Magnetization transfer imaging adds information to conventional MRIs to differentiate inflammatory from fibrotic components of small intestinal strictures in Crohn’s disease

Abstract

Objectives:

Identifying inflammation- or fibrosis-predominant strictures in Crohn’s disease (CD) is crucial for treatment strategies. We evaluated the additive value of magnetization transfer (MT) to conventional MRI for differentiating CD strictures using surgical histopathology as reference standard.

Methods:

Twenty-eight consecutive CD patients who underwent MRI preoperatively were recruited. MRI parameters included T2-weighted imaging (T2WI) hyperintensity, bowel wall thickness, enhancement pattern changes over time, enhancement pattern and gain ratio in dynamic contrast-enhanced phases, and MT ratio. Correlation analysis was performed using Spearman’s rank test. Receiver operating characteristics curve analysis and Cohen’s κ were used. A model with combined MRI variables characterizing intestinal strictures was proposed and validated in 14 additional CD patients.

Results:

Significant correlations with histological inflammation scores were shown for wall thickness (r=0.361, P=0.001) and T2WI hyperintensity (r=0.396, P<0.001), whereas histological fibrosis scores were significantly correlated with MT ratio (r=0.681, P<0.001) and wall thickness (r=0.461, P<0.001). T2WI hyperintensity could differentiate mild from moderate-to-severe inflammation with a sensitivity of 0.871 and a specificity of 0.800. MT ratio could discriminate mild from moderate-to-severe fibrosis with a sensitivity and a specificity of 0.913 and 0.923, respectively. Combining MT ratio and T2WI hyperintensity, the MRI classification moderately agreed with the pathological stricture classification (P<0.01, κ=0.549). In the validation set, the diagnostic accuracy of T2WI hyperintensity and MT ratio were 86% and 89%, with good agreement between MRI and histopathological classification (P<0.01, κ=0.665).

Conclusions:

MT ratio combined with conventional MRI improves the differentiation of fibrotic from inflammatory components of small-bowel strictures in CD patients.

Introduction

Crohn’s disease (CD) is a chronic inflammatory bowel disorder in which intestinal strictures followed by obstruction represent a common major complication [1, 2]. Recent evidence demonstrates that most stricture lesions in CD contain a mixture of fibrotic and inflammatory components to varying degrees [3–6]. Because predominantly inflammatory strictures are potentially treatable using medical therapies, but severely fibrotic strictures usually require endoscopic treatment or surgical resection, non-invasive techniques measuring the degree of bowel inflammation or fibrosis can optimize management strategies.

MRI with abundant imaging sequences and without ionizing radiation is widely accepted for assessing CD activity [7–9]. Several studies [3, 6, 10–13] using surgical histopathology as the reference standard have confirmed the value of conventional MRI for assessing CD inflammation. MRI findings, e.g., hyperintensity on T2-weighted imaging (T2WI) and mural thickness, are associated with inflammatory activity [3, 6, 10–13]. Although studies have shown that several MRI parameters including mural thickness [3], enhancement pattern [10, 11], and percentage of enhancement gain [11] are associated with the degree of pathological fibrosis, other studies using contrast enhancement in CD with bowel fibrosis reported conflicting results [10, 11].

Recently, magnetization transfer (MT) MRI has been reported to be a non-invasive modality that quantitatively distinguishes fibrotic from non-fibrotic CD in animal studies [14–16]. MT image contrast is mainly determined by the fraction of large macromolecules or immobilized phospholipid cell membranes in the tissue, and the MT ratio enables the indirect quantification of macromolecule concentrations in aqueous physiological environments [17, 18]. Our previous study showed that MT imaging (MTI) outperforms contrast-enhanced and diffusion-weighted MRI in detecting and distinguishing varying degrees of bowel fibrosis with or without coexisting inflammation [19]. However, MTI is sensitive to fibrotic, but not inflammatory, changes in the bowel [14]. Thus, to address the unmet need of simultaneously assessing inflammation and fibrosis in the same bowel segment, this study aimed to investigate whether the combination of MTI and conventional MRI could better characterize the type of CD strictures, using surgical histopathology as the reference standard.

Methods

Patients

Twenty-eight consecutive adult patients with established CD diagnosis were recruited as the derivation set from July 2014 through March 2017. Inclusion criteria were: (a) preoperative MR enterography within 2 weeks of elective surgery for symptomatic small-bowel stricture (defined as fixed luminal narrowing in an appropriately distended lumen) and (b) the ability to match resected bowel specimens with pathological bowel segments identified on MRI. Exclusion criteria were: (a) inadequate image quality and (b) other concomitant bowel diseases, e.g., intestinal tumours.

From April 2017 through March 2018, 14 additional CD patients were recruited for the validation study using identical criteria. This prospective study was approved by the institutional ethics review board, and written informed consent was obtained from all patients.

Regarding the CD activity index (CDAI), clinical remission was defined as a CDAI <150. Mildly active CD was defined as a CDAI of 150–220, moderate–severe disease activity as a CDAI of 220–450, and severe disease as a CDAI >450 [20].

MRI protocol

Bowel preparation was performed as described previously [21]. Patients ingested polyethylene glycol electrolyte solution for 6–8 h, followed by 1600–2000 mL 2.5% mannitol solution as an oral contrast agent, 1 h before the MRI. A spasmolytic (10 mg raceanisodamine hydrochloride; Minsheng Pharmaceutical Group Co.) was injected intramuscularly 10 min before scanning.

The MR images were acquired using a 3.0T MR system (Magnetom Trio; Siemens Medical Solutions) with multi-channel phased-array body coils. The conventional MRI included axial and coronal T2WI. Dynamic, fat-saturated T1WI was conducted before and 30 s, 70 s, and 7 min after intravenous contrast injection (0.2 mL/kg gadopentetate dimeglumine, Beilu Pharmaceuticals). MTI was performed using two axial gradient-echo sequences with and without off-resonance prepulse (Supplementary Table 1).

MRI interpretation

Conventional MRI

Conventional MR images were analysed using the RadiAnt DICOM Viewer 3.4.1 software (Medixant) by two radiologists with 5 (Z.F.) and 8 (S.H.) years of experience in interpreting bowel MRIs; they were blinded to clinical, laboratory, and pathological information.

Conventional MRI findings associated with pathological bowel segments, including wall thickness, mural signal intensity (SI) on T2WI, enhancement pattern, enhancement progression from 70 s to 7 min, and the percentage of gadolinium enhancement gain were investigated. The following parameters were evaluated: (a) wall thickness, measured three times in the transverse section of the bowel segment on T2WI and averaged; (b) mural SI, defined as normal or hyperintensities on T2WI (Supplementary Fig. 1; [11]); (c) enhancement pattern, observed at 30 s, 70 s, and 7 min after contrast injection and classified as mucosal only, homogeneous, or layered (Supplementary Fig. 2; [3, 10]); (d) enhancement progression in deep layers from 70 s to 7 min, recorded as stability or progression of enhancement over time (Supplementary Fig. 3; [11]); and (e) percentage of gadolinium enhancement gains between 30 s or 70 s and 7 min, calculated for superficial (mucosal) and deep (submucosa/muscularis propria) layers according to the following formula: percentage of enhancement gain at 7 min vs 30 s = ((SI 7 min – SI 30 s) / (SI 30 s)) × 100 or at 7 min vs 70 s = ((SI 7 min – SI 70 s) / (SI 70 s)) × 100 [3, 10, 11]. Postcontrast wall SIs at 30 s, 70 s, and 7 min were measured using three ROIs in matching locations within the pathological segment on T1WI.

MT imaging

MT ratios were calculated as follows: MT ratio (%) = (1 − (M_sat / M₀)) × 100, where M_sat and M₀ refer to the SIs with and without off-resonance prepulse saturation, respectively (Supplementary Fig. 4; [14, 22]). Colour MT ratio maps were generated using in-house MATLAB software (Math Works Inc.). Three irregular ROIs were drawn on MT ratio maps in selected bowel walls by two radiologists with 7 (X.L.) and 8 (S.H.) years of experience in bowel MRI who were blinded to other patient data. These MT ratios were averaged (Fig. 1b).

Fig. 1.

A 41-year-old man with ileal CD. (a) Axial T2-weighted image showing marked bowel wall thickening and hyperintensity in the ileum (arrow). Colour MT ratio map (b) demonstrating the MT effect in the ileum (arrow; MT ratio=39.22%) similar to that in the skeletal muscle (MT ratio=44.59%). H&E (c) and Masson trichrome (d) stainings depicting severe inflammation (score=3) and marked transmural fibrosis (blue area [arrows]; score=3), respectively (0.5× magnification). CD Crohn’s disease, H&E haematoxylin and eosin, MT magnetization transfer

Region-by-region correlation between MRI and surgical specimens

Region-by-region correlations between MRI and resected specimens were performed by a radiologist not blinded to imaging and surgical information (C.S.) who has extensive experience in bowel MRI interpretation. Bowel lesions or anatomic landmarks were matched to bowel walls on preoperative MR images and surgical specimens [19]. The most stenotic areas were selected for histological assessment and MRI correlation. We obtained 1–4 specimens per patient according to the disease extent [11, 23].

Histopathological analysis

The resection margin was intraoperatively determined by the colorectal surgeon based on inspection. Resected bowel segments were stained with haematoxylin and eosin (H&E) and Masson trichrome for histological inflammation and fibrosis scoring, respectively [19]. A pathologist (Q. C.) with 11 years of experience in bowel pathology who was blinded to clinical and MRI information assigned the histological scores 0 (none), 1 (mild), 2 (moderate), or 3 (severe) using a semi-quantitative scoring system modified from previously described methods (Supplementary Table 2; [5]).

Statistical analysis

Statistical analyses were performed using SPSS v23.0 (IBM Corp.) with significance defined by P<0.05. Quantitative data are expressed as the mean ± standard deviation; qualitative data are presented as percentages and/or absolute values. The correlation between inflammation and fibrosis scores was determined using Spearman’s rank test. Interobserver agreement on MRI parameters was performed with Cohen’s kappa (κ) for qualitative parameters (high signal intensity on T2WI, enhancement pattern) and intraclass correlation coefficient (ICC) for quantitative parameters (MT ratio, wall thickness, and percentage of enhancement gain). We characterize κ values of 0–0.20 as poor, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as good, and 0.81–1.00 as excellent. ICC was classified as poor (<0.40), fair (0.40–0.59), good (0.60– 0.74), or excellent (0.75–1.00). The relationships between categorical variables were assessed using the chi-squared test. Group differences were analysed using the Kruskal–Wallis test. Histopathological inflammation and fibrosis scores were correlated with MRI variables using partial correlation analysis. A logistic regression analysis determined MRI variables that indicate the presence of moderate-to-severe fibrosis or inflammation. Subsequently, a ROC curve analysis was performed, and the AUCs were calculated to determine the diagnostic accuracy of the MRI indicators. Cohen’s κ was used to measure the agreement between combined MRI indicators and histological findings for the classification of intestinal strictures.

Results

Demographic and clinical data

Of the 28 patients recruited into the derivation study, one was excluded (inadequate MRI quality). The remaining 27 patients with 82 resected small-bowel specimens of acceptable MRI quality were analysed. Of these specimens, 54 were from the terminal ileum and 28 from the proximal ileum and jejunum. The demographic and clinical characteristics of the patients are summarized in Table 1.

Table 1.

Demographic and clinical characteristics of the patients operated for Crohn’s disease

	Derivation set	Validation set
Sex: female / male, n (%)	17 (63) /10 (37)	9 (64) /5 (36)
Age, mean ± SD, years	33.56 ± 8.67	33.93 ± 10.43
Disease duration, median (IQR), months	38 (24–120)	55 (27–121)
Montreal classification, n (%)
L1	7 (26)	5 (36)
L2	0 (0)	0 (0)
L3	20 (74)	9 (64)
L4	6 (22)	3 (21)
B1	0 (0)	0 (0)
B2	9 (33)	4 (29)
B3	18 (67)	10 (71)
P	12 (44)	8 (57)
Concomitant medication(s)a, n(%)
5-ASA	11 (41)	6 (43)
Corticosteroids	8 (30)	5 (36)
Azathioprine	9 (33)	6 (43)
Methotrexate	2 (7)	0 (0)
Tumour necrosis factor antagonists	5 (19)	1 (7)
Interval between MRE and surgery, median (IQR), days	9.00 (3.00–13.00)	8.50 (4.75–12.25)
Less than 7 days, n (%)	10 (37)	6 (43)
Between 7 and 14 days, n(%)	17 (63)	8 (57)
Surgery type, n (%)
Ileocecal resection	19 (70)	10 (71)
Partial small-bowel resection	8(30)	4 (29)
Number of bowel specimens, n (%)
Terminal ileum	54 (66)	26 (70)
Ileum+ jejunum	28 (34)	11 (30)
CDAI, median (range)	253.10 (148.43–323.25)	321.00 (223.60–467.50)
CDAI<150/CDAI≥150 b, n (%)	6 (22)/ 21 (78)	2 (14)/ 12 (86)
CRP, mean ± SD, mg/L	31.06 ± 39.14	39.19 ± 45.31
ESR, mean ± SD, mm/h	37.78 ± 23.82	40.29 ± 20.62

5-ASA 5-aminosalicylic acid, CDAI Crohn’s disease activity index, CRP C-reactive protein, ESR erythrocyte sedimentation rate, IQR interquartile range, SD standard deviation

^aPatients may receive more than one concomitant medication

^bClinical remission is defined as a CDAI<150. Active Crohn’s disease is defined as a CDAI≥150

Histological bowel wall evaluation and interobserver agreement of MRI parameters

The histological inflammation gradings on H&E stainings were as follows: none (n=0), mild (n=20), moderate (n=44), and severe (n=18) grades. The fibrosis gradings based on Masson trichrome stainings were as follows: none (n=6), mild (n=7), moderate (n=42), and severe (n=27) grades. The positive correlation between histological inflammation and fibrosis scores (r=0.532, P<0.001) highlights the necessity to control for one factor when analysing the other.

The interobserver agreement was fair to excellent. The highest agreement was obtained in the wall thickness evaluation with an ICC of 0.911 (95% confidence interval [CI], 0.865–0.928), whereas the lowest agreement was observed in the enhancement pattern at 7 min with a κ of 0.406 (95% CI, 0.212–0.600; Table 2).

Table 2.

Agreement between observers in the derivation set

MRI parameters	Agreement between observers
MTR	0.899 (95% CI, 0.848–0.934) b
High signal intensity on T2WI	0.819 (95% CI, 0.680–0.958) a
Wall thickness	0.911 (95% CI, 0.865–0.928) b
Enhancement pattern at 30 s	0.818 (95% CI, 0.700–0.936) a
Enhancement pattern at 70 s	0.651 (95% CI, 0.865–0.905) a
Enhancement pattern at 7 min	0.406 (95% CI, 0.212–0.600) a
Enhancement progression	0.756 (95% CI, 0.615–0.897) a
Percentage of mucosa enhancement gain at 7 min vs 30 s	0.818 (95% CI, 0.732–0.879) b
Percentage of submucosa enhancement gain at 7 min vs 30 s	0.434 (95% CI, 0.241–0.595) b
Percentage of mucosa enhancement gain at 7 min vs 70 s	0.629 (95% CI, 0.478–0.744) b
Percentage of submucosa enhancement gain at 7 min vs 70 s	0.737 (95% CI, 0.619–0.822) b
Percentage of mucosa enhancement gain at 30 s vs precontrast	0.868 (95% CI, 0.803–0.913) b
Percentage of mucosa enhancement gain at 70 s vs precontrast	0.813 (95% CI, 0.724–0.875) b
Percentage of mucosa enhancement gain at 7 min vs precontrast	0.803 (95% CI, 0.710–0.868) b

MTR magnetization transfer ratio, CI confidence interval, T2WI T2-weighted imaging

^aCohen’s kappa;

^bIntraclass correlation coefficient

MRI parameters associated with the degree of bowel inflammation

Significant differences were found in MRI variables including MT ratio (P=0.002), wall thickness (P<0.001), T2WI hyperintensity (P<0.001), and enhancement patterns at 30 s, 70 s, and 7 min (P=0.002, 0.004, and 0.015, respectively) among mild, moderate, and severe grades of inflammation (Table 3; pairwise comparison of MT ratio and wall thickness among inflammation groups are shown in Supplementary Figs. 5, 6) Significant correlations with histological inflammation scores were detected for wall thickness (r=0.361, P=0.001) and T2WI hyperintensity (r=0.396, P<0.001; Supplementary Table 3). However, after controlling for the influence of histological fibrosis scores, the inflammation scores showed only weak correlations with MT ratio (r=−0.037, P=0.742) and enhancement patterns at 30 s, 70 s, and 7 min (r=0.159, 0.124, and −0.100; P=0.158, 0.270, and 0.928, respectively).

Table 3.

Association between MRI parameters and histological inflammation score

	Histological inflammation score
MRI parameters	Mild (n=20)	Moderate (n=44)	Severe (n=18)	P
MT ratio, mean ± SD (range)	29.85±7.43 (19.12–40.93)	37.68±6.46 (21.53–50.79)	35.87±7.46 (19.10–48.12)	0.002 b
Wall thickness, mean ± SD (range)	5.35±2.39 (2.50–10.40)	10.81±3.31 (3.31–22.80)	11.11±3.40 (6.73–18.40)	<0.001 b
High signal intensity on T2WI	4 (20.0%)	39 (88.6%)	15 (83.3%)	<0.001 a
Enhancement pattern 30 s				0.002 a
1: homogeneous	18 (90.0%)	18 (40.9%)	9 (50.0%)
2: mucosa only	1 (5.0%)	20 (45.5%)	9 (50.0%)
3: layered	1 (5.0%)	6 (13.6%)	0
Enhancement pattern 70 s				0.004 a
1: homogeneous	17 (85.0%)	16 (36.4%)	9 (50.0%)
2: mucosa only	2 (10.0%)	22 (50.0%)	9 (50.0%)
3: layered	1 (5.0%)	6 (13.6%)	0
Enhancement pattern 7 min				0.015 a
1: homogeneous	18 (90.0%)	22 (50.0%)	14 (77.8%)
2: mucosa only	1 (5.0%)	16 (36.4%)	4 (22.2%)
3: layered	1 (5.0%)	6 (13.6%)	0
Enhancement progression				0.143 a
1: Stable over time	13 (65.0%)	17 (38.6%)	9 (50.0%)
2: Progression over time	7 (35.0%)	27 (61.4%)	9 (50.30%)
Percentage of mucosa enhancement gain at 7 min vs 30 s, mean ± SD (range)	58.64±61.38 (−11.15–169.89)	56.95±52.01 (−3.03–200.57)	65.16±56.78 (5.21–192.03)	0.745 b
Percentage of submucosa enhancement gain at 7 min vs 30 s, mean ± SD (range)	63.71±56.82 (−3.13–169.89)	84.60±60.66 (−10.93–237.47)	84.77±63.89 (17.14–274.10)	0.366 b
Percentage of mucosa enhancement gain at 7 min vs 70 s, mean ± SD (range)	8.99±24.91 (−17.62–86.18)	6.95±16.63 (−16.66–73.11)	5.33±9.48 (−8.42–26.80)	0.957 b
Percentage of submucosa enhancement gain at 7 min vs 70 s, mean ± SD (range)	17.26±35.09 (−17.62–124.50)	29.11±31.93 (−16.66–136.11)	22.02±23.18 (−8.42–73.25)	0.140 b
Percentage of mucosa enhancement gain at 30 s vs precontrast, mean ± SD (range)	85.74±60.05 (−1.92–244.60)	109.99±77.08 (−5.22–294.02)	88.48±70.86 (11.65–229.13)	0.382 b
Percentage of mucosa enhancement gain at 70 s vs precontrast, mean ± SD (range)	153.10±59.41 (26.54–264.63)	188.94±77.16 (33.78–358.21)	175.58±82.71 (57.76–408.60)	0.232 b
Percentage of mucosa enhancement gain at 7 min vs precontrast, mean ± SD (range)	169.10±60.23 (74.88–311.48)	204.07±76.48 (35.24–342.81)	187.94±77.01 (66.99–365.79)	0.264 b

MT magnetization transfer, SD standard deviation, T2WI T2-weighted imaging

^aChi-squared test;

^bKruskal-Wallis test

MRI parameters associated with the degree of bowel fibrosis

MRI parameters significantly associated with the degree of histological fibrosis were MT ratio (P<0.001), wall thickness (P=0.001), T2WI hyperintensity (P=0.001), and enhancement pattern at 70 s (P=0.048; Table 4; pairwise comparison of MT ratio and wall thickness among fibrosis groups are shown in Supplementary Figs. 7, 8). Significant correlations with histological fibrosis scores were observed for MT ratio (r=0.681, P<0.001) and wall thickness (r=0.461, P<0.001; Supplementary Table 4). After controlling for the influence of histological inflammation scores, the histological fibrosis scores exhibited weak correlations with T2WI hyperintensity (r=0.189, P=0.091) and enhancement pattern at 70 s (r=0.073, P=0.516).

Table 4.

Association between MRI parameters and histological fibrosis score

	Histological fibrosis score
MRI parameters	None or Mild (n=13)	Moderate (n=42)	Severe (n=27)	P
MT ratio, mean ± SD (range)	23.87±4.39 (19.10–33.01)	35.41±5.54 (21.53–47.13)	40.85±4.89 (32.31–50.79)	<0.001 b
Wall thickness, mean ± SD (range)	4.94±2.46 (2.50–10.30)	9.42±2.72 (3.34–17.90)	11.96±4.08 (6.86–22.80)	<0.001 b
High signal intensity on T2WI	2 (15.4%)	35 (83.3%)	21 (77.8%)	<0.001 a
Enhancement pattern 30 s				0.091 a
1: homogeneous	11 (84.6%)	18 (42.9%)	16 (59.3%)
2: mucosa only	2 (15.4%)	20 (47.6%)	8 (29.6%)
3: layered	0	4 (9.5%)	3 (11.1%)
Enhancement pattern 70 s				0.048 a
1: homogeneous	11 (84.6%)	16 (38.1%)	15 (55.6%)
2: mucosa only	2 (15.4%)	22 (52.4%)	9 (33.3%)
3: layered	0	4 (9.5%)	3 (11.1%)
Enhancement pattern 7 min				0.580 a
1: homogeneous	11 (84.6%)	26 (61.9%)	17 (63.0%)
2: mucosa only	2 (15.4%)	12 (28.6%)	7 (25.9%)
3: layered	0	4 (9.5%)	3 (11.1%)
Enhancement progression				0.192 a
1: Stable over time	9 (69.2%)	17 (40.5%)	13 (48.1%)
2: Progression over time	4 (30.8%)	25 (59.5%)	14 (51.9%)
Percentage of mucosa enhancement gain at 7 min vs 30 s, mean ± SD (range)	60.03±53.94 (−3.13–169.89)	59.24±56.14 (−11.15–192.03)	58.61±55.27 (−3.03–200.57)	0.962 b
Percentage of submucosa enhancement gain at 7 min vs 30 s, mean ± SD (range)	63.63±52.98 (−3.13–169.89)	88.89±67.38 (−10.93–274.10)	72.66±50.97 (11.02–200.57)	0.417 b
Percentage of mucosa enhancement gain at 7 min vs 70 s, mean ± SD (range)	11.60±18.78 (−17.62–49.88)	6.89±19.92 (−16.03–86.18)	5.26±13.11 (−16.66–42.44)	0.437 b
Percentage of submucosa enhancement gain at 7 min vs 70 s, mean ± SD (range)	10.94±17.88 (−17.62–49.88)	32.13±34.93 (−13.23–136.11)	19.66±27.11 (−17.10–106.77)	0.107 b
Percentage of mucosa enhancement gain at 30 s vs precontrast, mean ± SD (range)	82.89±35.22 (20.10–161.48)	90.50±65.86 (−1.92–244.60)	121.04±89.42 (−5.22–294.02)	0.473 b
Percentage of mucosa enhancement gain at 70 s vs precontrast, mean ± SD (range)	153.11±49.40 (88.64–225.54)	165.61±68.13 (26.54–314.14)	207.03±87.77 (57.76–408.60)	0.088 b
Percentage of mucosa enhancement gain at 7 min vs precontrast, mean ± SD (range)	184.67±80.22 (74.88–311.48)	176.72±59.68 (35.24–334.69)	219.29±84.26 (66.99–365.79)	0.069 b

MT magnetization transfer, SD standard deviation, T2WI T2-weighted imaging

^aChi-squared test;

^bKruskal-Wallis test

Independent MRI indicators of intestinal inflammation and fibrosis

Multivariate logistic regression was conducted using a forward stepwise procedure with P_in=0.05 and P_out=0.10. The dependent variable “inflammation” was defined as a binary variable, where 0 and 1 meant mild and moderate-to-severe inflammation, respectively. Similarly, the dependent variable “fibrosis” was defined as 0 and 1 representing no-to-mild and moderate-to-severe fibrosis, respectively.

The resulting inflammation model includes T2WI hyperintensity (odds ratio [OR]=13.075, 95% CI 2.251–75.933, P=0.004) and wall thickness (OR=2.097, 95% CI 1.406–3.129, P<0.001). The fibrosis model includes MT ratio (OR=1.496, 95% CI 1.169–1.916, P=0.001) and wall thickness (OR=1.888, 95% CI 1.120–3.182, P=0.017).

Based on the results of the group difference analysis and the multivariate logistic regression, we proposed that T2WI hyperintensity and MT ratio could serve as MRI markers to identify moderate-to-severe inflammation and fibrosis, respectively. Since wall thickness was included in both the inflammation and fibrosis logistic models, it was not considered a specific MRI marker. The ROC curve analysis (Fig. 2) showed that the AUC of T2WI hyperintensity for differentiating mild from moderate-to-severe inflammation was 0.835 (95% CI 0.722–0.949, P<0.001); the diagnostic sensitivity and specificity values were 0.871 and 0.800, respectively. The AUC of the MT ratio discriminating mild from moderate-to-severe fibrosis was 0.964 (95% CI 0.926–1.000, P<0.001) with a sensitivity of 0.913, a specificity of 0.923, and a cutoff value of 31.25% (Fig. 2).

Fig. 2.

ROC curve analysis demonstrating that T2WI hyperintensity has a high accuracy to distinguish moderate-to-severe from mild inflammation with an AUC of 0.835, whereas MT ratio has a high accuracy, with an AUC of 0.964, for differentiating moderate-to-severe from absent-to-mild fibrosis of the bowel walls. MT magnetization transfer, T2WI T2-weighted imaging

Characterization of intestinal strictures according to combined MRI indicators

Based on these identified indicators for moderate-to-severe inflammation and fibrosis of CD lesions, we categorized CD lesions into four groups considering both bowel inflammation and fibrosis degrees (Fig. 3). In this new MRI classification, CD lesions with T2WI hyperintensities were considered to indicate moderate-to-severe inflammation, whereas those with high MT ratios (>31.25%) were indicative of moderate-to-severe fibrosis. In the 82 specimens, 13.41% (11/82) specimens were classified as mild inflammation and mild fibrosis, 8.54% (7/82) as moderate-to-severe inflammation and mild fibrosis, 15.85% (13/82) as mild inflammation and moderate-to-severe fibrosis, and the remaining 62.20% (51/82) were both moderate-to-severe inflammation and fibrosis. The combined MRI indicators and the histological findings for the classification of intestinal strictures showed a moderate agreement (κ=0.549, 95% CI 0.382–0.716; P<0.01).

Fig. 3.

New proposed classification into four different categories according to the combination of T2WI SI and MT ratio. Low: low degree of both inflammation and fibrosis (normal SI on T2WI and MT ratio≤31.25%); high: high degree of both inflammation and fibrosis (high SI on T2WI and MT ratio>31.25%); only high fibrosis: high degree of fibrosis but low degree of inflammation (normal SI on T2WI and MT ratio>31.25%); and only high inflammation: high degree of inflammation but low degree of fibrosis (high SI on T2WI and MT ratio≤31.25%). MT magnetization transfer, SI signal intensity, T2WI T2-weighted imaging

MT ratio and T2WI hyperintensity characterizing bowel lesions in the validation set

In additional 14 prospectively identified patients, 37 bowel specimens were collected for the validation set. These specimens were histologically graded regarding inflammation as none (n=2), mild (n=18), moderate (n=15), or severe (n=2) and regarding fibrosis as none (n=0), mild (n=7), moderate (n=27), or severe (n=3). Using an MT ratio>31.25% as the cutoff level for distinguishing moderate-to-severe from absent-to-mild bowel wall fibrosis, 28 specimens were categorized as moderate-to-severe and 9 as absent-to-mild. The sensitivity, specificity, PPV, NPV, and diagnostic accuracy of the MT ratio were 90% (27/30), 86% (6/7), 96% (27/28), 67% (6/9), and 89% (33/37), respectively. T2WI hyperintensity had a sensitivity of 94% (16/17) and a specificity of 80% (16/20) for diagnosing moderate-to-severe bowel wall inflammation, and the PPV, NPV, and diagnostic accuracy were 80% (16/20), 94% (16/17), and 86% (32/37), respectively. According to the previously introduced classification method, good agreement between combined MRI indicators and histological classifications was confirmed in the validation set (κ=0.665, 95% CI 0.465–0.865; P<0.01).

Discussion

This study found that the MT ratio could independently differentiate moderate-to-severe from absent-mild fibrosis, whereas T2WI hyperintensity could distinguish moderate-to-severe from mild inflammation. By combining these parameters, bowel fibrosis and inflammation could be simultaneously characterized. We combined T2WI hyperintensity and MT ratio to develop a novel MRI classification method for evaluating inflammation and fibrosis in the same bowel segment, with moderate and good agreements in our derivation and validation sets. This MRI classification may help clinicians to determine whether a stricture is predominantly fibrotic requiring surgery or dilation or mainly inflammatory with mild fibrosis that is more likely responding to medical management.

MT ratio has been reported as a quantitative MRI indicator of bowel fibrosis in experimental [14–16] and human [22] CD. Previously, Pazahr et al demonstrated that MT ratio has sufficient power to identify fibrotic segments (AUC 0.98±0.01, 95% CI; [22]). Their optimal cutoff was 28.7% with a sensitivity of 95% and a specificity of 90%. In our study, the MT ratios of bowel walls increased with the fibrosis degree allowing the differentiation between mild and moderate-to-severe fibrosis with an AUC of 0.964. Besides, it is worth mentioning that MT ratio significantly correlated with fibrosis degree but only weakly with inflammation degree, which indicates that MT ratio is sensitive to changes in fibrosis severity but relatively insensitive to inflammation. Similar to our study, Rimola et al identified oedema as an independent predictor of severe inflammation. They used the combination of T2WI SI and enhancement gain (70 s and 7 min) to simultaneously assess fibrosis and inflammation [11]. Their moderate agreement between MRI and pathology was not validated in independent patient cohorts.

Contrast-enhanced MRI has been reported to help characterize disease activity [7, 24–26], but attempts to distinguish between inflammatory and fibrotic stricture have been controversial. Rimola et al [11] demonstrated that the percentage of enhancement gain is a reliable MRI parameter for detecting the presence of severe fibrosis regardless of any coexisting inflammation, whereas other investigators [27, 28] suggested that the degree of mural enhancement increases with the grade of intestinal inflammation. We found that mural enhancement ratios were associated neither with inflammation scores nor fibrosis scores. This is supported by Punwani et al [10] demonstrating that acute inflammatory scores are not correlated with mural enhancement at 30 and 70 s. Besides, mural enhancement-related MRI parameters, e.g., enhancement pattern at 7 min and percentage of submucosa enhancement gain at 7 min vs 30 s, showed a low interobserver reproducibility, potentially limiting their use to assess intestinal inflammation and fibrosis.

Some studies reported that mural thickness is associated with inflammation [6, 10, 29], whereas other groups reported that it correlates with both inflammation and fibrosis [3, 11]. Similar to these studies, wall thickness in our study correlated with both inflammation and fibrosis. Although recent consensus [30] indicates that wall thickness generally correlates with endoscopic and histologic inflammation severity, wall thickness is not an appropriate parameter to discriminate between inflammatory and fibrotic strictures, because both acute inflammation and fibrosis deposition can lead to wall thickening [10].

Our study has several limitations. First, our enrolled patients were all selected for surgery, representing a population with higher grades of inflammation and fibrosis. Hence, only a small number of patients had low inflammation or fibrosis degrees. Second, the precise location correlations between surgical specimens and intestinal MRI is difficult due to intestinal motility. Finally, our study is partly retrospective and includes only small-bowel lesions to create a more homogeneous and focused patient population.

In conclusion, our study suggests that MT combined with conventional MRI can simultaneously characterize bowel fibrosis and inflammation. T2WI hyperintensity and MT ratio enable the accurate differentiation of inflammatory and fibrotic intestinal strictures in CD patients.

Key points.

• MT ratio from magnetization transfer imaging combined with T2WI from conventional MRI can simultaneously characterize bowel fibrosis and inflammation in adult Crohn’s disease.

Abbreviations:

CD

Crohn’s disease

CDAI

Crohn’s Disease Activity Index

CI

confidence interval

H&E

haematoxylin and eosin

ICC

intraclass correlation coefficient

MT

magnetization transfer

MTI

magnetization transfer imaging

OR

odds ratio

SI

signal intensity

T1WI

T1-weighted imaging

T2WI

T2-weighted imaging