Whether diffusion-weighted imaging (DWI)-MRI is of value in detecting and assessing inflammation of ileal Crohn's disease (CD) remains poorly investigated.
Whether diffusion-weighted imaging (DWI)-MRI is of value in detecting and assessing inflammation of ileal Crohn's disease (CD) remains poorly investigated.
To compare DWI-MR enterography (MRE) with conventional MRE in estimating inflammation in small bowel CD, to determine an apparent diffusion coefficient (ADC) threshold to differentiate active from non-active lesions and to assess inter-observer agreement.
Thirty-one CD patients from the Clermont-Ferrand IBD unit with ileal involvement were consecutively and prospectively included between April and June 2011. All patients underwent DWI-MRI to detect the digestive segment with the most severe lesions, which was then used to calculate the ADC. Qualitative and quantitative results were compared with conventional MRE including MaRIA (Magnetic Resonance Index of Activity) score calculation and independent activity predictors (wall thickening, oedema, ulcers). Each examination was interpreted independently by two radiologists blinded for clinical assessment.
Seventeen patients (54.8%) had active CD as defined by the MaRIA score ≥7. DWI hyperintensity was highly correlated with disease activity evaluated using conventional MRE (P = 0.001). Qualitative analysis of DW sequences determined sensitivity, specificity, positive predictive value and negative predictive value as 100%, 92.9%, 94.4% and 100% respectively. Quantitative analysis using a cut-off of 1.6 × 10−3 mm²/s for ADC yielded sensitivity and specificity values of, respectively, 82.4% and 100%. Inter-observer agreement was high with regard to DWI hyperintensity (κ = 0.69, accuracy rate = 85.7%) and ADC (correlation = 0.74, P < 0.001, and concordance = 0.71, P < 0.001).
DWI-MR enterography is a well-tolerated, non-time-consuming and accurate tool for detecting and assessing inflammation in small bowel Crohn's disease.
Crohn's disease (CD) is a chronic relapsing and destructive disorder that can involve the entire length of the digestive tract. While colonoscopy is considered as the gold standard for assessing colonic inflammatory activity, small bowel investigation requires magnetic resonance imaging (MRI) or computed tomography (CT-scan) to detect involvement. Wireless capsule endoscopy is an alternative tool in the event of nonstricturing disease. Owing to its non-ionising characteristics, MRI should be preferred to CT-scan especially in CD patients, who will undergo repeated examinations because of the natural history of the disease. In addition, the emergence of new therapeutic goals such as mucosal healing, preventing digestive damage or limiting disability[5, 6] has led IBD physicians to look for new non-invasive modalities to monitor the course of the disease. MRI is potentially a very promising procedure.
The MaRIA (Magnetic Resonance Index of Activity) score was recently prospectively validated and so can be considered as the benchmark for assessing inflammatory activity using MRI for ileocolonic CD.[7, 8] Some patients in the IBD centres undergo MRE preceded by bowel cleansing and rectal preparation, which can greatly decrease tolerability.
Diffusion-Weighted Imaging (DWI) is a method deriving its image contrast from differences in the motion of water molecules between tissues. It has been established as an important tool in the detection of early ischaemic changes in clinical settings. DWI also has the capability to detect inflammatory foci.[10-12] In body imaging, DWI is used in solid organs such as liver, pancreas, spleen and kidney.[13-15] Diffusion-weighted imaging (DWI) is a promising diagnostic tool in CD that has yet to be properly investigated. No threshold of apparent diffusion coefficient (ADC), the quantitative parameter of DWI, has been established for defining inflammation in CD. DWI in colonic CD has been assessed in two studies.[16, 17] Assessment of its use in detecting small bowel inflammatory activity in CD is restricted to one prospective study, in which MRI was compared with conventional barium radiology or surgery, and one retrospective study.
The aim of this study was (i) to compare the ability of DWI-MRE with no oral or rectal preparation and that of the standard MRE including gadolinium injection and MaRIA score calculation to accurately detect and assess small bowel inflammation, (ii) to determine an apparent diffusion coefficient (ADC) threshold in detecting inflammatory activity and (iii) to assess inter-observer agreement.
The study was performed in accordance with the Declaration of Helsinki, Good Clinical Practice and applicable regulatory requirements. Study ethics approval was obtained on (CECIC Rhône-Alpes-Auvergne, Grenoble, IRB 5921).
We performed an observational study of a single-centre cohort in which standardised evaluation was used by experienced clinicians for all patients. Thirty-one patients treated at the IBD unit of the Clermont-Ferrand University Hospital, France were consecutively and prospectively included between April and June 2011. All patients had established diagnosis of CD according to Lennard-Jones criteria. The patients enrolled in the study had either symptomatic disease or were undergoing imaging to monitor CD activity under treatment. Patients with claustrophobia or other common MRI contraindications such as pace maker, metallic intraocular foreign body, or allergy to gadolinium or severe renal failure (MDRD <30 mL/min) were excluded. All included patients underwent complete clinical assessment, had blood samples obtained for cell blood counts, C-reactive protein (CRP) measurements and evaluation of renal function (MDRD). Clinical disease activity was assessed using the Crohn's Disease Activity Index (CDAI). All patients underwent MRI within 72 h. Each examination was interpreted independently by two radiologists who were unaware of clinical or biological data.
On the day of MRI-DWI, patients had to have been fasting for at least 4 h before the examination. All MRI were performed with a 1.5 Tesla (T) MR unit (Optima MR450w; General Electric HealthCare, Fairfield, CT, USA). Patients were placed in the supine (more comfortable) position in the MR imager. A combination of two surface coils was used for signal reception to allow coverage of the whole abdominal area. To reduce bowel peristalsis, 1 mg of glucagon was injected intravenously after dynamic sequences. To achieve an adequate distension of the distal ileum, 1000 mL of PEG solution (Fortrans, Ipsen Pharma, Paris) was given to all patients, 25 min before MR in patients with previous small bowel resection or 40 min before in non-operated patients. The included patients did not receive any bowel cleansing or rectal preparation, to improve tolerability. The acquisition protocol used is described in Table 1. It included a Cine coronal True FISP (FIESTA, General Electric), a two-dimensional (2D) FIESTA sequence in the coronal plane and 2D FIESTA fat sat (FS) sequence in the axial plane, a 2D Single-Shot TSE/FSE (SSFSE, General Electric) sequence in the axial and coronal planes, a diffusion-weighted sequence in the axial and coronal planes, Volume Interpolated GRE (LAVA, General Electric) sequences before and after intravenous administration of 0.2 mL/Kg body weight of Gadobemate Dimeglumine (Multi-Hance, Bracco Imaging, Evry, France) for dynamic study in the coronal plane (30, 70, 130 and 210 s after intravenous injection) and finally a T1 LAVA sequence in the axial plane late after injection. The diffusion-weighted sequence used a diffusion factor b fixed at 0 and 800 s/mm2 in the axial plane because these values represent the best compromise between SNR (signal to noise ratio) and lesion detection sensitivity on our MRI system. The entire examination lasted about 30 min.
|Sequences||Plane||FOV (cm)||TE (ms)||TR (ms)||Flip angle (°)||Slice thickness (mm)||Acquisition duration (s)|
|2D FIESTA FS||Axial||34||2||4.8||85||6.5||40|
|T1 FS LAVA||Coronal||42||1.9||4||12||3.2||240|
|T1 FS LAVA||Axial||40||1.9||4||12||3.8||45|
Image analysis was performed by one experienced radiologist (VP) and one junior radiologist (AJ) using a dedicated postprocessing workstation (Advantage Windows Work Station; General Electric, USA). The small bowel was examined to detect the segment with the most severe lesions on the basis of the following criteria: bowel wall thickness (mm), presence of mucosal ulcers (defined as deep depressions in the mucosal surface), presence of mural oedema (hyperintesity on T2-weighted sequences of the bowel wall relative to the signal of the psoas muscle), presence of enlarged regional mesenteric lymph nodes, presence of fistula or abscess, and relative contrast enhancement (RCE) of the intestinal wall. Quantitative measurements of wall signal intensity (WSI) were obtained from the areas with the greatest thickening [region of interest (ROI)] before and after intravenous injection of gadolinium (70 s). RCE was calculated according to the following formula: RCE = [(WSI postgadolinium−WSI pregadolinium)/(WSI pregadolinium)] × 100 × (SD noise pregadolinium/SD noise postgadolinium). Dynamic sequences increased the detection of lesions, especially stenosis. The segment with the most severe lesions was used to calculate the MaRIA score [=1.5 × wall thickening (mm) + 0.02 × RCE (relativecontrast enhancement) +5 × oedema + 10 × ul-cers].[7, 8] Quantitative analysis used measurement of apparent diffusion coefficient (ADC) in the axial plane by a ROI placed on the largest area covering the bowel wall in the ADC map. Measurement of ADC during DWI sequences and conventional MRE sequences with gadolinium injection associated with MaRIA score calculation were performed by two blinded radiologists independently, with a wash-out period of at least 48 h.
Quantitative variables are given as means and standard deviation (s.d.) or as medians in the case of an abnormal distribution. Proportions are expressed as percentages and 95% confident intervals (CIs). Differences in quantitative measures were tested by Student's test or Kruskal Wallis' test if the conditions of Student's test were not met. Qualitative variables were compared by exact Fisher's test or Chi-squared test.
A threshold was determined by calculating receiver operating characteristic (ROC) curves. All correlations were studied using Spearman's nonparametric correlation coefficients (r and P-value respectively). Inter-observer agreement between paired evaluations of MR by two radiologists (VP and AJ) was performed with kappa statistics for comparisons of qualitative variables and Pearson correlation coefficient or Lin concordance correlation coefficient for quantitative data. A P value of 0.05 was considered significant. Calculations were done with stata v10 (STATA Corp., College Station, TX, USA).
All the data retrieved from the 31 patients undergoing MRI-DWI were suitable. Baseline characteristics of the patients are given in Table 2.
|Male gender (n, %)||11, 35.5|
|Female gender (n, %)||20, 64.5|
|Age at diagnosis (years, range)||26 (12–63)|
|Median disease duration (years, range)||5 (1–21)|
|Previous intestinal resection (n, %)||13, 41.9|
|5-ASA (n, %)||9, 29.0|
|Corticosteroids (n, %)||12, 38.7|
|Immunosuppressives (n, %)||10, 32.3|
|Anti-TNF agents (n, %)||8, 25.8|
|Median CDAI (points, range)||103 (23–298)|
|Clinical remission (n, %)||23, 74.2|
|Active disease (n, %)||8, 25.8|
|Median CRP (mg/L)||4.5 (0.1–187.6)|
|CRP <5 mg/L (n, %)||16, 51.6|
|CRP ≥5 mg/L (n, %)||15, 48.4|
The quality of the MRI examination was considered as high in 24 patients (77.4%) and intermediate in 7 (22.6%). Distal ileum distension was not fully achieved in 9 patients (29.0%) and bowel peristalsis induced motion artefact in 3 (9.6%). Only one patient had poor tolerance during the examination. Of note, there was no difference in the choice of the more severe segment between the two observers. Active lesions were reported in 23 patients (74.2%) defining as a MaRIA score ≥7. Disease location was ileal or ileocolonic, respectively, in 21 (67.7%) and 2 patients (6.5%). No lesion was found in 8 patients (25.8%). MRI characteristics are listed in Table 3 and examples are given in Figure 3.
|Population (n = 31)|
|Median wall thickening (mm, range)||4 (1.5–9)|
|T2-weighted hyperintensity (n, %)||11, 35.5|
|DWI hyperintensity (n, %)||18, 58.1|
|Median ADC (10−3 mm2/s, range)||1.7 (0.8–2.8)|
|Median relative contrast enhancement||176 (60–322)|
|Ulcers (n, %)||4, 12.9|
|Median MaRIA score||12 (3.5–31.8)|
|Qualitative MaRIA score (≥7)||17, 54.9|
MRI-DWI results were correlated with MRI activity defined as both qualitative (MaRIA score ≥7) (P = 0.0001) and quantitative criteria (P = 0.0001). In detecting active disease, DWI hyperintensity had sensitivity, specificity, predictive negative value and predictive positive value of, respectively, 100% (95% CI: 83.9–100.0), 92.9% (95% CI: 66.1–99.8), 94.4% (95% CI: 77.2–99.9) and 100% (95% CI: 75.3–100.0). ADC was inversely correlated with the quantitative MaRIA score (r = −0.77; P = 0.0001) (Figure 1). Using ROC curves (Figure 2), we determined 1.6 × 10−3 mm²/s as the threshold for differentiating active from non-active disease with high sensitivity (82.4%) and specificity (100%). Area under the ROC curve was 0.96 (95% CI: 0.89–1.00). ADC was also correlated with the different components of the MaRIA score such as wall thickening (−0.74, P < 0.0001), RCE (−0.62, P < 0.0001) and oedema (P = 0.0004).
A linear regression was performed using both the factors included in the MaRIA score (except RCE, which needs gadolinium injection) and ADC obtained from DWI sequences to create an MRI activity index for small bowel CD. No difference was found between the original MaRIA score (R² = 0.998) and the new MRI-DWI score (R² = 0.989) (Figure 3). Coefficients were obtained using the linear regression as given in Table 4, thus DWI-MaRIA could be calculated as:−1.321 × ADC (mm2/s) + 1.646 × wall thickening (mm) + 8.306 × ulcers +5.613 × oedema + 5.039.
|Coefficient||Standard error||95% CI|
|ADC||−1.321||0.473||−2.294 to −0.348|
The kappa index and the accuracy rate were, respectively, 0.94 (S.E.: 0.17) and 97.1% for the MaRIA score, 0.77 (S.E.: 0.16) and 88.6% for oedema, 0.80 (S.E.: 0.17) and 94.3% for ulcers and 0.69 (S.E.: 0.17) and 85.7% for DWI hyperintensity.
Pearson correlation coefficient and Lin concordance correlation coefficient were, respectively, 0.90 (P < 0.001) and 0.89 (P < 0.001) for MaRIA score, 0.83 (P < 0.001) and 0.81 (P < 0.001) for wall thickening, 0.65 (P < 0.001) and 0.54 (P < 0.001) for RCE and 0.74 (P < 0.001) and 0.71 (P < 0.001) for ADC.
As the small bowel is difficult to reach with conventional endoscopy, several non-invasive tools have been developed in the last two decades in the investigation of CD. MRI is one method that is increasingly used. MR Enterography (MRE) to diagnose, to assess disease extent, to detect inflammation or to show complications such as stenosis or fistula has been widely assessed. Recently Rimola et al. developed the MaRIA score, which is able to assess inflammation in ileal and colonic CD.[7, 8] This index underwent an internal and external validation in an independent cohort. However, the MaRIA score is a validated index for describing the severity of inflammation, but it is not a gold standard for describing CD, which is much more complex. The MaRIA score is a first attempt at describing severity, but does not reflect the overall burden of disease. Despite this limitation, it can be considered as the most validated tool for evaluating inflammation by MRI in CD.
The efficacy of conventional MRE has been widely studied in CD,[7, 8, 22-31] but little attention has been paid to the interest of DWI.[16-19] This study is the first observational prospective study to assess the accuracy of DW sequences in detecting and assessing small bowel inflammation in CD by comparing findings of DWI-MRE and conventional MRE including gadolinium injection and MaRIA score calculation. Two studies have evaluated DWI in colonic CD. The first was a small sample size retrospective study, and the other, involving 96 patients, was performed in the referral centre of Nancy, France. The latter yielded disappointing results compared with those reported in ulcerative colitis, because the sensitivity and the specificity in detecting colonic inflammation in CD were, respectively, 58.3% and 84.5%. Only one study has investigated whether adding DW sequences was of value in the detection of small bowel inflammation in CD. Unfortunately, data were not compared with those of conventional MRE and two types of gold standard were used, conventional barium radiology or histological examination of surgical resections performed within 1 month before or after MRI. The authors concluded that the accuracy of DWI was high in evaluating disease activity among the 31 patients enrolled, especially in the small bowel, and that ADC facilitated quantitative analysis of disease activity.
Unlike in most studies using endoscopic findings as gold standard, we took conventional modalities of MRI as our reference. Although we are aware that this choice is debatable, we suggest that it could be an interesting and original approach for several reasons. As the MaRIA score is a very recently validated tool, all of the trials in the same field were performed before the validation of this index. It seemed easier, more accurate and more reproducible to compare one segment using two methods of MRI than to compare imagery with endoscopy. MRI modalities are more comfortable and acceptable for patients than endoscopy, which symptomatic CD patients may decline to undergo. Bowel cleansing is highly unpleasant, and importantly, intubation of the distal ileum can only be completed in 80% of the patients in such series. Recent evidence indicates that cross-sectional imaging techniques are highly accurate in the assessment of mucosal lesions and are reliable alternatives to ileocolonoscopy. In addition, they offer certain advantages, including the assessment of extramural complications, such as the presence of fistulas and abscesses, and determining the functional repercussion of a stricturing lesion. Of note, our patients included those who had been operated on and thus reflected more closely a real-life situation. Unlike the Spanish research group that created the MaRIA score, we used a 1.5 T MR unit. Although a 3 T MR unit could have theoretically a better spatial resolution, to our knowledge, no specific data are available comparing 1.5 T and 3 T MR units in digestive disease. In addition, the 1.5 T unit is currently more widespread in tertiary IBD units.
In the one quarter of patients who did not achieve full intestinal distention, no analysis was incomplete because of insufficient ileal loop distension. High image quality was reported in almost two-thirds of the examinations. Although almost one-third were of intermediate quality, none of the radiologists considered the images as of poor quality.
Among our 31 CD patients, we observed that both ADC and DWI hyperintensity were strongly correlated with the MaRIA score retrieved from conventional MRI sequences. We determined a threshold (ADC = 1.6 mm²/s), which can accurately separate active from non-active CD. A previous study involving 35 patients reported mean ADC values of 1.61 and 2.56 (P < 0.0001) for, respectively, active and non-active small bowel CD. We observed also that ADC is highly correlated with all items previously reported as of good value in detecting inflammation i.e. oedema, ulcers and wall thickening. It would be interesting to know if DWI parameters were correlated with biomarkers such as CRP and faecal calprotectin.
We attempted to create an index integrating DWI parameters such as ADC to assess inflammation in small bowel CD. Details of the MaRIA-DWI index are given in Table 4. As this index was strongly correlated with the MaRIA score (Figure 4), we could consider that DWI is a useful tool in evaluating inflammation in small bowel CD. Diffusion-weighted MR imaging (DWI) reflects molecular diffusion, which is the thermally induced Brownian motion of water molecules, without the administration of contrast material. These methods do not significantly extend the duration of examination and could avoid gadolinium injection, which is invasive, can induce nephrogenic systemic fibrosis and is contraindicated in case of severe renal failure. In our experience, a protocol of DWI-MRE without injection would be shorter than a conventional MRE with gadolinium injection.
The two radiologists were specifically chosen to include one experienced practitioner with high expertise in IBD (V.P.) and one less experienced junior radiologist (A.J.), to assess the reproducibility of DWI. We recorded good inter-observer agreement for the variables evaluated, which are independent predictors of activity. Our results seem comparable to those of Rimola et al., who reported a kappa index for oedema of 0.86 (P < 0.001), and of 0.91 (P < 0.001) for ulcers. The intraclass correlation coefficients were r = 0.96 (P < 0.001) for wall thickness, and r = 0.89 (P < 0.001) for contrast signal intensity. Hence, these data demonstrate that DW sequences allowed easy and reproducible measurements, which could lead to a greater standardisation of MRI reports in IBD, especially in clinical trials.
Several factors could be considered as strengths in our study. The protocol is novel in that it is the first observational prospective study to compare DWI and conventional MRI using the MaRIA score to assess inflammation in small bowel CD. MRE required bowel cleansing and rectal preparation in studies made at a Spanish referral centre.[7, 8] Our protocol was shorter, to improve tolerability for patients, but without limiting final analysis. However, in our experience, small bowel distension with PEG (500–1000 mL) remains essential and cannot be avoided to date.
Our study has certain limitations. We could be criticised for not choosing endoscopy as the gold standard, but we preferred to use conventional MRI because the comparison of two cross-sectional imaging modalities was easier, more accurate and remains the final goal of our study. However, adding an endoscopic assessment to this protocol would probably be of great value. Unlike in most studies about MRI modalities in CD, we did not perform DWI measurement in each segment of the digestive tract and focused only on the segment with the most severe lesions. Of note, there was no variability between the two radiologists in defining the most severe inflammatory segment. In the era of mucosal healing or digestive damage,[3, 4] the aim of DWI-MRE is to detect inflammation, so as to monitor and optimise therapy. Thus, the presence of intense inflammation in at least one of the digestive segments makes the patient eligible for therapeutic tailoring.
In conclusion, DWI-MRE is a well-tolerated, non-time-consuming, accurate and reproducible tool for detecting and assessing inflammation in small bowel CD. Further investigations are necessary to complete an external validation of the MaRIA-DWI index and thereby to avoid gadolinium injection in current practice.
Guarantor of the article: A. Buisson.
Author contributions: Anthony Buisson designed the research study and wrote the paper. Anthony Joubert performed the research, collected and analysed the data and contributed to the design of the study. Pierre-François Montoriol and David Da Ines contributed to the design of the study. Constance Hordonneau proofread the article and contributed to the design of the study. Bruno Pereira performed the statistical analysis. Jean-Marc Garcier contributed to the design of the study. Gilles Bommelaer proofread the article and contributed to the design of the study. Virginie Petitcolin performed the research, collected and analysed the data, and designed the research. All authors approved the final version of the manuscript.
Declaration of personal and funding interests: None.