Introduction
The incidence of aortic diseases is increasing with an ageing population [
1]. In Western society, the incidence is rising further due to a growing group with overweight, hypertension and diabetes [
2,
3]. Aortic disorders include dissection and aneurysms of the aorta, often in combination with aortic valve problems and atherosclerosis [
4,
5]. Detailed imaging of the thoracic aorta with assessment of its dimensions is of major importance to monitor and detect aortic diseases in order to make valid clinical decisions [
5,
6].
Computed tomography angiography is often used for clinical imaging of the thoracic aorta, because of speed and excellent spatial resolution [
7]. However, it is less suitable for regular follow-up because it requires iodinated contrast and ionizing radiation [
5,
7], making magnetic resonance angiography (MRA) a good alternative [
5‐
8]. Contrast-enhanced (CE) MRA uses a gadolinium contrast bolus for 3-dimensional (3D) aortic depiction [
5,
9]. Imaging of this contrast bolus requires a short acquisition time and does not allow for ECG triggering, resulting in movement artefacts close to the heart [
10]. To avoid overestimation or underestimation of aortic diameters, it was therefore advised not to use ungated sequences in the sinus of Valsalva and sinotubular junction [
9]. Furthermore, the possibility of gadolinium nephrotoxicity in patients with reduced renal function and the unknown effects of gadolinium deposition in the brain on later life caused the European Medicine Agency to advise minimisation of gadolinium contrast agent use [
11‐
14]. Three-dimensional balanced steady-state free precession (SSFP) MRA of the thoracic aorta allows ECG triggering and navigator gating, and since it uses phase contrast, an additional contrast agent is not required [
15]. Although ECG-triggered imaging was already proven to show superior image quality and similar aortic dimensions compared to CE-MRA [
16‐
19], these sequences result in longer acquisition times and are therefore more prone to artefacts in patients with arrhythmias and clinically less-stable patients [
20].
Measurements performed on either CE-MRA or SSFP-MRA must be interchangeable and reproducible so the best suited sequence can be chosen based on the patient’s current medical condition, without influencing follow-up decision making. A standardised measurement operating procedure is equally important, as differences in slice selection can cause (additional) variability [
19]. Our objective of this study was to test the hypothesis that dimensions of thoracic aortic landmarks [
5,
6] assessed with CE-MRA and SSFP-MRA are comparable and that intra- and inter-observer reproducibilities are improved on SSFP-MRA for landmarks close to the aortic root caused by better image quality.
Discussion
In this study, we compared non-ECG-triggered CE-MRA and SSFP-MRA to investigate their interchangeability and reproducibility in aortic dimension assessment. Aortic dimensions were assessed at nine predefined landmarks using a standardised protocol adjusted with spacing rules for landmarks without clear surrounding anatomy to accurately measure aortic dimensions, resulting in similar dimensions between CE-MRA and SSFP-MRA (Fig.
3, Supplementary Table
2).
The high correlation of aortic dimensions between non-ECG-triggered CE-MRA and SSFP-MRA was in agreement with previous studies focusing on thoracic aortic dimensions [
19,
26‐
28]. Aortic dimensions were not overestimated or underestimated with SSFP-MRA compared to CE-MRA in studies reporting dimensions assessed on a cross-sectional overview [
15,
19,
26,
28]. Conversely, Veldhoen et al [
27] reported underestimation of dimensions with SSFP-MRA. They, however, acquired 2D instead of 3D images and analysed aortic dimensions on the para-sagittal plane, leading to perhaps less accurate dimensions. The Society of Cardiovascular Magnetic Resonance recommends aortic dimension assessment on double-oblique multiplanar images, requiring 3D sequences [
9]. In our study, the difference between CE-MRA and SSFP-MRA tended to decrease for landmarks more distal from the sinus of Valsalva. This finding was confirmed by Von Knobelsdorff-Brenkenhoff et al [
19], the only other study reporting differences between 3D SSFP-MRA and CE-MRA for multiple landmarks along the thoracic aorta. However, in their study, the mid-ascending aorta showed relatively high variability compared to other aortic landmarks. As their study population primarily consisted of patients with suspicion or control of ascending aortic aneurysms (74%), they suggested that inaccurate slice selection resulted in large differences of aortic diameters within aortic aneurysms, emphasizing the importance of accurate slice selection. In our study, spacing rules were used for locations without clear landmarks, which possibly explains the lower variability. Moreover, as few of our included subjects were suffering from an ascending aortic aneurysm, less dimension variability can be expected.
The intra- and inter-observer variabilities of the sinus of Valsalva and sinotubular junction were significantly smaller in SSFP-MRA than in non-ECG-triggered CE-MRA (Table
1) which is in accordance with previous studies [
29‐
31] as is the similar variability of landmarks distal from the aortic root [
22,
26,
29,
30,
32]. Only Von Knobelsdorff-Brenkenhoff et al [
19] reported similar intra- and inter-observer variabilities for all landmarks, suggesting no reproducibility improvement of the aortic root measurements with SSFP-MRA.
Landmarks close to the aortic root are sensitive to cardiac motion in untriggered sequences [
10]. We assessed the image quality on three landmarks along the aorta to study differences in image quality. The image quality of the sinus of Valsalva was significantly better with SSFP-MRA (
p < 0.05), while the image quality of the mid-ascending and mid-descending aortas was similar with non-ECG-triggered CE-MRA (Fig.
6). Most previous studies assessed the image quality of the thoracic aorta with ECG-triggered SSFP-MRA versus non-ECG-triggered CE-MRA using visual classification criteria [
18,
19,
28,
29,
31,
33], whereas one lacked these criteria [
15] and another two studies used a semi-automatic approach [
24,
30]. Despite the variation of methods, all studies reported improved image quality with SSFP-MRA for the aortic root [
15,
19,
28‐
31,
33,
34]. Some studies also reported increased image quality of the mid-ascending aorta [
19,
24,
29,
30,
34] and mid-descending aorta [
19,
29]. These improvements were small compared to improvements in the aortic root, probably caused by similarly good average image quality of CE-MRA for these landmarks.
Only a limited number of the studies mentioned above reported results of both observer reproducibility and image quality in SSFP-MRA and non-ECG-triggered CE-MRA [
19,
29‐
31]. Potthast et al [
30] demonstrated superior image quality and reproducibility with SSFP-MRA for the sinus of Valsalva, sinotubular junction and ascending aorta. Bannas et al [
31] also reported significantly improved image quality and reproducibility with SSFP-MRA compared to CE-MRA for the sinus of Valsalva and sinotubular junction; however, the reproducibility of other landmarks was not reported. Furthermore, van Kesteren et al [
29] reported the lowest and highest image qualities with CE-MRA, with corresponding lowest and highest inter-observer agreements in measuring diameters, in the sinus of Valsalva and the distal ascending aorta, respectively. Although Von Knobelsdorff-Brenkenhoff et al [
19] showed improved image quality for all landmarks with SSFP-MRA compared to CE-MRA along the aorta, the overall intra- and inter-observer reproducibilities were similar between both sequences, suggesting that image quality does not influence observer reproducibility. To our knowledge, our study is the first to directly link the reproducibility of aortic dimensions with image quality, since previous studies only reported values of reproducibility and image quality for the entire study population. We demonstrated decreased reproducibility of dimensions in images with impaired quality which is in accordance with the main literature findings, suggesting that impaired CE-MRA quality at the aortic root causes impaired reproducibility.
Surgery indications for thoracic aortic aneurysms are mainly based on either absolute aortic diameters or size increase over 1 year [
5]. The lower threshold for aneurysms is stated in guidelines and depends on the aortic segment of the dilation—ascending aorta, aortic arch or descending aorta—and on the medical condition of the patient. The size increase of > 3 mm/year should be measured using repetitive measurements obtained with the same imaging technique on the exactly the same aortic level, and this increase should be checked using an alternative technique to test its consistency when it impacts the therapeutic decision [
5]. To prevent overestimation or underestimation of aortic size increase between repetitive measurements when using both non-ECG-triggered CE-MRA and SSFP-MRA in follow-up, similarity of dimensions is required which was demonstrated in this study. However, the variability in and between observers within the same sequence will inevitably add some uncertainty about the actual size increase, which is especially the case in measuring the aortic root in CE-MRA and is caused by motion artefacts. When aortic root dilatation is suspected, ideally SSFP-MRA should be performed to minimise both false negative and positive surgery indications. The use of CE-MRA or SSFP-MRA does not influence the accuracy of the assessed aortic size increase in suspected aortic dilation in other segments, since variation was found to be similar between sequences. To ensure the selection of the exactly the same aortic level in follow-up measurement, it may be helpful to measure the distances from that aortic level towards the closest proximal and distal anatomic landmark and add this information to the medical report.
This study has some limitations. In 13 of the 30 subjects, the SSFP-MRA was performed after gadolinium administration where it ideally should have been performed prior to gadolinium administration to prevent image quality bias from contrast agents. Nevertheless, there was no measurable image quality and reproducibility improvement in the SSFP-MRA acquired after compared to before gadolinium administration (p = 0.45 and p = 0.29, respectively). Furthermore, the second observer performed the analysis once and therefore the intra-observer variability was based on the reproducibility measurements of the first observer.
The aortic dimensions of predefined locations were similar on SSFP-MRA and non-ECG-triggered CE-MRA using a standardised operating procedure to ensure reproducible slice selection. SSFP-MRA showed improved reproducibility for landmarks close to the aortic root, since these landmarks in non-ECG-triggered CE-MRA are sensitive to cardiac motion and therefore result in impaired image quality. Contrast-free SSFP-MRA seems to be a good alternative for assessment of thoracic aortic dimensions with improved image quality and reproducibility.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.