To determine relative diagnostic value of MR diffusion and perfusion parameters in detection of active small bowel inflammation in patients with Crohn's disease (CD).
To determine relative diagnostic value of MR diffusion and perfusion parameters in detection of active small bowel inflammation in patients with Crohn's disease (CD).
We reviewed 18 patients with active CD of terminal ileum (TI) who underwent MR enterography (MRE; including dynamic contrast enhanced MRI and diffusion-weighted MRI). Conventional MRI findings of TI were recorded. Regions of interest were drawn over TI and normal ileum to calculate apparent diffusion coefficient (ADC), the volume transfer constant (Ktrans) and the contrast media distribution volume (ve). Receiver operating characteristic analysis was used to determine their diagnostic performance.
Among conventional MR findings, mural thickening and increased enhancement were present in all actively inflamed small bowel. Ktrans, ve, and ADC values differed significantly between actively inflamed TI and normal ileum (0.92 s−1 versus 0.36 s−1; 0.31 versus 0.15 ± 0.08; 0.00198 mm2/s versus 0.00311 mm2/s; P < 0.001). Area under the curve (AUC) for Ktrans, ve, and ADC values ranged from 0.88 to 0.92 for detection of active inflammation. Combining Ktrans and ADC data provided an AUC value of 0.95.
Dynamic contrast-enhanced MRI (DCE-MRI) and diffusion-weighted imaging (DWI) provide quantitative measures of small bowel inflammation that can differentiate actively inflamed small bowel segments from normal small bowel in CD. DWI provides better sensitivity compared with DCE-MRI and combination of ADC and Ktrans parameters for analysis can potentially improve specificity. J. Magn. Reson. Imaging 2011;. © 2011 Wiley-Liss, Inc.
CROHN's DISEASE (CD) IS a chronic inflammatory disease of unknown origin involving the entire gastrointestinal tract, affecting more than half a million Americans and is increasing in frequency in many areas of the world, including the United States (1). Recently, MR imaging has emerged as a valuable cross-sectional imaging tool in the detection of bowel abnormalities and evaluation of disease activity in CD (2–6). In the recent prospective studies, MR enterography (MRE) was found to have a similar accuracy, area under the receiver operating characteristic (ROC) curve and sensitivity for detecting active inflammation in CD compared with CT enterography (CTE; 7–9). In addition, MRE has the potential advantage of providing functional and quantitative information about bowel wall (e.g., diffusion, perfusion, motility) that cannot be obtained by CT.
Diffusion-weighted imaging (DWI) and quantitative dynamic contrast-enhanced MRI (DCE-MRI) have been investigated recently in the assessment of bowel inflammation in CD (10–13). Initial results from studies evaluating both small bowel and colon appear promising despite their limitations such as small sample size or an imperfect reference standard (i.e., barium enema). DWI with parallel imaging allows detection of bowel inflammation in patients with CD (11, 12). In two different small series (11 patients each), quantitative analysis of DCE-MRI data could differentiate inflamed bowel from normal bowel and certain MR perfusion parameters derived using this analysis correlated with disease chronicity and angiogenesis in patients with CD (10, 13). Both DWI and DCE-MRI may provide quantitative information that can potentially serve as imaging biomarkers for patients with CD. Furthermore, because they can be performed on commercially available scanners, they can potentially be turned into clinical tools in the future if their clinical value is well-defined and validated.
In our study, we wanted to evaluate the use of both techniques in small bowel using endoscopic and histologic findings as a reference standard. Our aims were to determine the relative diagnostic value of apparent diffusion coefficient (ADC) values derived from DWI and perfusion parameters derived from quantitative DCE-MRI analysis in detection of active small bowel inflammation in patients with CD.
This research was conducted with institutional review board (IRB) approval, waiver of informed consent due to the retrospective study design and was in compliance with the Health Insurance Portability and Accountability Act. We searched our data base for patients who were referred to our department for MRE between June 2007 and February 2009, with the clinical indication of known diagnosis of CD involving the small bowel. Inclusion criteria required that the patients had active disease involving the terminal ileum diagnosed by histopathology (endoscopic or surgical) and/or ileocolonoscopy within two months of MRE. The patients did not undergo interval treatment for CD between MRE and endoscopy. Endoscopic findings accepted as indicative of active disease were erosions, ulceration, granularity, or friability. Histopathologic findings accepted as indicative of active disease were the presence of crypt abscesses, mucosal ulceration, neutrophilic infiltration and edema.
MRI examinations were performed with a 1.5 Tesla (T) GE Signa MR scanner (GE Healthcare, Milwaukee, WI). Patients fasted for 6 h before the MRI examination. A total of 1350 mL of Volumen (E-Z-EM Inc) was administered orally over the course of 45 min before scanning. Immediately before the examination, 1 mg of intramuscular glucagon (Glucagen, Bedford Laboratories, Bedford, OH) was administered after the patient was placed in the scanner.
After acquiring a standard three-plane scout image, the following sequences were obtained through the abdomen and pelvis using a four-channel phased array body coil: (i) Axial and coronal FIESTA (fast imaging using steady state acquisition) with and without fat suppression (TR/TE = 3.4/1.4 ms, matrix size = 224 × 224, flip angle = 45°, slice thickness/gap = 7 mm/0 mm). (ii) Axial and coronal T2-weighted single-shot fast spin echo (SSFSE) with and without fat suppression (repetition time/echo time [TR/TE] = infinite/90 ms, matrix size = 256 × 256, slice thickness/gap = 6 mm/0 mm). (iii) Precontrast T1-weighted LAVA (three-dimensional [3D] spoiled gradient echo pulse sequence with fat suppression). Following contrast administration, dynamic postcontrast images were obtained. Gadodiamide (Omniscan; GE Healthcare, Princeton, NJ) was administered intravenously at a dose of 0.1 mmol/kg, followed by a 20-mL saline flush at the rate of 2.0 mL/s. For the DCE-MRI examinations, T1-weighted, 3D, gradient-echo, and free-breathing coronal DCE-MR images covering the entire abdomen were acquired (TR = 3.5–3.9 ms, TE = 1.6–1.9 ms, matrix size = 160 × 256, flip angle = 10°, interpolated slice thickness = 3 mm) with temporal resolution of 5 to 12 s for approximately 3 to 7 min. The dynamic scans were started immediately with the injection of contrast without delay. Postcontrast high resolution T1-weighted images were obtained after completion of the DCE-MRI sequence acquisition with parameters similar to precontrast LAVA images. (iv) Axial and/or coronal diffusion-weighted images (b values = 0 and 600 s/mm2, TR/TE = 8000/75 ms, matrix size = 128 × 128–224, slice thickness/gap = 7 mm/0 mm, number of signals acquired = 4). For each sequence, the upper abdomen and pelvis were scanned separately, and acquisition time for each sequence ranged from 5–8 min. In all sequences, field of view ranged between 32 and 40 cm and ASSET factor of 2 was used. Total scan time was between 40 and 50 min.
A gastrointestinal radiologist with 12 years of experience in body MRI, reviewed the MRE examinations. For each patient, this radiologist evaluated the terminal ileum and noted the presence or absence of mural hyperenhancement (segmental increased intensity of the bowel wall compared with the intensity of normal-appearing ileum), mural thickening (>3 mm), increased T2 signal in the wall, mural stratification (visualization of two or three layers within the bowel wall), adjacent fat stranding (streaky decreased signal within the mesenteric fat on nonfat suppressed T2-weighted images), adjacent enlarged lymph nodes (> 5 mm in shortest diameter), penetrating disease (sinus tract, abscess, phlegmon, or fistula), and the comb sign (prominent vasa recta).
This gastrointestinal radiologist then defined two bowel segments to be used in the quantitative analysis for each patient: a normal-appearing control ileal loop and the most enhancing segment of the terminal ileum. In patients who had previous ileocolectomy and ileocecal anastomosis, the small bowel segment (up to 10 cm) anastomosed to the colon (neo-terminal ileum) was regarded as “terminal ileum.”
DCE-MRI analysis included evaluation of the terminal ileum and normal-appearing ileal loop wall enhancement. A single freehand region-of-interest (ROI) per bowel loop (range, 20–174 mm2) was outlined in the abnormal bowel wall by an MR imaging physicist (with 10 years of MR physics experience) and a radiologist (with 1 year of MR imaging experience) in consensus. The two observers placed the ROI on the part of the terminal ileum and normal ileum (as determined by the gastrointestinal radiologist during conventional MRI analysis) that was most reliably identified through the DCE-MR data set and with least noise. All the ROIs were drawn on DCE-MRI images based on information from high resolution coronal T2-weighted images and delayed postcontrast images to select right slice and location.
A simple two-compartment model (TCM; 14) of the intravascular versus the extravascular extracellular space (EES) was used to describe the distribution of Gadodiamide after bolus injection for the average signal calculated over the ROI. The model predicts a change in contrast concentration as function of time in the tissue, C(t), as follows (15):
where Ktrans (min−1) is the volume transfer constant between intravascular and EES, ve is the volume of EES per unit volume of tissue, and Cp(t) is the arterial input function (AIF). The DCE-MRI data analysis was performed using computer programs written in-house IDL (Research Systems, Inc., Bolder, CO). To fit the differential form of TCM numerically, a golden section search method was implemented to maximize a goodness-of-fit parameter R2. The search section was narrowed down for Ktrans and ve until the difference between the lower and upper boundary was less than 10−3 min−1 or less than 10−3, respectively. C(t) curves were derived from their counterpart signal intensity curves for each ROI by using an approximation formula published previously with muscle as a reference tissue (16). For each patient, the AIF was obtained from the ROI of aorta (17) contrast concentration (Ca(t)) by using Cp(t) = Ca(t)/(1-Hct), with Hct = 0.45 to correct for the hematocrit (Hct; 18). Please note that above contrast concentration conversion method was also applied to the arterial blood and the AIF did not fit any model.
The ADCs were calculated from the wall of the terminal ileum and normal-appearing ileal loop (as determined by the gastrointestinal radiologist during conventional MRI analysis). ADC measurements were performed independently for each segment by two different radiologists (12 and 1 year of experience in body MRI respectively) on a workstation with commercially available diffusion analysis software (Advantage Windows version 4.2.3, GE Healthcare, Milwaukee, WI) in separate sessions one month after the review of DCE-MR images. To obtain the ADC measurements, the images were magnified and oval ROIs were placed on the largest possible area covering the bowel wall. The measurements were made from the area of brightest signal in the bowel wall on the DWI image. ROI areas varied between 27 and 144 mm2. Because two ADC values were obtained for each ileal segment in each patient, the mean ADC was defined as the ADC for each segment.
All data analysis was performed with PASW Statistics (Chicago, IL) Version 17.0. Presence or absence of each conventional MRI finding in terminal ileum of patients with active CD was recorded and summarized.
Average values and standard deviations were calculated for quantitative MRE parameters (DCE-MRI parameters (Ktrans, ve) and ADC values) in the terminal ileum and a normal ileal loop of patients with active CD. Inter-rater agreement for ADC measurements made by two different observers was calculated with a Pearson correlation coefficient. Statistical significance was interpreted as a correlation with P < 0.05. Quantitative MRE parameters were compared between the terminal ileum and normal ileal loop of active CD patients by using the two-sample t-test. Statistical significance was assessed at P < 0.05. ROC analysis was performed to compare the diagnostic performance of the parameters. The area under ROC curve (Az) was used to choose a threshold value for each test parameter by maximizing the combination of sensitivity and specificity.
Ninety-three patients underwent MRE during the specified time period. Eighteen of these patients (11 female and 7 male; mean age, 33.2 years; age range, 20–53 years) who underwent colonoscopy within 2 months of MRE (mean: 20.6 ± 22 days; median, 14 days; range, 0 to 62 days) and were found to have active inflammation in their terminal ileum, were included in the study. In 17/18 patients (94.4%), the diagnosis of active inflammation in the terminal ileum was confirmed histopathologically (surgery: 10 patients; endoscopic biopsy: 7 patients) and in one patient (5.6%) the diagnosis was confirmed by endoscopy only. Specific information for each patient in the study is shown in Table 1.
|Patient||Age (yr)||Sex||Time between MRE and surgery/endoscopy (days)||Histopathological diagnosis|
|Endoscopic biopsy||Surgical biopsy|
Conventional MR findings of actively inflamed terminal ileum are summarized in Table 2. Mural thickening, increased enhancement and increased T2 signal were the most common findings (77.8–100%; Figs. 1–3). Fat stranding, mural stratification, fibrofatty proliferation, enlarged lymph nodes, comb sign, and abscess were present in a relatively small percentage of patients (27.8–50%).
|Parameter||No. of patients|
|Wall thickness||18/18 (100%)|
|Increased enhancement||18/18 (100%)|
|T2 signal||14/18 (77.8%)|
|Fat stranding||9/18 (50%)|
|Fibrofatty proliferation||8/18 (44.4%)|
|Lymph node||8/18 (44.4%)|
|Comb's sign||7/18 (38.9%)|
|Mural stratification||7/18 (38.9%)|
For DCE-MRI, Ktrans and ve values were significantly different between the actively inflamed terminal ileum and normal ileal loop of patients ([0.92 min−1 ± 0.43 versus 0.36 min−1 ± 0.19; P < 0.001] and [0.31 ± 0.13 versus 0.15 ± 0.08; P < 0.001]; Figs. 1–3). Ktrans and ve values were significantly higher in the terminal ileum.
For DWI, ADC value (0.00198 mm2/s ± 0.00055 versus 0.00311 mm2/s ± 0.00056; P < 0.001) was also significantly different between the actively inflamed terminal ileum and normal ileal loop of patients (Table 3). ADC value was significantly lower in the terminal ileum. Pearson's product-moment correlation coefficient comparing the two gastrointestinal radiologists' ADC measurements was 0.930 (P < 0.001), indicating excellent inter-observer agreement.
|Normal ileal loop||Terminal ileum||t-test analysis P value|
|Active CD||Ktrans (min-1)||0.36 ± 0.19||0.92 ± 0.43||<0.0001|
|ve||0.15 ± 0.08||0.31 ± 0.13||<0.0001|
|ADC (×10-3 mm2/s)||3.11 ± 0.56||1.98 ± 0.55||<0.0001|
|Inactive CD||Ktrans (min-1)||0.18 ± 0.13||0.31 ± 0.11||0.14|
|ve||0.12 ± 0.06||0.17 ± 0.10||0.31|
|ADC (×10-3 mm2/s)||2.48 ± 0.63||2.44 ± 0.52||0.88|
Results of the ROC analysis for parameters Ktrans, ve, ADC, and ADC + Ktrans are shown in Table 4 for distinguishing TI and normal ileum. The Az values obtained through ROC analysis for these parameters ranged between 0.88 and 0.92. By using a threshold Ktrans value of 0.52 min−1, it was possible to differentiate actively inflamed terminal ileum from a normal ileal loop with a sensitivity of 83% and specificity of 89%. The threshold ADC value, 2.4 × 10−3 mm2/sec, provided differentiation of actively inflamed terminal ileum from normal ileal loop with a sensitivity of 94% and specificity of 88%. By combining Ktrans and ADC comparisons against their cutoff values to create a combined ordinal parameter, the AUC value for differentiation of actively inflamed terminal ileum from normal ileum was found to be 0.95.
|Ktrans (TI vs. IL)||0.92 ± 0.05||0.52 min−1||0.83||0.89|
|ve (TI vs. IL)||0.88 ± 0.06||0.21||0.78||0.83|
|ADC (TI vs. IL)||0.92 ± 0.05||2.4 x 10−3 mm2/s||0.94||0.88|
|ADC + Ktrans (TI vs. IL)||0.95 ± 0.04||N/A||0.82||1.00|
Our study showed that actively inflamed small bowel segments in patients with CD can be differentiated from normal small bowel loops based on DWI and quantitative DCE-MRI parameters. Terminal ileum wall with active inflammation has restricted diffusion (indicated by low ADC values) and increased perfusion (indicated by increased Ktrans and ve values) compared with normal bowel wall. Analysis based on ADC values alone provided better sensitivity for detection of active bowel inflammation compared with analysis based on DCE-MRI parameters alone. The combination of ADC and Ktrans parameters improved the specificity of the analysis.
Increased bowel wall enhancement is an important and well-known finding indicative of active inflammation in patients with CD (4, 6). Absolute and relative wall attenuation of the terminal ileum on contrast-enhanced CT and visual assessment of bowel wall enhancement on MRI have been shown to correlate with disease activity in patients with CD (3, 4, 19–23). Several studies using MRI have attempted to quantify the degree of bowel wall enhancement by calculating an enhancement ratio (ER; 24–26). Florie et al showed that ERdyn (SIend/SIbaseline) correlated modestly with the CD Activity Index (CDAI; r = 0.38, P = 0.016) and static ER correlated with all clinical reference indices of disease activity (26).
Quantitative analysis of DCE-MRI was recently investigated for detection of bowel wall inflammation (10, 13). It is possible to directly compare quantitative parameters when acquired serially in a given patient, in different patients imaged at the same time or different scanning sites (27). Tofts et al proposed the terms Ktrans and ve as outcome parameters derived from a two-compartment general kinetic model which is the most commonly accepted model (15). In this model, Ktrans refers to the transfer constant and is proportional to the capillary permeability and blood flow and ve refers to the volume of extravascular extracellular space per unit of tissue volume and is proportional to the leakage space. In a previous study performed in 11 patients and 51 bowel segments (colon and terminal ileum), inflamed bowel segments had higher Ktrans and ve values (10). In another study, Taylor et al showed that inflamed bowel had significantly higher mean ER and Ktrans but there was no such difference in ve values (13). In this study, there was no relationship between DCE-MRI derived parameters and histologic markers of inflammation or any clinical parameters (13). These results are inconsistent with the previous reports suggestive of correlation between enhancement level and activity. There is also a striking magnitude difference in the Ktrans values of both normal and inflamed bowel, between the two studies (Oto et al, Ktrans: 0.8 min−1 versus Taylor et al Ktrans: 0.05 min−1). Ktrans values found in Taylor's study are very low compared with previously reported Ktrans values from various normal tissues in the body (including muscle tissue). The results described herein are similar to those described by Oto et al (10), which included mostly colon segments, and our mean Ktrans values for small bowel with and without active inflammation were 0.92 min−1 and 0.36 min−1, respectively. We believe that part of the explanation for this discrepancy may lie in the method used for image acquisition and analysis in this study (28). In Taylor's study, during the acquisition of DCE-MRI data, the patients suspended their breathing for 30 s and then shifted to free breathing. This abrupt shift from breath-hold to breathing usually causes a few vigorous breaths, which could have caused significant motion artifacts in the acquired data. Another plausible reason for the discrepancy in the Ktrans values between this study and Taylor's study lies in the population arterial input function (AIF) used in Taylor's study—instead of a customized AIF—which may have lead to incorrect scaling and, ultimately, lower Ktrans values. The level of distension of bowel segments may also have an impact on the perfusion measurements. In a large CT enterography series, attenuation of the collapsed small bowel loops was consistently higher than the well distended loops (20). We did not specifically investigate the effects of distention in our study. However, we expect a similar effect that was observed on CT. The effect of distention on the calculation of perfusion measurements need to be investigated in future studies.
Increased blood perfusion and permeability are hallmarks of active inflammation, and, as demonstrated in our study, their increase is expected to cause increased Ktrans and ve values. In addition, angiogenesis appears to be an integral component of the pathogenesis of CD and has been shown to correlate with increased Ktrans and ve values in various tumor models (29–31). Increased vascularization is found only in the mucosa and submucosa of patients with CD where tissues are involved by active inflammation (32). Compared with normal mucosa, microvessel density and the levels of vascular endothelial growth factor, interleukin-8 and other angiogenic factors are increased in mucosal extracts of patients with inflammatory bowel disease (29). Brahme and Lindstrom performed in vitro angiography of the resected bowel specimen of patients with CD and found increased vascularity and edema correlating with the severity of inflammation (32).
The use of DWI for detection of bowel wall inflammation in CD has been recently addressed in two small series (11, 12). There were some limitations in these prior studies which limit the applicability of their results to small bowel. In one study, terminal ileum was compared with other colonic segments (11), and in another study, the predominant reference standard was barium enema (12). Both of these studies showed that inflamed bowel wall had decreased ADC values indicating restricted diffusion. Mean ADC values for inflamed (1.59 × 10−3 mm2/s versus 1.57 × 10−3 mm2/s) and normal (2.74 × 10−3 mm2/s versus 2.38 × 10−3 mm2/s) bowel wall were similar in both studies (11, 12). In our current study, ADC values measured from actively inflamed small bowel wall were significantly lower compared with normal bowel wall (actively inflamed bowel: 1.98 × 10−3 mm2/s versus bowel without active inflammation: 3.11 × 10−3 mm2/s). Slightly higher ADC values in our series can be explained by inclusion of only small bowel segments which have higher ADC values compared with colon segments (12).
The exact mechanism for restricted diffusion in actively inflamed bowel wall warrants further study. Restricted diffusion has been reported in various inflammatory processes, predominantly in the brain, but also in the abdomen, in the setting of pyelonephritis and hepatitis (33, 34). In active CD, the lamina propria and submucosa of small bowel are infiltrated by inflammatory cells, and lymphoid aggregates are the characteristic histologic findings (35). Increased cell density and viscosity, dilated lymphatic channels and granuloma development can narrow the extracellular space and contribute to restricted diffusion of water molecules in the inflamed bowel wall.
A variety of research and clinical scoring tools (such as the Crohn's disease Activity Index, biologic indices, endoscopic and imaging studies) have all been used to monitor the disease activity and evaluate the response to treatment, but there remains no established gold standard which accurately provides all this information. The main advantage of using DWI and DCE-MRI in addition to conventional sequences of MRE is the ability to provide quantitative, spatially encoded information about the walls of entire bowel segments. Objective measures of CD inflammatory activity are required to justify the use and evaluate the effectiveness of recently introduced therapies, as well as to improve inter-observer agreement and interpretive accuracy (36). Quantitative diffusion and perfusion parameters have been evaluated as surrogate imaging biomarkers for the evaluation of oncologic drugs and have the potential to serve in a similar function for the follow-up of patients with CD (27). DCE-MRI has already been found to be useful in help determining disease activity in perianal CD (37).
The different temporal resolutions could affect the AIF measurements and thereafter impact the Ktrans and ve measurements. The AIF obtained in this study is similar to previous results of Wang et al (38). Due to 5–12 s temporal resolution, the amplitude of the AIF measured directly from an aorta is much lower than the true AIF. This could cause some errors in the estimation of Ktrans and ve. However, despite the systematic errors, there were clear differences between actively inflamed TI and normal ileum.
Our study had several limitations due to its retrospective design and small sample size. Although we magnified images and used oval regions of interest to try to exclusively cover the bowel wall, we cannot completely exclude the possibility of a partial volume effect on ADC measurements, especially from normal bowel walls. However, very good inter-observer agreement between ADC value measurements is suggestive of relatively low contamination from a partial volume effect. In addition, for DCE-MRI covering the entire abdomen under free breathing, motion artifacts (respiration, peristalsis, pulsation) and compromised signal-to-noise ratio are inevitable, especially with relatively low temporal resolution. Distention of bowel loops with oral contrast and use of an antiperistaltic agent before the study helped decrease the motion related artifacts and improve the delineation of the bowel wall. Therefore, we could perform DCE-MRI analysis for the terminal ileum and normal ileal wall in all of the patients in the study. However, the errors in estimating Ktrans and ve due to motion were not investigated in this study. Finally, the assessment of activity of Crohn's disease in this study was limited to the endoscopic and histologic findings and the review of our experienced gastroenterologist. This was due to the retrospective nature of the study and the absence of a validated retrospective activity index involving these parameters. Future work will include standardized measurements of disease activity. In addition, the reproducibility of quantitative measurements was not assessed in this study, which should be incorporated into clinical research protocols to measure the effects on DCE-MRI techniques (39).
In conclusion, actively inflamed small bowel segments in CD patients demonstrate increased perfusion and restricted diffusion. DCE-MRI and DWI are both promising techniques for the detection of active small bowel inflammation and provide quantitative measures of bowel perfusion and diffusion that can differentiate actively inflamed small bowel segments from normal small bowel in CD. ADC parameter alone can provide better sensitivity for detection of active inflammation compared with perfusion parameters but the combination of DWI and DCE-MRI parameters can potentially improve specificity. Prospective studies using this analysis and its impact on treatment strategies are warranted.