|Home | About | Journals | Submit | Contact Us | Français|
Purpose e Cross-sectional imaging methods are important for objective evaluationof small intestinal inflammationinCrohn’sdisease(CD).The primary aim was to compare relative parameters of intestinal perfusion between contrast-enhanced ultrasonography (CEUS) and dynamic contrast-enhanced magnetic resonance enterography (DCE-MRE) in CD. Furthermore, we aimed at testing the repeatability of regions of interest (ROIs) for CEUS.
Methods This prospective study included 25 patients: 12 females (age: 37, range: 19–66) with moderate to severe CD and a bowel wall thickness>3mm evaluated with DCE-MRE and CEUS. CEUS bolus injection was performed twice for repeatability and analyzed in VueBox®. Correlations between modalities were described with Spearman’s rho, limits of agreement(LoA) and intraclass correlation coefficient(ICC). ROIrepeatability for CEUS was assessed.
Results s The correlation between modalities was good and very good for bowel wall thickness (ICC=0.71, P<0.001) and length of the inflamed segment (ICC=0.89, P<0.001). Moderate-weak correlations were found for the time-intensity curve parameters: peak intensity (r=0.59, P=0.006), maximum wash-in-rate (r=0.62, P=0.004), and wash-in perfusion index (r=0.47, P=0.036). Best CEUS repeatability for peak enhancement was a mean difference of 0.73dB (95% CI: 0.17 to 1.28, P=0.01) and 95% LoA from −3.8 to 5.3dB. Good quality of curve fit improved LoA to −2.3 to 2.8dB.
Conclusion The relative perfusion of small intestinal CD assessed with DCE-MRE and CEUS shows only a moderate correlation. Applying strict criteria for ROIs is important and allows for good CEUS repeatability
In Crohn’s disease (CD) the grading of disease activity has shifted from subjective clinical scoring systems towards more objective measurements, in combination with patient-reported outcomes 1. Endoscopy, although not completely objective, is often considered a gold standard for luminal disease in the colon, rectum, and sometimes the terminal ileum. However, endoscopy is of limited use in stricturing and proximal disease 2 and even well-recognized endoscopic scoring systems are not fully reliable 3. This calls for cross-sectional imaging methods with objective parameters of disease severity 1 4. Currently there is no single imaging modality as the gold standard for transmural disease of the small intestine 5.
The most consistent characteristic of disease activity on imaging is an increased bowel wall thickness of more than 3mm 4 6 7. Nevertheless, the intestinal wall may be thickened not only by active disease but also by fibrosis 7 8. Other features of inflammatory activity comprise ulcerations, T2-hypersignal, perimural signal, contrast enhancement, comb sign, enlarged lymph nodes, fistulas, abscesses and strictures described in the development of the MR intestinal activity score and MR enterography global assessment 7 9 10. Unfortunately, experts do not agree about the importance of the individual findings 11.
In recent classifications, increased contrast enhancement is considered a relevant marker of disease activity 7 9 12 13. This is in accordance with the characteristics of active inflammation including dilated leaking vessels 14 and neoangiogenesis 15. Additionally, microvascular density has been shown to correlate with intensity on dynamic contrast-enhanced ultrasound (CEUS) 16. Therefore, dynamic imaging techniques could potentially be used for evaluating disease activity and efficacy of treatment 7 17.
The 2 promising modalities to assess relative bowel wall perfusion are CEUS and dynamic contrast-enhanced MR enterography (DCE-MRE). However, there are significant differences in contrast behavior between modalities. MR gadolinium-based contrast agents are relatively small and exhibit extravasation over time, whereas CEUS gas-filled lipid-shell contrast acts as a true intravascular agent. The time-intensity curves (TICs) recorded from the former are therefore a combination of perfusion and permeability, rather than perfusion alone. The relationship between signal intensity and MRI contrast agent concentration is complex and depends on a number of parameters, such as the native tissue relaxation rate, relaxivity of the contrast agent, local field inhomogeneity and the applied flip angle and inversion-recovery time 18. US contrast agent on the other hand has a direct correlation with the signal intensity measured in dB 19. Since perfusion is difficult to measure if the bowel wall is less than 3mm thick 20, the parameters should only be used for grading disease activity or to follow treatment efficacy 6 21.
In the present study, we hypothesized that intensity and time parameters of the initial time-intensity curves correlate well between modalities as the amount of MR contrast which is extravasated during the initial pass is low.
The main objective of this study was to compare objective parameters of relative perfusion obtained with DCE-MRE and CEUS in patients with moderate to severe CD. Our secondary objectives were to test the repeatability of regions of interest (ROIs) for CEUS and to evaluate the inter-rater reliability of CD characteristics assessed with MRE and US.
This GCP monitored prospective double-blind observational study was approved by the Danish national authorities (2011-005886-19) and the local research ethics committee (1-10-72-340-12) for the off-label use of US contrast agents. All participants gave written informed consent before entering the study. Inclusion criteria were known CD with moderate to severe clinical activity based on either the CD Activity Index 22 (CDAI)>220 or Harvey Bradshaw Index 23 (HBI)>7. Furthermore, patients had to be ≥18 years of age and have a US-detectable intestinal segment with bowel wall thickness>3mm. Patients were excluded if they were pregnant, breastfeeding or had any contraindications for DCE-MRE or CEUS.
25 patients (mean age: 37 years; range: 19–66 years; 12 females) were recruited for the study from September 2012 to March 2014. Due to screening failure, 2 patients were excluded and another 2 were subsequently recruited to reach the desired inclusion of 25 patients, Fig. 1. The Montreal classification 24, CDAI, HBI, gastrointestinal symptoms, smoking status, and medical history were recorded and blood and stool samples were taken during the first visit. For full patient demographics, Table 1.
Participants were investigated after a 4-h fast. Ultrasonography (US) was performed by one physician (RW) with 2 years of experience with the procedure. The investigator was blinded to the MRE scan and biochemical results. However, the patients’ symptoms were known. An Acuson S3000 ultrasound machine with a 4–9MHz linear matrix transducer and a 1–6MHz curvilinear transducer was used (Siemens Medical Solutions, Malvern, PA). Color Doppler imaging (CDI) was set with a transmit frequency of 6.75MHz, gain 1dB, pulsed repetition frequency 1099, low wall filter of 2, and a scale of 6cm/s. The most severely inflamed bowel segment was identified based on wall thickness and the highest CDI signal score according to the Limberg classification 25. The total length of each affected segment, bowel wall pattern, presence of ulcers, stenosis and prestenotic dilatation were registered.
All CEUS scans were performed on the Acuson machine using the 9L4 probe. The settings were: fixed mechanical index of 0.06–0.08, dynamic range 80, frame rate of 10 per second, frequency 4MHz, and the focal zone beneath the bowel wall. Sulfur hexafluoride microbubbles (SonoVue®; Bracco Imaging, Milan, Italy) 2.4ml×2 were injected by trained nurses followed by a 5ml saline flush over 2s. Scans were recorded for 90s. The scan plane was kept constant and patients were instructed regarding gentle breathing. More than 5min after the first injection, the scan was repeated at the same spot, but in a different scan plane, to cover the segment in both the transverse and longitudinal axes. The chosen bowel segments were terminal ileum, neo-terminal ileum or proximal ileum. The location for CEUS was determined as the most inflamed area of the segment according to a prior classification 12.
Cine loop files were exported in DICOM format, re-linearized, and quantified on VueBox® 5.1 (Bracco Suisse SA, Geneva, Switzerland) as described earlier 17.
If possible, four ROIs were drawn using the following criteria: all ROIs had to be larger than 0.1cm2 and within the bowel wall at all times. Shapes and placement of ROIs were optimized to obtain a quality of fit (QoF) of the fitted curve larger than 90% or as high as possible. Built-in motion compensation was applied whenever beneficial. The first ROI was drawn as large as possible and typically covering the full bowel wall thickness of the anterior and posterior bowel wall avoiding the lumen. VueBox includes the possibility to apply a heat map for the parameters of interest. 3 additional ROIs were placed in areas with the highest peak enhancement according to the heat map and without overlapping, Fig. 2 and video (Online Resource). Analyses were then compared for repeatability between the largest ROI, the maximum peak ROI and the mean of the 3 latter ROIs, exhibiting a QOF>85%, entitled mean ROI. The average (log-converted) values of the best reproducible method were subsequently chosen for comparison with DCE-MRE results. Data post-processing was badge-analyzed by the same investigator more than 6 months after US and clinical scoring of the last patient to ensure effective blinding of data.
Patients were instructed to fast for 4 h and drink 1 l of oral contrast 1 h before the scan. Oral contrast comprised a suspension of 125ml mannitol 15% (Fresenius Kabi, Bad Homburg, Germany) in 875ml of tap water, 30ml psyllium HUSK® Fibre, and ice cubes. Peristalsis was suppressed by intravenous injection of 20mg hyoscine butylbromide (Buscopan®; Boehringer Ingelheim, Ingelheim, Germany) prior to non-dynamic sequences and repeated before contrast injection. Images were acquired using a 1.5T MR unit (Avanto; Siemens, Erlangen, Germany) with patients in the prone position. The intravenous contrast agent used was gadoterate meglumine (Dotarem®; Guerbet, Villepinte, France) with 0.2mg/kg bodyweight at 5ml/s followed by a 24ml saline flush. Patients were instructed to hyperventilate prior to a long breath hold followed by gentle breathing. The MR scanning protocols can be seen in Table 2.
Interpretation of MRE-based pathoanatomical data was performed individually by 2 radiologists with 9 (AHN) and 4 (VPH) years of experience, respectively. Both were blinded to the findings on US. The maximum wall thickness and total length of disease were described in continuous measurements for the most pathological bowel segment. Average values between readers were used for comparison with bowel wall thickness and length of involvement measured on US. The presence of mural edema, ulcers, wall enhancement pattern, perimural involvement, and presence of complications like stenosis and penetrating disease were also registered for the segment, based on the MRE global score 9 10.
The ROI for DCE-MRE analysis was placed in the bowel wall at the site of the largest wall thickness and highest enhancement within the same bowel segment examined by CEUS, using a custom-made program in MATLAB® (MathWorks®, Natick, MA). The ROI was manually moved in order to stay within the bowel wall during the dynamic series, Fig. 3. TICs were interpolated using a cubic spline. This interpolated curve was used to derive the parameters described in Table 3.
Statistical analysis was performed using Stata 13.1 for MAC (Stata Corp LP, College Station, TX). If no disease was observed on MRE, the bowel wall thickness was set to 3mm and the length to 0cm. Existing data in the literature were too scarce to allow for a power calculation. However, we estimated 25 patients to be sufficient. None of the linearized CEUS intensity data, expressed as arbitrary intensity units (AIU), followed a Gaussian distribution. Hence, they were log-converted as by default in US systems using 10×log10 (AIU) and expressed in dB for further analysis 12. Time parameters for both CEUS and DCE-MRE, C-reactive protein, and fecal-calprotectin were analyzed log-converted. Correlations between DCE-MRE and CEUS TIC parameters were described with Spearman’s correlation, since DCE-MRE data were slightly skewed 26 even with log-conversion. Correlation coefficients were interpreted as suggested earlier 26. CEUS repeatability was assessed with 95% limits of agreement (LoA), using a mixed effect model with independent residuals per ROI 27. Data for length of disease and MRE global score did not follow a Gaussian distribution regardless of log conversion. Hence only intraclass correlation coefficients (ICC) are reported for these data. P-values<0.05 were considered statistically significant. Data were not corrected for multiple testing. However, final conclusions were drawn having multiple testing in mind.
All but 2 patients had CEUS and DCE-MRE performed within the same day. The remaining 2 patients were scanned 4 days apart. All patients completed both examinations without adverse events or serious discomfort.
The thickest bowel wall segments had a mean of 7.9mm (range: 4–12mm) when assessed with US and 8.1mm (range: 4–14.5mm) when assessed with MRE. The mean difference was 0.22mm (LoA −4.3 to 3.9) and the corresponding ICC was 0.71 (0.44–0.86, P<0.001) (Fig. 4). The median length of the inflamed segment was 15cm (range 3–57cm) on US and 12cm (range 1–70cm) on MRE. The corresponding ICC was 0.89 (0.76–0.95, P<0.001).
Data from 3 MRE and 2 CEUS scans were excluded from further analysis, Fig. 1. All compared segments were either from the terminal ileum (n=19) or the ileum (n=1).
The total area under curve, including wash-in and wash-out for CEUS and wash-in and plateau-phase at 70s for DCE-MRE, had a low and insignificant correlation between the 2 methods (r=0.16, P=0.494). The wash-in area under curve also showed poor correlation (r=0.18, P=0.443). Likewise, the rise time and time to peak showed no correlation between modalities (r=0.11, P=0.659 and r=0.02, P=0.930, respectively). The slope and maximum slope for DCE-MRE and wash-in rate for CEUS correlated moderately well (r=0.60, P=0.005, and r=0.62, P=0.004), Fig. 5. The peak intensity and wash-in perfusion index determined by each of the 2 methods were moderately and moderately to weakly correlated (r=0.59, P=0.006 and r=0.47, P=0.036 respectively). No significant correlation was found between peak enhancement of CEUS and of DCE-MRE (r=0.41, P=0.076).
For CEUS, the smallest mean difference between 2 contrast injections was found for the maximum peak ROI. However, the narrowest limit of agreement was consistently found for the mean ROIs, Table 4. In a post hoc analysis restricted to ROIs with QoF>90%, or if 2 ROIs could not qualify for this, at least one ROI with QoF>85% and the other>90%, LoA could be further reduced, Fig. 6 and Table 4 for all LoA, Table 5 for QoF.
MRE interrater variability for bowel wall thickness showed an ICC=0.83 (0.66–0.92 P<0.001) and ICC=0.76 (0.51–0.89 P<0.001) for length of involvement. The mean difference was 1.2mm with 95% LoA from −3.8 to 3.6mm for wall thickness. For reproducibility on MR enterography global score, Table 6.
The present study compares CEUS and MRE for the description of the severity of ongoing small intestinal inflammation in CD. Even though correlations between basic pathoanatomical findings were good between the 2 modalities, our main finding was only a moderate to weak correlation when assessing relative changes in perfusion.
Since clinical activity scores for CD are poorly associated with the presence of active inflammation and equally poorly predict long-term outcome, their use should be supplemented by objective markers 1. Therefore, cross-sectional imaging is of paramount importance as an adjunct to endoscopy 4. Active inflammation is potentially treatable with effective medication but needs objective description and repeated follow-up to determine treatment response. Stenoses caused by fibrosis do not respond to medical treatment and need surgery 28. In contrast to fibrosis 29, active inflammation causes hyperemia and hyperperfusion 30 which may be quantified by CEUS and MRE.
A few previous studies have shown a significant correlation between dynamic contrast-enhanced cross-sectional imaging and clinical disease activity, biochemistry 31, or a combined score for response 32, the need for surgery 20, and change in medication 17 21 33. Other authors aimed at more objective endpoints like micro-vessel density 16 or mucosal healing or inactive disease defined by endoscopy 21. However, the studies do not agree about which TIC parameters are important. Romanini et al. [16, Saevik et al.17 and Horje et al.34 found a statistically significant difference for almost all TIC parameters and disease activity, whereas others only showed significance for time to peak 31, area under curve 32, or peak enhancement 33. In this present study, we found a significant correlation between the 2 modalities when describing peak and slope-related parameters but, surprisingly, not for area under curve, peak enhancement or rise time.
There is no consensus on how to perform or quantify intestinal perfusion measurement. Consequently, the heterogeneity between studies makes them difficult to compare or reproduce. For example, only a few authors have described the placement and analysis of ROIs for CEUS in detail 21 and only one group did log transformation of data before statistical analysis 35.
Several MRE studies use change in contrast enhancement as an indicator for disease activity 13 29 30. However, most studies have not applied a dynamic protocol and only use a few image acquisitions or the relative change over a predefined timespan after injection. Taylor et al. found an inverse correlation with slope of enhancement on MRE and micro-vessel density 36, which is the opposite of the finding by Romanini et al. using CEUS 16. These studies and our findings, showing a lack of good correlation, suggest that the 2 modalities measure somewhat dissimilar components of “perfusion”, with DCE-MRE TIC measurements being a mixture of perfusion and extravasation. Taylor et al. also found a direct correlation with slope of enhancement and disease duration and speculated that increased enhancement could be caused by ischemia and arteriolar stenosis 36.
In the present study, the interrater variability for structural MRE findings was comparable to those reported in previous studies 37. We only found a moderate correlation in wash-in rate and peak intensity could be established between DCE-MRE and CEUS. Lack of a strong correlation between modalities may likely be due to the dissimilar types and distribution nature of contrast agents, relatively poor MR time resolution, different field of view and scan planes and perhaps also the administration technique between modalities. In the optimal setting, absolute perfusion measurements of tissue blood flow, blood volume and mean transit time should be compared. However, this is complicated, even when using MR contrast agents which act as true intravascular agents, e.g., in cerebral perfusion 38.
The present study demonstrates the consequence of ROI selection in the quantification of perfusion in CD. Our data emphasize the importance of TIC QoF for reliability and reproducibility. Poor QoF 34, e.g., by fitting a burst-replenishment curve on a bolus injection examination 39, will obviously give unreliable results. We therefore recommend that curve fitting quality should be reported alongside test results in future publications. Also, using low perfused tissue as a reference will cause high uncertainty of the final results 40.
This study has some limitations. We did not apply Tofts (extended) model or any other model to reflect pharmacokinetic parameters, like absolute blood flow or permeability measures for pathological conditions 41, as our T1 measurements employing the variable flip angle technique gave unreliable results 42. As an alternative, we used the absolute signal difference technique instead, which has recently been shown to have a linear relationship to contrast agent concentration at low contrast concentrations 43 44.
Furthermore, CEUS was performed without deconvolution 45, thereby only providing semiquantitative measurements. Deconvolution is complex and relies on several assumptions 45 46 that are difficult to fulfil and thus rarely used in daily practice nor in scientific work. A method called bolus tracking and burst replenishment is described by Jirik et al. 47 but the repeatability is not yet established in humans.
Based on existing guidelines, bolus injection techniques were used for CEUS and DCE-MRE. A fixed dose and manual injection of SonoVue was chosen for CEUS quantification 48. For DCE-MRE, gadolinium dose was bodyweight-dependent and administered with an automatic pump. We chose the bodyweight-dependent dose over the fixed dose based on the general recommendation for MR contrast administration 13. We did not measure the exact length defined from an anatomical landmark, like the ileocecal valve, to ensure identical ROI location between modalities. Also, the CEUS scan planes were subjectively chosen and did not necessarily follow the standardized scan planes of MRI. The morphology of CD may vary even within short distances of the bowel and we cannot state that the exact same location was analyzed with the 2 methods 44 49. However, we attempted to do so by analyzing the same bowel segment and the thickest part of it in each patient. As a result of the disease complexity, grading disease activity should ideally involve all changes in segmental inflammation instead of narrow sampling as used in this study. However, complex scores limit use in everyday practice 50.
Since there are no guidelines on the optimal scan plane, 2 different scan planes were employed for the assessment of the repeatability of CEUS ROIs. Full repeatability of findings from the same segment in an identical scan plane along with reproducibility between investigators is still warranted. However, based on the present findings within patient repeatability seems acceptable for the clinical use of CEUS in CD, especially when applying strict criteria for size and QoF. Lack of strict criteria or the use of a low perfused tissue as the reference tissue will lead to poor reproducibility 35 40.
We chose to restrict the inclusion of patients to those with moderate to severe disease activity based on clinical symptoms. Investigating perfusion in a normal bowel wall is difficult because of peristalsis and a small ROI size results in poor QoF. However, clinical symptoms are often poorly correlated to objective signs of active disease. Based on wall thickness and biochemical findings, we covered the full disease spectrum of active small bowel disease.
In summary, there is only a moderate to weak correlation between CEUS and DCE-MRE slope-related and peak intensity parameters in CD. This is likely to be caused by the inherently different nature of the contrast agents and scanning modalities. Additionally, we have elucidated the importance of quality of fit for ROI selection in CEUS. The value of perfusion measurements as activity assessment in CD still remains to be clarified and validated against more objective endpoints.