A search in our cardiovascular MR imaging database (2014-2020) yielded 337 patients with MR diagnosis suggestive of pericarditis. Incomplete studies (n=51) or studies with insufficient coverage of the liver parenchyma of the T1/T2 maps (n=25) were excluded. Next, patients with concomitant liver pathology (e.g. liver cirrhosis) were excluded (n=12). This yielded 249 cardiovascular MR studies. In 25 studies (10%), increased ventricular coupling was found at free-breathing real-time cine MR imaging, defined as inspiratory septal inversion, and/or increased excursion of the ventricular septum between inspiration and expiration . These patients were defined as pericarditis with constrictive physiology (CP+group). We randomly selected 20 patients in the cohort without increased ventricular coupling and defined them as pericarditis without constrictive physiology (CP- group). Finally, we selected as control group 30 subjects with normal cardiovascular MR imaging findings, lack of increased central venous pressure at TTE and normal liver enzymes (control group). The study was conducted in accordance with the Declaration of Helsinki and was approved by the research ethics committee of the hospital. Because of the retrospective design, informed consent by the patient was waived.
MR imaging protocol.
Studies were performed on a 1.5T scanner (Ingenia, Philips Healthcare). The protocol consisted of scout images, T1 and T2 weighted fast spin-echo imaging, breath-hold steady-state free precession (SSFP) cine imaging along the different cardiac axes, T1 and T2 myocardial mapping, late gadolinium enhancement (LGE) imaging and post-contrast T1 mapping, and free breathing real-time imaging in midventricular cardiac short-axis to assess ventricular coupling. T1-weighted imaging was performed using a T1-weighted black-blood turbo spin-echo (TSE) in cardiac short-axis. Typical imaging parameters were flip angle, 90°; TR/TE 1 beats/8.6 ms; sense factor, 2.1; TSE factor, 24 with asymmetric profile order; matrix 320x266; FOV, 320x266 mm (incl. fold-over suppression, 380 mm); slice thickness, 8 mm; acquired resolution, 1.05x1.32 mm; reconstructed resolution, 0.5x0.5mm; phase percentage, 70-80%, bandwidth, 490 Hz, and black blood preparation (TI 180 ms) and thickness 20 mm. The ventricles were completely encompassed using 6 to 10 slices (variable gap). T2-weighted imaging was performed using a black blood short-tau inversion-recovery (STIR) turbo spin-echo (TSE) in cardiac short-axis. Typical imaging parameters were TR/TE 2 beats/85 ms; sense factor, 2; TSE factor, 33 with linear profile order; halfscan factor, 0.8; matrix 256x164; FOV, 320x280 mm (incl. fold-over suppression, 560 mm); slice thickness, 8 mm; acquired resolution, 1.4x1.7 mm; reconstructed resolution, 0.7x0.7 mm; phase percentage, 70-80% and black blood inversion preparation (TI 180 ms) and thickness 20 mm. The ventricles were completely encompassed using a slice gap of 2 mm. For assessment of LV dimensions and function balanced SSFP breath-hold cine images were acquired in the following orientations: vertical and horizontal long-axis, short-axis and left ventricular (LV) outflow tract view. Typical imaging parameters were: repetition time (TR) / echo time (TE), shortest/shortest (e.g. 3.6/1.8 ms); sense factor, 2 or compressed sense (CS), 2.5; halfscan factor, 0.625; flip angle, 60°; matrix 208x160; field of view (FOV) 350x274 mm (with fold-over suppression 360 mm); acquired resolution, 1.7x1.7 mm; reconstructed resolution, 1.0x1.0 mm; slice thickness, 8 mm; pixel bandwidth, 954 Hz, and number of phases, 30, phase percentage, 67%. In cardiac short-axis, the left ventricle was encompassed completely using a slice gap of 2 mm. These cine MRI images were acquired starting from one minute post-contrast administration, followed by a three-chamber cine view using similar sequence parameters as above. LGE cardiovascular MR imaging was performed using a breath-hold 3D turbo-field-echo inversion recovery sequence (IR-TFE) and/or phase-sensitive inversion-recovery (PSIR) sequence in short-axis, vertical and horizontal long-axis. LGE was started 10 minutes after injection of 0.15 mmol/kg of commercially available gadolinium-chelates (gadobutrol, Gadovist Bayer). Typical imaging parameters: TR/TE, shortest/shortest; sense factor, 2 or CS, 6.5; matrix sense 256x164 or CS 168x159; FOV, sense 350x350 or CS 320x320; acquired resolution, sense, 1.4x2.2x10mm or CS, 1.9x2.0x10mm; reconstructed, 1.3x1.3x5mm; flip angle, 15°. A Look-Locker sequence was used to determine the optimal inversion time to null myocardium (IR-TFE) or blood (PSIR). T1 mapping images were acquired in 4-chamber long-axis and midventricular (in most also basal and apical) 2D short-axis orientations using the modified look-locker inversion recovery (MOLLI) sequence. For native T1 mapping MOLLI we used a 5s(3s)3s protocol (hsec inv pulse, minimum inversion times 110 and 350ms, balanced SSFP readout with TR:2.0ms, TE:0.9ms, flip angle: 35°, slice thickness: 10mm, acquisition matrix: 152x150 reconstructed to 256x256, SENSE factor: 2, BW: 1082 Hz, FOV: 300x300mm, pixel size: 2.0x2.0mm, reconstructed pixel size 1.2x1.2mm). At 10 min T1 mapping was repeated using a 4s(1s)3s(1s)2s protocol (minimum inversion times about 110, 230 and 350ms, all other parameters as for MOLLI native). T2 mapping was performed using a gradient-echo spin-echo sequence (GRASE) (9 echoes with TE:10-100 ms, TR/TE, 1 beat/shortest; sense factor,2; matrix 152x139; FOV, 300x300mm (with fold-over suppression 420 mm); slice thickness, 10 mm; acquired resolution, 2.0x2.0 mm; reconstructed resolution, 1.0x1.0 mm). Typically, 3 short-axis T2 maps were obtained covering basal, mid and apical part of the LV. Finally, real time cines images were acquired over 250 dynamics during deep respiration using a single-shot balanced TFE sequence with specific parameters: repetition time (TR) / echo time (TE), shortest/shortest (e.g. 1.9/0.8 ms); sense factor, 2; halfscan factor, 0.625; flip angle, 50°; matrix 96x65; FOV 360x304 mm (no fold-over suppression); acquired resolution, 3.8x4.7 mm; reconstructed resolution, 2.5x2.5 mm; slice thickness, 10 mm; pixel bandwidth, 3519 Hz.
Images were analyzed off-line using IntelliSpace Portal software (Philips Healthcare) by an European Association Cardiovascular Imaging (EACVI) level-III CMR reader. Pericardial thickness was measured, and the images were assessed for presence of effusion, edema (T2-weighted imaging) and inflammation (LGE imaging). Cine images were used to calculate LV and right ventricular (RV) volumes, function, and LV mass. Left atrial and right atrial area was contoured on the end-systolic horizontal long-axis image. The diameter of the IVC and descending aorta were measured at the level of the diaphragm. Hepatic vein size was visually assessed as normal or dilated. Presence of pleural fluid and ascites were evaluated. The early-diastolic inspiratory shape of the ventricular septum was described at real-time cine imaging as normal, flattened or inverted, and the total shift of the septum between inspiration and expiration was measured . As mentioned above, this information was used to define the CP+ and CP- group.
On the T1 and T2 map, myocardial T1 and T2 was measured in the ventricular septum, and a standardized method was used to calculate myocardial ECV. For the liver, a region of interest (ROI) was drawn freehand with a minimum number of 500 pixels. The ROI was typically chosen in the subcapsular liver parenchyma in the vicinity of the heart. Care was taken not to include the centrally located liver vessels. A similar formula as for the myocardium was used to calculate ECV liver (23), but the RV cavity instead of the LV cavity was used to measure the T1 of blood. Intra- and inter-reader reproducibility of T1/T2 liver mapping was assessed in 30 randomly chosen patients (10 from each group).
Statistical analysis was performed using R programming language for statistical computing v.4.0.0. (The R-Foundation for Statistical Computing). Histograms and Q-Q plots with graphical bootstrap were used to test the normality of the data distributions. Continuous variables were expressed as mean ± SD or medians with interquartile ranges (IQR) as appropriate. Categorical variables were expressed as frequency with percentage. One-way analysis of variance (ANOVA) or the Kruskal-Wallis test were used to assess differences among the different groups, and post hoc pairwise comparisons with Bonferroni adjustment were performed by using the Student t test or Wilcoxon Rank Sum test, respectively. The Chi-square test or Fisher’s exact test was used to compare categorical variables. ROC curve analysis was performed to examine differences in performance of each variable for prediction of clinical symptoms of RHF (i.e., elevated central venous pressure, lower limb edema, NYHA class >2). The 95% confidence interval (CI) of each area under the curve (AUC) was estimated and compared with the R “pROC” package using the DeLong method . The optimal cut-off points for each continuous predicator was defined from the ROC analysis in terms of maximum sensitivity and specificity by using the Youden index . The Youden index (J) is defined as , with J=1 indicating complete separation and J=0 indicating complete overlap of distributions. The amount of overlap in values among different patient group was assessed using density plots and quantified using the R “overlapping” package with bootstrap estimates . An adjusted 2-tailed p-value <0.05 was considered to indicate a statistically significant difference. Intra-reader and inter-reader agreement of multiparametric liver mapping were assessed by using intraclass correlation coefficients (ICCs) and Bland-Altman analysis with 95% limits of agreement (LOAs). ICCs of greater than 0.75 and of 0.4-0.75 indicate strong and average agreement, respectively. A difference between ICCs was considered to be statistically significant when there was no overlap between their respective 95% CI limits.
Human subject research.
This retrospective, observational study was approved by the Ethics Committee Research UZ/KU Leuven (S64242). Because of the retrospective design, informed consent by the patient was waived by the Ethics Committee Research UZ/KU Leuven.