In this cross-over single-centre retrospective study, a cohort of 30 patients (60.9±16.1 yo, 7 females) with various chronic aortic diseases (aortic dissection type A: 10, aortic dissection type B: 7, aortic aneurysm: 8; status after contained aortic rupture: 3, Marfan syndrome: 1, stenosis of left subclavian artery and impact of coronary artery bypass: 1) underwent a clinical follow up routine MRI examination between July and October 2022. The standard TWIST was complemented by a GRASP sequence after the study was approved by the local institutional review board and all patients gave their written informed consent prior to the MRI examination.
MRI data acquisition
All data was acquired on a 1.5T clinical scanner (Magnetom SolaFit, Siemens, Erlangen, Germany) equipped with a 32 channel body coil.
With each CE-trMRA acquisition the same amount of Gd contrast agent (CA) (Gadovist 1.0M, Bayer, Switzerland AG, Zurich) was administered (0.075 ml/(kg bw), flow rate 4 ml/s), i.e. twice during each protocol. All images were acquired during free breathing in the oblique coronal plane. To reduce bias due contrast enhancement in the vascular systems during the second CA administration, TWIST and GRASP sequences were acquired in reverse order for half of the patients (n=15), respectively. For all examinations, there was a three minutes pause between CA administrations.Reconstruction of GRASP data has been performed inline at the scanner within about 30 s.
All acquisition parameters are summarised in Table 1. For the TWIST sequence, 25% of the k-space center was used to reconstruct a time frame, whereas 33% of the remaining k-space periphery was sampled between each acquisition of k-space center [25]. For the GRASP sequence, 13 radial projections were used to reconstruct one time frame.
Qualitative image analysis
Qualitative analysis was performed using original non-subtracted 3D dynamic images. Three experienced radiologists (with 20 (#1), 5 (#2), and 8 (#3) years of experience) assessed the overall image quality independently. To this end, vascular contrast, vessel sharpness and image artefacts of TWIST and GRASP were assessed by grading the images on a 4-point Likert scale. In detail: for overall image quality, vascular contrast and vessel sharpness: 1=excellent; 2=acceptable (good); 3=poor (still diagnostic); 4=non-diagnostic. For image artefacts: 1=no artefacts; 2=minor artefacts (not interfering with diagnostic content), 3=moderate artefacts (degrading diagnostic content, image still diagnostic), 4=severe artefacts (non-diagnostic image).
When assessing image artefacts special attention was paid to streaking artefacts for GRASP and fold-over artefacts for TWIST. For the overall image quality index, readers focused on vascular enhancement visibility and perfusion on the vessel of the ascending aorta, supra-aortic vessels, intercostal arteries, visceral branches, particularly inferior mesenteric artery, renal cortex (primary entry and additional small communication channels).
Quantitative image analysis
Quantitative image analysis was performed by MATLAB 9.12 (MathWorks, Natick, MA, USA) using original non-subtracted images. Circular regions-of-interest (ROIs) were placed at three aortic levels as shown in Figure 1: ascending aorta (AA), descending aorta at the level of the pulmonary trunk (DA), and abdominal aorta at the level of the infrarenal arteries (AbA). In cases of an aortic dissection the true lumen was chosen for the placement of the ROIs. When drawing circular ROIs, those slices and temporal frames were selected, where the anatomy of interest was best visible for the readers. For each patient identical ROIs at the same positions were used for both sequences. ROIs were drawn on TWIST images first for half of the patients. The temporal behaviour of the sequences was quantified by obtaining the maximum slope of the contrast agent uptake c (maxslope) and the full width at half maximum (FWHM) from the normalised and interpolated signal intensity time courses within the ROIs. Time signal intensity curves were interpolated by a factor of 10 and a smoothing function was applied to exclude the influence of noise on the outcome.
Spatial blurring was quantified by calculating vessel sharpness as follows: at the same levels of the aorta as described above a straight line perpendicular to the vessel wall was drawn across the aorta. Vessel sharpness (vs) of the boundary was then calculated from the resulting signal intensity profile as vs=1/d. The value d (millimetres) is the distance between those points on the drawn line, between which the signal intensity changed from 20% to 80% of the absolute intensity difference, i.e. the difference between the maximal and the minimal signal intensity values [26] (Figure 2).
Signal-to-Noise ratio (SNR) calculation was performed by adopting the first method described in the previous work [27]. For noise, images from the second and the third time frames (prior to the bolus onset) were subtracted. The first time frame was not used as it has slightly different acquisition parameters for TWIST..
Statistical analysis
Mean and standard deviation values were calculated for all outcomes of the qualitative and quantitative assessment. Statistical analysis was performed using R, version R 4.2.2 (R Foundation for Statistical Computing, 2021) and Excel, version 16.66.1 (22101101), (Microsoft, Redmond, Washington, USA). For all tests, the statistically significant difference is set to p < 0.05.
Multilevel mixed-effect proportional-odds models were used for the qualitative analysis. Multilevel mixed-effect proportional-odds models include the scores as an ordinal dependent variable, the sequence (GRASP vs. TWIST) as the fixed factor and patient pseudo-ID and reader ID as random factors. The link function is logit (proportional odds).