Secondary Logo

Journal Logo

Original Articles: Gastroenterology: Inflammatory Bowel Disease

Texture Analysis of Magnetic Resonance Enterography Contrast Enhancement Can Detect Fibrosis in Crohn Disease Strictures

Tabari, Azadeh∗,‡; Kilcoyne, Aoife†,‡; Jeck, William R.§; Mino-Kenudson, Mari§; Gee, Michael S.∗,‡

Author Information
Journal of Pediatric Gastroenterology and Nutrition: November 2019 - Volume 69 - Issue 5 - p 533-538
doi: 10.1097/MPG.0000000000002454
  • Free


See “Intestinal Wall Texture Analysis: Finding Fibrosis in Pediatric Patients With Crohn Disease" by Phelps on page 513.

What Is Known/What Is New

What Is Known

  • Stricture types are important to distinguish in Crohn disease as inflammatory strictures may be relieved by medical therapy, whereas fibrotic strictures typically require mechanical intervention.
  • Collagen deposition in fibrotic strictures occurs in the submucosal and subserosal bowel layers; fibrosis cannot be assessed by endoscopic visualization or biopsy and currently requires full-thickness bowel resection.
  • Magnetic resonance texture analysis, a noninvasive alternative for stricture characterization, quantitatively assesses patterns of signal intensity heterogeneity among voxels within a region of interest.

What Is New

  • Novel quantitative magnetic resonance enterography biomarkers using texture analysis differentiate fibrotic from inflammatory strictures that can guide therapy in pediatric patients with Crohn disease.

Crohn disease (CD) is a relapsing inflammatory condition that can involve any portion of the gastrointestinal tract (1–3). The incidence of CD in children is increasing in the United States, with approximately 20% of patients now diagnosed in childhood or adolescence (2–4). CD is characterized by progressive transmural intestinal inflammation. One of the most common complications of CD are strictures, occurring in approximately 10% of patients at the time of diagnosis and 1/3 of patients over time with incidence increasing with disease duration (5,6). Strictures and their resultant obstructed bowel cause acute symptoms, including abdominal pain and vomiting, frequently associated with the need for hospital admission and surgical bowel resection (7). In addition, there is recent evidence that penetrating disease complications from small bowel CD occur in the setting of long-standing strictures (8). Strictures are typically classified by their histological composition and are characterized as inflammatory, fibrotic, or mixed (9). Inflammatory strictures are associated with mucosal active inflammation, fibrotic strictures are associated with submucosal and/or serosal collagen deposition, whereas mixed strictures contain elements of both. Fibrotic strictures are important to identify as they are typically unresponsive to medical treatment and often require mechanical dilation or surgical resection to alleviate patient symptoms.

As fibrosis is not discernible by endoscopic visualization or mucosal biopsy and currently requires full-thickness surgical bowel histology to diagnose, cross-sectional imaging has been explored as a noninvasive alternative for stricture characterization. Magnetic resonance (MR) enterography provides superior soft tissue contrast without the need for ionizing radiation; therefore, it has been the most extensively studied imaging modality to date; however, its performance for detecting fibrotic strictures is limited as superimposed active inflammation can mask transmural fibrosis on MRI (9,10). Combined fluorodeoxyglucose (FDG)-positron emission tomography (PET) MRI has improved performance for fibrotic stricture detection compared with MR enterography alone (11). However, the ionizing radiation exposure of PET is a significant impediment to its routine use in CD patients, who already are at risk of high cumulative radiation exposure from imaging (12).

Texture analysis is a novel tool for quantitative image analysis that has great potential for characterizing tissue composition on MR and computed tomography (CT) images. Texture analysis quantitatively assesses patterns of signal intensity heterogeneity among voxels (or pixels in 2D) within a region of interest (13), with texture analysis features providing indirect information regarding tissue composition (14). We hypothesized that the microscopic process of collagen deposition in fibrotic strictures may lead to a pattern of bowel wall contrast enhancement that is distinct that due to neutrophilic infiltration characterizing inflammatory strictures, in ways that might be detectable by quantitative texture analysis but not visually apparent. The purpose of our study was to explore whether bowel wall heterogeneity features based on texture analysis of T1-weighted fat-suppressed contrast-enhanced MR enterography images can discern small bowel stricture composition compared with histological reference in pediatric CD patients.


Study Population

This retrospective single institution Health Insurance Portability and Accountability Act-compliant study was approved by the institutional review board, with a waiver of the requirement for patient informed consent. A query of radiology and pathology report databases was performed to identify pediatric patients <18 years old who underwent elective small bowel or colonic resection within 30 days of contrast-enhanced MR enterography from the years 2009 to 2017. Electronic medical records were then queried to identify all patients with known CD and clinical evidence of stricturing disease before bowel resection.

Magnetic Resonance Enterography Imaging Protocol

MRE examinations were performed with a 1.5-T (HD Excite; GE Healthcare, USA) or 3T (Magnetom Trio; Siemens Healthineers, Erlangen, Germany) clinical MRI system with a multichannel phased-array body coil configuration. Oral contrast preparation consisted of dilute barium and sorbitol (VoLumen, E-Z-EM) or polyethylene glycol (Miralax) solution given over a 45-minute period before image acquisition. Patients consumed 450, 900,or 1350 mL of IV contrast (weighing < 25, 25–50, and >50 kg, respectively). The IV contrast agent (0.1 mmol/kg and injection rate of 1 mL/s) administered for the examinations was gadopentetate dimeglumine (Magnevist, Bayer HealthCare Pharmaceuticals, Wayne, New Jersey, USA) or Dotarem (Guerbet, Netherlands B.V.). Per clinical protocol, after contrast administration, coronal volumetric noncontrast T1-weighted fat-suppressed and contrast-enhanced fat-saturated T1-weighted images through the abdomen and pelvis were obtained at multiple time points (15). The enteric phase (55–70 seconds) postcontrast image set from each study was used for image analysis.

Magnetic Resonance Enterography Stricture Identification and Imaging-Histologic Correlation

The pathology reports from surgical bowel resection were reviewed to identify strictured regions that had histologic sections available for fibrosis assessment. All MRE studies were reviewed by 2 subspecialty-trained abdominal radiologists, with 12 and 7 years of experience interpreting MR enterography studies, respectively. The 2 radiologists in consensus identified strictures on each study, defined as areas of bowel wall thickening and luminal narrowing that corresponded to the location of bowel resection identified in the pathology reports. Only strictures involving the small bowel were included. Coronal enteric phase T1-weighted postcontrast images were then reviewed to identify the locations within each stricture corresponding to sites of histologic sections were then identified, based on distance from the ileocecal valve as described on the gross pathology and surgery reports. Regions of interest were manually drawn encompassing the entire bowel wall at these sites over a length of 2 to 3 cm. The regions of interest were drawn in a nonblinded fashion due to the need to identify sites corresponding to regions with histologic correlation.

Magnetic Resonance Enterography Texture Analysis

Region of interest-containing images were uploaded into proprietary software for textural analysis (TexRAD, Cambridge, UK) (14). Textural analysis with a filtration-histogram technique was performed within the region of interest using previously published methodology (14). The initial filtration step employs a Laplacian of Gaussian (LoG) band-pass filtration, which extracts and highlights image features at different sizes corresponding to spatial scale filter-0 (SSF0). Heterogeneity within the region of interest was quantified without image filtration using the following histogram parameters: mean, standard deviation (SD), kurtosis, skewness (asymmetry of gray-level pixel distribution), and entropy (inhomogeneity of pixel distribution). These heterogeneity features have been described previously (14). Entropy reflects the irregularity of gray-level distribution. 

Where I is the pixel value (between I = 1 to k [where k is the highest pixel value]) in the ROI and P(I) is the probability of the occurrence of that pixel value. Higher entropy and lower uniformity represent increased heterogeneity (16).

Skewness reflects asymmetry of the pixel distribution; it is related to the average brightness of the highlighted features (predominantly bright features give positive values, predominantly dark objects give negative values), which tends to 0 with increasing number of features highlighted and moves away from 0 with intensity variation in highlighted features (17).

These heterogeneity features have previously been shown to be significantly associated with tissue composition within tumors, including necrosis, hemorrhage, and fibrosis (18).

Histopathological Assessment of Bowel Inflammation and Fibrosis From Full Thickness Resection Specimens

The histology slides were reviewed by 2 pathologists in consensus (M.M.-K. and W.J.), blinded to imaging findings, for presence/absence of intestinal fibrosis and active mucosal inflammation. All surgical bowel resection specimens were fixed in formalin, paraffin-embedded, sectioned, and stained with hematoxylin-eosin per routine clinical protocol. Bowel strictures considered fibrotic (mural fibrosis) if the bowel wall, at least submucosal and mucosal layers, exhibited collagen fiber replacement. The minimal requirement for mural fibrosis was at least 33% of the entire horizontal length of submucosa involved by collagen deposition. Active inflammation was also assessed for each sectioned bowel segment, with segments scored positive if significant mucosal activity (cryptitis, crypt abscesses, and/or erosion), and/or transmural acute inflammation (ulcer, fissure, and/or abscess formation) were present. Mixed strictures were defined whenever both fibrosis and active inflammation are present.

Laboratory inflammatory markers including erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) were recorded from blood samples that were collected within ±30 days of the MRE imaging to assess the correlation between the inflammatory markers and texture features. ESR was considered as elevated when the levels were higher than 12 mm/h and CRP, when levels were over 2.90 mg/L in accordance with the reference values used in our hospital.

Statistical Analysis

Two-way analysis of variance (ANOVA) was performed to assess significance of differences in mean values among stricture groups (inflammatory, fibrotic, and mixed). Bivariate and stepwise multivariate logistic regression were performed to identify texture analysis predictors of stricture fibrosis. Statistical significance was based on Chi-square analysis of Wald Z statistics, whereas model selection was based on Akaike Information Criteria (AIC) minimization. Goodness of fit C-statistic (AUC) was also calculated for the optimal model. All statistical analysis was performed using Stata version 13 (College Station, TX). P-values <0.05 were considered significant.


A total of 25 pediatric patients with known CD were identified who underwent bowel resection and MRE within 30 days of each other (F : M 13 : 12, mean age 16 ± 2 years). A total of 64 small bowel strictured regions were identified from these patients who had imaging-histology correlation. The histologic distribution of these bowel segments included 9 segments that showed active inflammation without fibrosis, 23 segments that showed fibrosis without active inflammation, and 32 mixed segments with concomitant active inflammation and fibrosis (Table 1). There were no correlation between the inflammatory serum marker levels (ESR and CRP) and the quantitative texture parameters (P > 0.05). The bowel wall thickness measurements on coronal images showed no significant difference among the 3 groups (6.6 ± 0.4, 7.8 ± 1.0, and 6.9 ± 1.1 mm in active, fibrotic, and mixed segments, respectively). Multiple texture analysis features were assessed for the strictured regions based on histologic classification. At SSF = 0, texture features demonstrated higher values for mean, SD, and entropy whereas lower values for skewness in the presence of fibrosis. Significant differences in mean values of mean (1176 ± 703 vs 402 ± 238, P = 0.001), SD (212 ± 131 vs 98 ± 49, P < 0.001), entropy (5.2 ± 1 vs 0.75 ± 1, P < 0.001) and skewness (−0.5 ± 0.6 vs 0.25 ± 0.8, P = 0.006) were observed based on the presence or absence of fibrosis using ANOVA test. Examples of both the MR appearance and the texture analysis feature quantitation associated with the 3 stricture types is shown in Figure 1.

Baseline characteristics of small bowel stricture types; the position and number of bowel strictures, levels of markers of inflammation, presence of fistulas and abscess in groups of inflammatory/fibrotic/mixed
Representative Magnetic resonance enterography texture analysis of Crohn disease terminal ileum strictures from 3 patients with histology confirmation. Enteric phase contrast-enhanced T1-weighted fat-suppressed (A–C), single shot T2-weighted (D–F) images, along with corresponding full thickness H&E (G–I) histology (20× magnification) are shown for patients with 3 different stricture histologic types. Texture analysis regions of interest are shown displayed over the postcontrast images (A–C) and texture analysis quantitative parameters for each stricture are shown (J–L). An 11-year-old girl demonstrates a stricture enhancement pattern (A) associated with low texture analysis mean and entropy values and positive skewness (J), with histology showing active mucosal inflammation without fibrosis (G). A 17-year-old boy demonstrates a stricture enhancement pattern (B) exhibiting high mean and entropy with negative skewness (K) values on texture analysis and histology showing submucosal fibrosis without mucosal active inflammation (H). An 11-year-old girl with a stricture enhancement pattern (C) demonstrating texture analysis high mean and entropy values and negative skewness (L), and histology showing marked submucosal fibrosis as well as mucosal active inflammation (I). White arrows indicate stricture location on imaging. On H&E images, white arrows indicate microscopic active inflammation and black arrows indicate histologic fibrosis. Note that all 3 strictures demonstrate wall thickening and edema on T2-weighted images and mural hyperenhancement on T1-weighted fat-suppressed postcontrast images, and stricture type is not easily distinguished based on visual characteristics.

Bivariate regression analysis was then performed on these texture analysis variables as predictors of absence/presence of fibrosis. Several texture analysis variables (mean, standard deviation, entropy and skewness) individually are significant predictors of fibrosis (Table 2). Stepwise logistic regression was then performed on all the MRE-Texture analysis variables to identify the best predictive model of fibrosis based on these features, using Akaike Information Criteria minimization. A model including mean, skewness, and entropy values performed best for predicting fibrosis with mean considered as the optimum regression model (P < 0.001, Goodness-of-fit C-statistic AUC of 0.995; Table 3). Using a combination of these 3 TA parameters as a threshold based on regression modeling coefficients (mean >415, entropy >2.9, Skewness <0.06), MR texture analysis was able to correctly classify 100% of the cases (9/9 strictures without fibrosis, 55/55 strictures with fibrosis).

Comparison of image heterogeneity values between fibrotic and nonfibrotic strictures
Optimal model for predicting bowel stricture fibrosis based on multivariate logistic regression of MR texture features


In this study, our focus was on the development of a novel quantitative MRE biomarker using texture analysis to discriminate fibrotic from inflammatory strictures in pediatric patients with CD, which, to our knowledge, has not been established in the literature. These types of strictures are important to distinguish as inflammatory CD strictures may be relieved by medical therapy, whereas fibrotic strictures typically require mechanical intervention. As collagen deposition in fibrotic strictures occurs in the submucosal and subserosal bowel layers, fibrosis cannot be assessed by endoscopic visualization or biopsy, and currently requires full-thickness bowel resection. This highlights the need for reliable noninvasive determination of stricture composition. Previous studies have associated bowel wall enhancement intensity on MRI postcontrast images with presence of fibrosis; however, no clear discriminatory pattern has been identified and the fact that many CD strictures are mixed makes detection of intestinal fibrosis difficult (19–25).

In this study, we applied quantitative analysis of bowel stricture enhancement pattern using texture analysis to discriminate stricture type, using full-thickness surgery pathology specimens as the reference standard. Our hypothesis was that texture analysis could identify differences in mural enhancement characteristics not discernible by radiologist visual inspection. We selected enteric phase T1-weighted fat-suppressed postcontrast images because normal small bowel, which is perfused by branches of the superior mesenteric artery, is known to demonstrate relatively uniform enhancement during this phase. MR texture analysis identifies multiple measures of bowel wall heterogeneity that are significantly associated with the presence of mural fibrosis. Mean (P = 0.0003), SD (P = 0.006), and entropy (P < 0.0001) values all were significantly increased in strictures with fibrosis (fibrotic and mixed types), whereas skewness was significantly increased in inflammatory strictures without fibrosis (P = 0.002). Taken together, these results suggest that the presence of fibrosis within a stricture is associated with a more heterogeneous mural enhancement pattern on enteric phase postcontrast images. This is consistent with previous studies that have shown that transmural hyperenhancement is an MRE feature of active CD inflammation (26,27).

Logistic regression analysis demonstrates that a combination of 3 texture analysis features (entropy, mean, and skewness) provides the optimum discrimination of fibrotic and nonfibrotic strictures, with a high degree of accuracy (AUC = 0.995). In addition, a combination of threshold values for these 3 texture analysis parameters based on logistic regression analysis (mean >415, entropy >2.9, skewness <0.06) was able to correctly classify all 64 strictures in our study cohort, including 9/9 strictures without fibrosis and 55/55 strictures with fibrosis (fibrotic or mixed). This suggests that texture analysis of MR enterography images can be a useful biomarker to predict the presence of fibrosis within strictures. Importantly, MRE-Texture analysis can identify the presence of fibrosis within mixed strictures, despite the presence of concomitant active inflammation. This represents an advance beyond prior studies investigating MRE visual interpretation, bowel wall perfusion, and quantitative analysis that have shown limitations in detecting fibrosis within mixed strictures (9,11,25,28,29). Our results showing the benefits of texture analysis of MRE images are consistent with previous studies showing texture analysis of T2-weighted images to be helpful for assessing CD activity (30). A benefit of texture analysis in general is its ability to quantitatively analyze patterns of signal intensity that are beyond the ability of the human eye to discern. Texture analysis is particularly helpful for analysis of MRE postcontrast sequences, where patterns of signal intensity are likely to be more reliable than absolute signal intensity, given the lack of standardization of intensity values on MRI. We acknowledge that texture analysis may not be easily deployed in radiology departments without this software capability. Of note, SD as an individual feature demonstrates good performance for detecting stricture fibrosis (Table 1; with higher values observed in fibrotic strictures) and is typically available as an image analysis tool on most clinical Picture Archiving and Communication Systems.

The limitations of our study are the retrospective design and relatively small number of bowel segments included. The need for full-thickness surgical bowel resection specimens as well as a close time interval between MRE and surgical bowel resection substantially reduced the number of eligible patients. In addition, we only included strictures involving the terminal ileum and excluded colonic strictures as we wanted to eliminate variations in enhancement patterns because of differential blood supply to the colon, which further reduced the number of eligible patients. Because of the requirement for surgical bowel resection, there is a potential selection bias toward patients with sufficiently advanced disease to merit surgical resection. This also resulted in a relatively small number of segments without fibrosis for analysis. Also, our ROI selection was based on manual segmentation of the bowel wall within a stricture on a single coronal image. Future studies will explore the use of volumetric segmentation tools that can perform texture analysis on bowel strictures in their entirety.

Our results focused on texture analysis of enteric phase T1-weighted fat-suppressed images for fibrosis assessment, but in the future it would be useful to combine texture analysis quantitation of multiple MRE pulse sequences including T2-weighted, and delayed postcontrast images. In fact, the inherently quantitative nature of texture analysis, with the high number of voxels analyzed per region of interest, lends itself well to artificial intelligence algorithms that can extract complex patterns from medical image data (31). This offers great potential to enhance radiologist interpretation of MRE examinations, and derive quantitative MRE biomarkers of activity and treatment response, in the future.


Texture analysis of MRE enteric phase T1-weighted fat-suppressed postcontrast images can distinguish fibrotic from nonfibrotic strictures, providing a noninvasive biomarker of stricture composition that can guide therapy.


1. Rimola J, Rodriguez S, Garcia-Bosch O, et al. Magnetic resonance for assessment of disease activity and severity in ileocolonic Crohn's disease. Gut 2009; 58:1113–1120.
2. Kappelman MD, Moore KR, Allen JK, et al. Recent trends in the prevalence of Crohn's disease and ulcerative colitis in a commercially insured US population. Dig Dis Sci 2013; 58:519–525.
3. Malaty HM, Fan X, Opekun AR, et al. Rising incidence of inflammatory bowel disease among children: a 12-year study. J Pediatr Gastroenterol Nutr 2010; 50:27–31.
4. Carmen Cuffari MD. Diagnostic considerations in pediatric inflammatory bowel disease management. Gastroenterol Hepatol (N Y) 2009; 5:775–783.
5. Jianguo Z, Faming Z, Fei L, et al. Identifying the inflammatory and fibrotic bowel stricture: MRI diffusion-weighted imaging in Crohn's disease. Radiol Infect Dis 2015; 2:128–133.
6. Rieder F, Zimmermann EM, Remzi FH, et al. Crohn's disease complicated by strictures: a systematic review. Gut 2013; 62:1072–1084.
7. Vester-Andersen MK, Vind I, Prosberg MV, et al. Hospitalisation, surgical and medical recurrence rates in inflammatory bowel disease 2003-2011—a Danish population-based cohort study. J Crohns Colitis 2014; 8:1675–1683.
8. Orscheln ES, Dillman JR, Towbin AJ, et al. Penetrating Crohn disease: does it occur in the absence of stricturing disease? Abdom Radiol (NY) 2018; 43:1583–1589.
9. Gee MS, Nimkin K, Hsu M, et al. Prospective evaluation of MR enterography as the primary imaging modality for pediatric Crohn's disease assessment. AJR Am J Roentgenol 2011; 197:224–231.
10. Seemann NM, Radhakrishnan S, Gazendam A, et al. The role of imaging in the preoperative assessment of children with inflammatory colitis. J Pediatr Surg 2017; 52:970–974.
11. Catalano OA, Gee MS, Nicolai E, et al. Evaluation of quantitative PET/MR enterography biomarkers for discrimination of inflammatory strictures from fibrotic strictures in Crohn disease. Radiology 2016; 278:792–800.
12. Desmond AN, O’Regan K, Curran C, et al. Crohn's disease: factors associated with exposure to high levels of diagnostic radiation. Gut 2008; 57:1524–1529.
13. Ganeshan B, Miles KA. Quantifying tumour heterogeneity with CT. Cancer Imaging 2013; 13:140–149.
14. Miles KA, Ganeshan B, Hayball MP. CT texture analysis using the filtration-histogram method: what do the measurements mean? Cancer Imaging 2013; 13:400–406.
15. Moy MP, Kaplan JL, Moran CJ, et al. MR enterographic findings as biomarkers of mucosal healing in young patients with Crohn disease. AJR Am J Roentgenol 2016; 207:896–902.
16. Ganeshan B, Miles KA, Young RC, et al. Hepatic entropy and uniformity: additional parameters that can potentially increase the effectiveness of contrast enhancement during abdominal CT. Clin Radiol 2007; 62:761–768.
17. Miles KA, Ganeshan B, Rodriguez-Justo M, et al. Multifunctional imaging signature for V-KI-RAS2 Kirsten rat sarcoma viral oncogene homolog (KRAS) mutations in colorectal cancer. J Nucl Med 2014; 55:386–391.
18. Sieren JC, Smith AR, Thiesse J, et al. Exploration of the volumetric composition of human lung cancer nodules in correlated histopathology and computed tomography. Lung Cancer 2011; 74:61–68.
19. Rimola J, Planell N, Rodríguez S, et al. Characterization of inflammation and fibrosis in Crohn's disease lesions by magnetic resonance imaging. Am J Gastroenterol 2015; 110:432–440.
20. Fornasa F, Benassuti C, Benazzato L. Role of magnetic resonance enterography in differentiating between fibrotic and active inflammatory small bowel Stenosis in patients with Crohn's disease. J Clin Imaging 2011; 1:35.
21. Young KA. Role of computed tomography enterography/magnetic resonance enterography: is it in prime time? Clin Endosc 2012; 45:269e73.
22. Nyree G, Lee AG, Simon A, et al. Small bowel MR enterography: problem solving in Crohn 's disease. Insights Imaging 2012; 3:251–263.
23. Stasi C, Falchini M, Milani S. Imaging modalities for the noninvasive assessment of fibrosis in Crohn's disease. Sci World J 2012; 2012:450151.
24. Dagia C, Ditchfield M, Kean M, et al. Imaging for Crohn disease: use of 3-T MRI in a paediatric setting. J Med Imaging Radiat Oncol 2008; 52:480–488.
25. Barkmeier DT, Dillman JR, Al-Hawary M, et al. enterography-histology comparison in resected pediatric small bowel Crohn disease strictures: can imaging predict fibrosis? Pediatr Radiol 2016; 46:498–507.
26. Gale HI, Sharatz SM, Taphey M, et al. Comparison of CT enterography and MR enterography imaging features of active Crohn disease in children and adolescents. Pediatr Radiol 2017; 47:1321–1328.
27. Quencer KB, Nimkin K, Mino-Kenudson M, et al. Detecting active inflammation and fibrosis in pediatric Crohn's disease: prospective evaluation of MR-E and CT-E. Abdom Imaging 2013; 38:705–713.
28. Wilkens R, Hagemann-Madsen RH, Peters DA, et al. Validity of contrast-enhanced ultrasonography and dynamic contrast-enhanced MR enterography in the assessment of transmural activity and fibrosis in Crohn's disease. J Crohns Colitis 2018; 12:48–56.
29. Li XH, Sun CH, Mao R, et al. Assessment of activity of Crohn disease by diffusion-weighted magnetic resonance imaging. Medicine (Baltimore) 2015; 94:e1819.
30. Makanyanga J, Ganeshan B, Rodriguez-Justo M, et al. MRI texture analysis (MRTA) of T2-weighted images in Crohn's disease may provide information on histological and MRI disease activity in patients undergoing ileal resection. Eur Radiol 2017; 27:589–597.
31. Hosny A, Parmar C, Quackenbush J, et al. Artificial intelligence radiology. Nat Rev Cancer 2018; 18:500–510.

imaging biomarker; inflammatory bowel disease; magnetic resonance imaging; pediatrics; texture analysis

Copyright © 2019 by European Society for Pediatric Gastroenterology, Hepatology, and Nutrition and North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition