronal imaging demonstrates greater robustness to motion artifacts, maintaining clear visualization of the vessel wall. Motion artifacts in transverse imaging are evident as an ill-defined lumen region (C) and loss of wall signals (D).
Reproducibility of T1 and T2 values
In the 370 measurements without motion interference, LoA for T1 and T2 values ranged from - 501.39 to 506.01 (ICC = 0.67) and - 37.99 to 38.72 (ICC = 0.79) for transverse acquisitions, alongside - 452.28 to 452.13 (ICC = 0.71) and - 28.30 to 27.06 (ICC = 0.74) for coronal acquisitions. In the 230 measurements with motion interference, LoA for T1 and T2 values ranged from - 1016.03 to 841.81 (ICC = - 0.05) and - 54.13 to 51.96 (ICC = 0.22) for transverse acquisitions, and - 649.00 to 973.02 (ICC = 0.31) and - 52.65 to 51.60 (ICC = 0.28) for coronal acquisitions (Figure 4). These results suggest superior reproducibility for coronal acquisitions, particularly in motion-affected scans.
FIG 4.

Bland-Altman plots illustrate inter-session agreement for T1 and T2 values without motion interference (N = 370) and with motion interference (N = 230). The dashed red and green lines represent the upper and lower 95% LoA, respectively. Motion interference resulted in widened LoA for both imaging orientations; however, coronal acquisitions consistently exhibited narrower LoA than transverse acquisitions.
DISCUSSION
This study evaluated and compared transverse and coronal acquisitions of the qMATCH technique for carotid MR imaging, focusing on image quality, motion robustness, and reproducibility. Data from multiple study sites were included in the analysis to account for variation in scanner models and software versions, and enhance the generalizability of the results. The main finding of our study showed that coronal acquisition appeared to perform better than transverse acquisition overall. With the same scan time, coronal acquisition yielded fewer motion artifacts, superior image quality, and better reproducibility.
Advantages of coronal acquisition likely stem from multiple factors. The carotid bifurcation is highly susceptible to various types of motion, notably swallowing and arterial pulsation.5 Motion may result in blurring and, more importantly, aliasing artifacts that can severely obscure regions of interest. Transverse acquisition covered the entire cross-section of the neck, allowing motion artifacts from outside the carotid regions, particularly the anterior aspect, such as the jaw, to affect the carotid arteries. In coronal imaging, anterior-posterior coverage was limited to the carotid region, eliminating artifacts from outside the slab. The overall advantage of coronal imaging in image quality was primarily due to its motion robustness, as we observed in cases without substantial motion, transverse imaging achieved equal or better ratings. Coronal imaging also outperformed transverse imaging in apparent vessel wall and lumen SNR, with or without motion. These results may reflect the 28% larger voxel size in coronal imaging (1 mm3 vs. 0.781 mm3). Lastly, Bland-Altman analysis showed coronal acquisitions may provide more reproducible T1 and T2 maps, particularly in motion-affected scans, potentially supporting enhanced delineation of plaque components such as fibrous cap, lipid-rich necrotic core, and intraplaque hemorrhage. Although sagittal imaging is technically feasible with qMATCH, it was not included in this study. Given the anatomy of carotid arteries, sagittal orientation would require approximately 2 times as many slices as coronal acquisition to cover both sides of the vessels, making it much less time-efficient.
Clinically, these findings suggest coronal imaging should be prioritized for patients prone to motion, such as those with neurological disorders, tremors, or difficulty following instructions. Its superior reproducibility supports use in longitudinal studies, including plaque monitoring and treatment response. Coronal acquisition may also complement transverse imaging in hybrid protocols, where transverse sequences provide anatomical information and coronal sequences improve reliability in motion-prone patients.
There were several limitations to this study. First, the subjective nature of qualitative image quality evaluation and motion artifact classification may have led to potential bias. To alleviate this issue, we averaged assessments from two independent readers. Second, preliminary analysis across three 3T Siemens systems showed higher image quality, wall aSNR, and fewer motion-corrupted cases on the platform where the sequence was originally developed and optimized, likely reflecting developer bias (Supplemental Data). However, the trend between the two imaging orientations appeared to be consistent across different sites. Third, excluding the severely motion-corrupted scan in analysis may potentially underestimate motion prevalence in clinical practice and bias the results toward less severe motion cases, although this effect is expected to be minimal. Future studies with larger, potentially more sex-balanced cohorts are needed to validate these findings and enable more detailed analyses of vessel geometry and plaque characterization, ideally with histological correlation.
CONCLUSION
qMATCH with coronal orientation offers a more motion-robust and reproducible solution for carotid assessment compared with transverse orientation, making it particularly suitable for longitudinal studies and patients prone to motion artifacts.
Supplementary Material
ACKNOWLEDGMENTS
This work was supported partly by NHLBI R01HL159200, NHLBI R01HL159803, VA MERIT I01CX002208, VA MERIT I01CX001901, and NIA R01AG070987. We thank the DTECT Consortium for coordinating data acquisition.
ABBREVIATIONS:
- aCNR
apparent contrast-to-noise ratio
- aSNR
apparent signal-to-noise ratio
- ICC
Intraclass Correlation Coefficient
- LoA
limits of agreement
- qMATCH
Quantitative Multi-Contrast Atherosclerosis Characterization
Footnotes
The authors declare no conflicts of interest related to the content of this article.
REFERENCES
- 1.Bos D, Arshi B, van den Bouwhuijsen QJA, et al. Atherosclerotic Carotid Plaque Composition and Incident Stroke and Coronary Events. Journal of the American College of Cardiology 2021;77:1426–35. [DOI] [PubMed] [Google Scholar]
- 2.Luca Saba, Riccardo Cau, Alessandro Murgia, et al. Carotid Plaque-RADS. JACC: Cardiovascular Imaging 2024;17:62–75. [DOI] [PubMed] [Google Scholar]
- 3.Xie Y, Christodoulou A, Wang N, et al. Quantitative multi-contrast atherosclerosis characterization (qMATCH): Comprehensive quantitative evaluation of atherosclerosis in a single-scan. In: Processing of the 25th Annual Meeting of ISMRM. Honolulu, HI; 2017:3122. [Google Scholar]
- 4.Underhill HR, Yarnykh VL, Hatsukami TS, et al. Carotid Plaque Morphology and Composition: Initial Comparison between 1.5- and 3.0-T Magnetic Field Strengths. Radiology 2008;248:550–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Boussel L, Herigault G, de La Vega A, et al. Swallowing, arterial pulsation, and breathing induce motion artifacts in carotid artery MRI. Journal of Magnetic Resonance Imaging: An Official Journal of the International Society for Magnetic Resonance in Medicine 2006;23:413–5. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
