MR Measures of Small Bowel Wall T2 Are Associated With Increased Permeability

Increased small bowel permeability leads to bacterial translocation, associated with significant morbidity and mortality. Biomarkers are needed to evaluate these changes in vivo, stratify an individual's risk, and evaluate the efficacy of interventions. MRI is an established biomarker of small bowel inflammation.

transport and transcellular permeability. Increased gut permeability is implicated in the pathophysiology of a number of gastrointestinal diseases characterized by gut wall inflammation such as celiac disease and inflammatory bowel disease, but also in disorders without overt gut inflammation such as liver cirrhosis, irritable bowel syndrome, obesity, diabetes, HIV, and those who go on to develop or have inflammatory bowel disease in remission. 2,3 Passage of nutrients through the gut wall is a normal physiological process, but an increase in gut wall permeability can result in bacterial translocation (BT), 4 which is defined as passage of bacteria and/or bacterial products across the apparently intact gut wall 5 either via the mesenteric lymph nodes or directly through the portal circulation. There is evidence that the small bowel is the principle focus of pathological BT. 6,7 BT is common in cirrhosis 8,9 and leads to a systemic inflammatory response that exacerbates the hyperdynamic circulation, resulting in increased portal pressure. 4,10,11 BT is also implicated in complications of cirrhosis including variceal bleeding, spontaneous bacterial peritonitis, and hepatic encephalopathy. 10 In Crohn's disease, increased bowel permeability has been reported before macro-and microscopic manifestation of the disease 12,13 and is reversible with biological therapy. 14 It has been suggested that ankylosing spondylitis and multiple sclerosis can be triggered by BT. 15,16 Increased gut permeability has also been associated with hyperglycemia, which in mouse models has been shown to drive intestinal barrier permeability by altering tight and adherence junction integrity. 3 The concept that increased permeability is directly associated with leakage through the gap-junctions of the mucosal barrier has been demonstrated using direct visualization of bowel wall function in vivo using confocal endomicroscopy and a peripherally injected contrast agent. 17 However, visualization is limited to small regions of the bowel, and the cost and sedation required limit this technology to highly selected patients in specialist centers. Alternatively, measurement of intestinal permeability involves monitoring differential urinary excretion of sugars or sugar alcohols that are absorbed in the bowel and poorly metabolized (eg, lactulose, mannitol, rhamnose, sucralose). 18 The lactulose to mannitol excretion ratio (LMR) is the most widely used and validated marker, 3,19 as evidenced by its inclusion as a treatment efficacy endpoint in clinical trials. 20,21 The LMR test has the benefit of using the ratio of two molecules rather than the measured amount of a single molecule, which is thought to correct for interindividual differences in processing of the molecules (eg, bowel transit, renal function, and tissue distribution) that are unrelated to permeability. 18 Administration of oral indomethacin is a well-validated, safe provocation that increases small bowel permeability, 22 with a 2-week washout period being demonstrated to be adequate in preventing cross-contamination. 19 Animal models have also demonstrated that indomethacin causes an acute stimulation of gut motility. 23,24 There is a pressing need for standardized, widely available, noninvasive markers of gut wall changes related to increased permeability and BT. Such measures would allow study of the underlying mechanisms of altered gut permeability and the effects of interventions and could improve management of therapies in key patient groups in a personalized-medicine approach. Various aspects of bowel structure and function can be measured with magnetic resonance imaging (MRI).
Several publications have identified subjective T 2 -weighted measures as being important in the MR assessment of the bowel wall, particularly in relation to Crohn's disease. 25 Aside from gut wall enhancement with contrast, a meta-analysis of MR enterography showed the parameters with consistently highest sensitivity and specificity for bowel wall inflammation were wall thickness and motility. 25 The terminal ileal motility score showed good agreement with endoscopic and histopathologic activity in Crohn's disease, suggesting that it is sensitive to gut inflammation. 26 We hypothesized that quantitative MRI measures of small bowel wall thickness, T 2 , and motility would relate to increased small bowel permeability in healthy volunteers exposed to an indomethacin challenge. 22 Our aim was to undertake a single-institution validation of these quantitative MR small bowel measures as a test of intestinal permeability compared to LMR as a reference standard in an indomethacin-challenged healthy volunteer model of increased intestinal permeability. Due to the semiautomated analysis of the quantitative MR measures, we hypothesized that there would be an excellent intraclass correlation with minimal intrasubject variability.

Materials and Methods
The study protocol was approved by the University of Nottingham

Design
This was a single-center study with two healthy volunteer cohorts. The provocation cohort underwent a double-blind, two-way crossover provocation study administering two doses of 75 mg slowrelease indomethacin or placebo. Participants returned after a minimum 2-week washout period for a repeat study day on the alternative allocation. The order of indomethacin or placebo administration was randomized and blinded, with both pills manufactured to appear identical. The second cohort underwent exactly the same study visits, again a minimum of 2 weeks apart, but without the indomethacin/placebo administration. The study ran from April 2016 until December 2016. The second cohort was performed without the use of indomethacin to test the interobserver agreement and intrasubject test-retest variability of T 2 measurements.

Participants
Participants in the provocation cohort were screened for eligibility and consented prior to randomization. To be eligible, participants had to have no exclusion factors known to increase small bowel permeability. Exclusion criteria included: pregnancy, chronic gastrointestinal disorders or symptoms, diabetes mellitus (type 1 or 2), smoking, psychiatric disease, celiac disease, food allergy, history of atopy, allergy or intolerance to nonsteroidal antiinflammatory drugs (NSAIDs), first-degree relative with inflammatory bowel disease, celiac disease or type 1 diabetes mellitus, alcohol dependency, estimated glomerular filtration rate <45 mL/min, or any contraindications to MRI. For 2 weeks prior to a study visit, volunteers were instructed not to take any regular medications other than oral contraceptives. Participants were informed not to smoke, drink alcohol, or ingest any artificial sweeteners for 72 hours prior to either study visit. In addition, all NSAIDs were prohibited throughout the study. Height and weight were recorded at the screening visit to calculate the body mass index (BMI) of each participant prior to both study visits.

Randomization and Blinding
All participants in the provocation cohort were randomized to receive either indomethacin or placebo administration first. After a minimum 2-week washout period they returned for a second study visit for repeat measures on the alternative treatment arm to act as their own controls. The indomethacin and placebo tablets were manufactured to appear identical. All analyses were performed blind to the treatment allocation and other biomarker results.

Interventions and Procedures
PROVOCATION COHORT. The order of the procedures for each study day (2 study days in total for each participant) is shown in Supplementary Figs. S1 and S2. Participants fasted on the day of the study. They took time-stamped digital photographs of themselves consuming the treatment tablet (placebo or indomethacin) 16 and 4 hours before the planned midway point of a 2-hour lactulose/mannitol urinary excretion test (usually at 10:30 am). Upon arrival at the test center, a cannula was inserted, and blood samples were taken prior to the test. Subjects emptied their bladders and within 5 minutes ingested 5 g of lactulose and 2 g of mannitol dissolved in 100 mL of water, in a 1-minute time window in the presence of an investigator. Thirty minutes after sugar administration, 500 mL of water was given to aid in the collection of urine. Water was allowed ad libitum thereafter. The LMR in the urine collected in the first 2 hours after ingestion was used to quantify the small bowel permeability.
Prior to the MRI scan, participants were given an oral contrast solution (consisting 1 L of water, 25 g/2.5% mannitol, and 2.0 g/0.2% locust bean gum). Forty-five minutes prior to the start of the MRI scan 0.5 L of the solution was given. The remaining 0.5 L was ingested equally over the 15 minutes prior to the start of the MRI data acquisition to obtain optimal distension of the small bowel and terminal ileum.
MRI ACQUISITION. All images were acquired using a whole-body Philips (Best, Netherlands) 3T Achieva (N = 46) with a 16-channel XL Torso coil or Philips 3T Ingenia (N = 2) with a 32 channel dStream Torso coil (Philips Healthcare). Participants lay in the prone position with their arms by their head scanned feet first. After acquisition of the anatomical scans to locate the regions of interest (ROIs), small bowel motility scans were acquired. Subjects were then given two doses of 20 mg intravenous Buscopan (hyoscine N-butylbromide) separated by a minimum of 10 minutes followed by the T 2 and bowel wall thickness scans.
To provide images to measure bowel wall thickness, a 2D balanced turbo field echo (bTFE) sequence was acquired covering the entire small bowel in two 16-second breath-holds. Twenty coronal slices were acquired at resolution 1.2 × 1. For the T 2 measurement, a single-slice, spin echo-prepared, 2D bTFE (TE/TR = 1.68/3.4 msec; flip angle = 50 , half Fourier acquisition (0.625)) was acquired at echo times of 20, 50, 80, 120, 180, and 300 msec. 35 Each spin echo was acquired in a separate breath-hold with a minimum wait time of 15 seconds between scans to ensure full recovery of the magnetization before the next acquisition. Therefore, the number of echo times was limited by the time that Buscopan remains effective (7 minutes 36 ). The images were acquired at 1.3 × 1.5 mm 2 in-plane resolution and reconstructed to 1 × 1 mm 2 over an FOV of 340 × 350 mm 2 . A 5-mm thick coronal imaging slice was placed in the plane where the terminal ilium enters the cecum. This limited the amount of small bowel in the imaging plane but ensured consistency across the 2 study days.
We used a slightly altered version of a previously published protocol for measuring small bowel wall motility. 27,28 2D bTFE images were acquired free breathing at a rate of 1 acquisition per second for 60 seconds. A flip angle of 50 was used with a TE/TR of 1.16/2.32 msec, half Fourier acquisition (0.7), and a SENSE factor of 1.5. Images were acquired at 1.5 × 1.5 mm 2 in-plane resolution reconstructed to 1 × 1 mm 2 . The number of slices was set to cover the entire small bowel wall and ranged from 7−10, depending on the participant.

MRI ANALYSIS: SMALL BOWEL WALL THICKNESS.
Software developed in-house with IDL (Research Systems, Boulder, CO) was used to calculate the mean small bowel wall thickness on the higher-resolution bTFE images. ROIs were manually selected as freeform shapes in the right lower quadrant around the terminal ileum at sites where loops of small bowel lay adjacent to each other (red line in Fig. 1). R.S. (3 years of experience) performed the manual ROI segmentation and saved electronic copies of all the drawn profiles, which were then reviewed by H.W. (2 years of experience). Both readers were supervised and ROIs checked by C.H. (>10 years of experience). The right lower quadrant of the abdomen was chosen to ensure that similar regions of the wall were being sampled at both visits. The program then automatically measured the small bowel wall thickness by generating a profile of the intensity values perpendicular to the wall at adjacent points along the ROI. As profiles were drawn across adjacent loops of bowel, the thickness measured was twice the wall thickness. A minimum of 200 profiles were used to calculate the mean small bowel wall thickness for each individual at each study visit.

MRI Analysis: Small Bowel T 2
The single-slice spin-echo prepared bTFE images were used to measure T 2 of the small bowel wall. We developed in-house software using MatLab (MathWorks, Natick, MA) to identify the bowel wall from T 2 -weighted images. A semiautomated analysis pipeline was set up to isolate the bowel wall and extract the signal from it (UK patent application 2002582.1).
In brief, the images were first registered to the first echo time image via nonlinear registration using MatLab's image registration function imregister, an intensity-based image registration process. A second motion correction step was applied using MatLab's function for estimating displacement fields aimed at correcting local image distortions. The motion correction was run as a bulk process for all datasets, taking 3-4 minutes for each dataset. Three freeform ROIs from different locations were then manually drawn in the content of the bowel to measure the signal intensity of the content at each echo time, which was used in subsequent thresholding and partial volume correction. This was performed by H.W. (with 2 years of experience) under the supervision of P.G. (with >10 years of experience) with saved copies of the electronic masks reviewed by C.H. (with >10 years of experience).
Following this, a series of automated steps isolated the bowel wall. A single mask of the area containing the bowel was created using thresholding to remove subcutaneous fat, muscle, and visceral fat (determined by histogram analysis of the different tissues). Only the areas inside this mask were used for further analysis (Fig. 2b). Images were normalized to allow consistent threshold values to be used throughout the analysis. Next, a binary mask of the bowel wall was created using edge detection and thresholding for each echo time. These masks were then combined to produce a mask that only contained voxels that were identified as wall at every echo time (Fig. 2c).
Following this, a manual quality control step was used to ensure that only small bowel wall was included in the final mask (performed by H.W., under supervision by P.G. and C.H.). The removal process was done by visually inspecting the mask overlaid on all six images and then drawing around areas that did not cover the bowel wall, including the wall of the colon, stomach, uterus, and bladder (Fig. 2d). These areas were removed from the mask.
The final mask was automatically split into smaller sub-ROIs to allow for the heterogeneity along the bowel wall to be investigated. The signal for each sub-ROI at each echo time was extracted. Datasets that contained three or less sub-ROIs were excluded from analysis, as these either had a lack of bowel in the imaging plane or through-plane motion that could not be corrected for in the image registration step. After the initial batched image registration, the analysis for each dataset took 1-2 minutes, including the two manual steps (ie, drawing three ROIs in the content and the manual removal of misidentified sections of bowel).
The T 2 fit took the full bTFE readout into account. 35 To overcome partial volume effects, the signal from each ROI was fit to a two-compartment model (small bowel and content). The T 2 of the content was taken as the average T 2 measured from three ROIs located within the lumen of the bowel. Any ROI that produced an R-squared value of less than 0.9 for the T 2 fit was removed from further analysis. The median and interquartile range of T 2 across all sub-ROIs were calculated and used as the data values for subsequent statistical analysis.

MRI Analysis: Small Bowel Motility
Analysis of the MR measurement of the small bowel motility has been described previously in detail. 27,28 In brief, initially robust data decomposition registration (RRDR) was used to remove the effects of respiratory motion. The global motility index was determined using GI-Quant (Motilent, London, UK) applied to an ROI (performed by R.S. with 3 years of experience, saved maps reviewed by C.H. with >10 years of experience) encompassing the entire visible small bowel region across all slices acquired according to published protocols. 27,28 This index was generated from nonlinear registration parameters generated over the entire image dataset. 27

Analysis of LMR Data for Small Bowel Permeability
The in vivo permeability test is a standard differential urinary sugar excretion test using hydrophilic interactions liquid chromatography (HILIC) with electrospray ionization tandem mass spectrometry (ESI-MS/MS). 29,30 After collection, the total urine volume was noted and 1.5 mL sample aliquots were filtered with 450 nm filters (Merck Millipore, Billerica, MA) and stored at -20 C until batch analysis was performed. All the samples were coded without reference to the test condition. The measurements were performed by a lab technician (C.O.) blinded to the test condition.
To precipitate any excess salt, 20-μL aliquots were diluted with 980 μL 90% acetonitrile to which internal standards xylitol and raffinose were premixed at 0.5 μg/mL final concentration. These were vortexed, incubated at -20 C overnight, and centrifuged, and the supernatant was decanted into amber high-performance liquid chromatography (HPLC) vials. Calibration standards were made as a dilution series from 2.5-500 μg/mL of mannitol and lactulose from stocks made in water. The method was validated by creating six independently prepared dilutions of 5, 50, and 500 μg/mL. To accurately identify lactulose, sucrose standards were also prepared.
For convenience, two liquid chromatography columns, a Sequant ZIC-pHILIC (5 μm) 100 × 2.1 mm and a ZIC-HILIC (5 μm) 150 × 2.1 mm from Merck KgaA (Darmstadt, Germany), were used in series and kept at 15 C. The mobile phase was acetonitrile and 5 mM ammonium acetate adjusted dropwise to pH 6.85 with 0.05% ammonium hydroxide solution. The flow rate was 0.3 mL/min. The detector was a Sciex 4000 QTrap (Framingham MA) operating in -ve ion electrospray mode with the source at 350 C with curtain, nebulizer, and auxiliary gases were set to 10, 40, and 20, respectively. The ion-spray voltage was -4200 V. As the two analytes had very different ranges of concentrations, samples were quantified against the appropriate region of the line. A minimum of 5 points were used for each analyte.

Interobserver and Intrasubject Reproducibility
The second cohort of 10 healthy volunteers was scanned twice a minimum of 2 weeks apart to look at the intrasubject reproducibility of T 2 in the small bowel wall in healthy volunteers. The only difference to the provocation protocol cohort was that no indomethacin or placebo was given. Furthermore, the T 2 measurement slice was not constrained to be over the terminal ileum but was chosen to maximize the amount of small bowel imaged. The analysis was performed by two observers to allow interobserver agreement to be tested (performed by H.W. with 2 years of experience and A.A. with 2 years of experience, supervised by C.H. and P.G., both with >10 years of experience).

Statistical Analysis
A previous study measured mean, healthy, small bowel thickness to be 1.5 mm with a standard deviation (SD) of 0.5 mm. 31 Assuming a 66% increase in bowel thickness as a result of indomethacin provocation 22 would be comparable to active Crohn's disease, 25,32,33 we anticipated that 24 participants in the provocation cohort would give us more than 90% power to reject the null hypothesis with alpha of 0.05 and between group correlation of 0.5.
The data were tested for normality using the Shapiro-Wilk's test. Paired testing was then carried out to determine whether paired differences between measures with placebo and indomethacin provocation were significant. Data found to be normally distributed were tested using a paired t-test; otherwise, data were compared using a Wilcoxon signed rank test. The relationship between T 2 and LMR was investigated using the Pearson's correlation coefficient. In this exploratory study a covariate of interest was to describe the variability of T 2 data across ROIs between subjects (interquartile range across sub-ROIs for each participant).
Interrater variability between two independent observers for the second cohort of healthy volunteers who had repeat measures on 2 study days a minimum of 2 weeks apart was reported by Bland-Altman. 34 The interobserver variability was assessed by calculating the intraclass correlation with a two-way random model of absolute agreement and interpreted as follows: 0.81-1: almost perfect correlation; 0.61-0.8: good correlation; 0.4-0.6: moderate correlation; 0.21-0.4: fair correlation; 0.0-0.2: poor correlation. 34 The repeatability was defined as poor when the coefficient of variation (CoV) was >30%, acceptable when CoV was between 20% and 30%, good when CoV was between 10-20%, and excellent when CoV ≤10%. 35 Statistical analyses were performed using SPSS v. 22 (IBM, Armonk, NY) or GraphPad Prism v. 8.0 for Windows (GraphPad Software, La Jolla, CA).

Provocation Cohort
Twenty-four healthy volunteers consented to the provocation study ( Supplementary Fig. S3). All participants attended both study days with placebo and indomethacin administration, the order of which was randomly and blindly allocated. Two participants were excluded (one male, one female) from the per-protocol final analyses, as one was noncompliant with the study protocol and one had an incidental finding of an asymptomatic thickened terminal ileum prior to the intervention on review of the MRI data ( Supplementary Fig. S3). This participant was subsequently diagnosed with inflammatory terminal ileal Crohn's disease on colonoscopy, confirmed by histology. No participants suffered any adverse events caused by administration of indomethacin. Fifteen of the volunteers were female (63.6%). The median age was 23 years (interquartile range [IQR] [22][23][24][25], and median BMI was 23.9 (IQR 21.6-28.0) kg/m 2 . The median interval between study visits was 21 (IQR 18-27) days.

MRI Measures Associated With Indomethacin Provocation
SMALL BOWEL WALL THICKNESS. There was no significant measurable difference (P = 0.17) between small bowel wall thickness around the terminal ileum between placebo (1.28 mm, IQR 1.21-1.36 mm) and provocation with indomethacin (1.29 mm, IQR 1.25-1.36 mm).
SMALL BOWEL WALL T 2 . For the T 2 measurements, six datasets were not used because the number of final sub-ROIs was too small (three or fewer) due to significant respiratory or bowel motion that could not be corrected. Figure 3 shows that indomethacin provocation induced a statistically significant increase in small bowel wall T 2 compared to placebo (mean T 2 ± SD: 0.115 ± 0.063 seconds vs. 0.070 ± 0.036 seconds, respectively, P < 0.05. There was also a nonsignificant trend toward increased variation in T 2 along the bowel wall after administration of indomethacin compared to placebo (0.16 vs. 0.10 seconds, P = 0.065, Fig. 3c).

LMR Test Quantification of Bowel Permeability
Indomethacin induced significantly increased LMR from 0.019 (IQR 0.016-0.026) on placebo to 0.025 (IQR 0.021-0.039) on indomethacin provocation (P < 0.05). There was a significant positive correlation (r = 0.68, P < 0.05) between LMR and SB wall T 2 (Fig. 3b). There was also a significant positive correlation (r = 0.63, P < 0.05) between the change in LMR and the change in T 2 induced by the indomethacin challenge for each subject (Fig. 3c).

Interobserver Agreement and Intrasubject Reproducibility
Two subjects' data were excluded from the analysis due to having less than three sub-regions due to significant peristaltic movements during the acquisition. Figure 4 shows the results obtained from the remaining eight healthy volunteers with two study visits per participant, separated by a minimum of 2 weeks (16 study days in total). Interobserver agreement was excellent for T 2 measurements (N = 16, intraclass correlation = 0.89, P < 0.05). Bland-Altman estimated bias was 0.005 seconds for observer A (95% confidence interval [CI] limit of agreement -0.04 to +0.05 seconds) and 0.0006 for observer B (95% CI limit of agreement -0.05 to +0.06 seconds). The coefficient of variation was 19% for observer A and 21% for observer B. The results are summarized in Figure 4.

Discussion
We have shown that small bowel wall T 2 increased following indomethacin provocation and correlated with increased permeability, as demonstrated by a 2-hour lactulose/mannitol urinary excretion ratio (LMR) test. MRI measures of small bowel wall thickness and motility were unchanged by indomethacin provocation. We also showed that the test-retest repeatability of small bowel wall T 2 measurement was acceptable, with the variation in values lower than the difference seen from the indomethacin provocation. In addition, the interobserver reproducibility was excellent.
The prospective double-blind crossover study design minimizes confounding factors and increases the power of the study. All the participants included in the per protocol analysis were well-phenotyped, healthy volunteers. All analysis was performed blind to treatment allocation and compared to small bowel permeability as defined by 2-hour LMR, the current standard measure of small bowel permeability. 2,29 The MRI measures obtained are quantitative, in contrast to qualitative MRI measures that are commonly used to assess the small bowel, 25 which may improve the power of studies involving repeated measurement within and between subjects. These MRI techniques do not require administration of intravenous contrast and are, therefore, safe and appropriate for repeated measurements. 36 The protocols are based on widely available scan sequences and, as such, can be rapidly adopted into clinical and research protocols. In order for quantitative  T 2 to become a viable clinical measure, the analysis must be as fast and automated as possible. The T 2 analysis method developed here can be applied to any images in which the small bowel content and wall have a different signal intensity, not just for T 2 mapping.
While two doses of 75 mg indomethacin is a known positive control and increased small bowel permeability as expected, 19,22 the dose was relatively low and the intervention was only transient. Nonetheless, it was sufficient to induce a change in small bowel wall T 2 . Larger or more frequent doses of oral NSAIDs are known to cause variable patchy small bowel erosions, 37,38 which may explain the heterogeneity and increased intrasubject range of the small bowel wall T 2 measurements calculated here. NSAID enteropathy increases permeability by direct injury to the intestinal mucosa with inflammation and edema, but also disrupts the tight junctions between cells, which permits the passage of ions and water. 2,3 The changes in small bowel wall T 2 could reflect direct inflammation and/or shifts in water through the tight junction defects associated with increased permeability.
Although indomethacin is known to cause an acute stimulation of motility in animal models, a feature thought to be important in the secondary bacterial penetration of the mucosal barrier, 23 the longer-term effect is more dominated by the inhibitory effect of mucosal inflammation, as is seen in humans with Crohn's disease. 26 At the doses used here the mucosal changes would be predicted to be much less severe than is seen in the animal models, where hemorrhage and marked ulceration is common. 24 This may account for the lack of any effect of indomethacin on small bowel motility (increase or decrease) observed in this study. It is intriguing to note that motility is a sensitive marker of inflammation when compared to endoscopy or histology. 26 Indomethacin provocation did not cause a change in motility but did cause a significant change in bowel wall T 2 . Hence, bowel wall T 2 may either be a more sensitive marker of mucosal inflammation than motility or measuring more subtle change within the bowel wall (eg, edema) that correlates with gut permeability.
Bowel gas is mainly located in the large intestine. With the subjects lying prone, gas was pushed away from the small bowel region (as it lies at the posterior edges of the large colon) minimizing its influence on the images. bTFE sequences are prone to artifacts from poor shimming due to field distortions and these would have been eliminated using the thresholding techniques.
The wall thickness measurements may not be sensitive enough to detect the subtle changes caused by this indomethacin intervention. Changes to these measurements are seen in Crohn's disease, where the damage to the bowel wall due to inflammation is much more extensive and prolonged. 25 The interobserver reproducibility of the T 2 measurement was found to be robust. Intrasubject variability was high in some cases, which was probably due to the fact that the imaging slice was placed to cover a large area of bowel rather than being restricted to the plane containing the terminal ilium, as was a requirement in the initial study. Defining the imaging plane based on a fixed anatomical location would be likely to reduce the intrasubject variability. This could be overcome by using multislice imaging. 39 Two out of the 10 subjects were removed due to the presence of peristalsis during the imaging, which prevented the T 2 of the bowel wall from being measured.
Our study suggests noncontrast-enhanced quantitative MR measurement of the small bowel wall T 2 could provide a sensitive biomarker of permeability. This has far-reaching implications if validated in a wider range of patient groups where increased small bowel permeability and bacterial translocation contribute significantly to the pathogenic process and are associated with clinical manifestations or outcomes. This method may have impact in non-GI diseases where increased permeability of the gastrointestinal tract has been considered a putative pathogenic mechanism; for example, in ankylosing spondylitis, diabetes, and multiple sclerosis. Arguably, the lack of robust, accessible, and affordable biomarkers of these potentially pathophysiological changes has hampered research in this area. A widely available, noninvasive, in vivo measure of small bowel structure and integrity would be an important tool for long-term, noninvasive mechanistic studies, and to evaluate the efficacy of specific interventions.

Limitations
The associated T 2 analysis tool requires manual inspection of bowel wall maps and removal of misidentified regions of wall, which is inherently subjective and only partly addressed by averaging several ROIs in each subject. This is an inherent weakness; however, the interobserver reproducibility suggests that this has minimal impact on the measurement of T 2 .
Although LMR is the most validated measure of small bowel permeability, it does have known shortcomings. 3,40 First, up to 30% of participants have detectable urinary mannitol at baseline (prior to administration of test sugars) or disproportionate excretion relative to the mass of mannitol administered for the test. This is hypothesized to be a result of inadvertent ingestion of mannitol in diet or medications. 40 Second, the measurement made at 0-2 hours mostly reflects small bowel permeability but may also reflect colonic permeability.
In this pilot exploratory study, six out of the 22 subjects were removed from analysis due to motion in the T 2 measurement images. This is likely to have statistically underpowered our study based upon the original sample size calculation informed by Crohn's data, but suggests quantitative T 2 is a sensitive imaging biomarker. This motion was largely due to respiratory motion resulting in the imaging slice moving between acquisitions. This is a weakness of single-slice imaging, which could be overcome by using simultaneous multislice methods. 39 A further shortcoming of the test-retest of the T 2 measurements was the small sample size.

Conclusion
We implemented a noncontrast MRI technique to measure T 2 of the bowel wall in vivo and showed that changes in bowel wall T 2 are related to changes in small bowel wall permeability following indomethacin provocation. Sensitive MR measures of bowel structure and function, including quantitative T 2 , could be used to characterize relevant patient populations where increased gut permeability is thought to be a key event in the pathophysiology towards clinical outcomes and measure the effect of interventions.