Magnetic resonance imaging for suspected perianal Crohn's disease in children: a multi-reader agreement study.

Objectives: We aimed to assess inter-radiologist agreement when interpreting pelvic MRI in children with newly diagnosed perianal Crohn's disease (CD).

Materials and methods: In this retrospective multi-reader study, we identified pediatric patients (< 18 years of age) who underwent a pelvic MRI examination for newly diagnosed perianal CD. Images were de-identified and uploaded to a cloud-based image platform for review by 13 fellowship-trained pediatric radiologists The reviewers assessed for the presence of a fistula and abscess, categorization of different imaging findings, and classification using the Parks and St James' University Hospital systems. Fleiss' kappa (κ) statistics and intra-class correlation coefficients (ICC) were used to measure inter-reader agreement, along with 95% confidence intervals (CI).

Results: Forty-six patients were included in our study (median age = 13.0 years [IQR: 10.5 to 16.0 years]); thirty-five (76.1%) were boys. Most imaging features showed fair agreement (κ = 0.21 to 0.35). There was moderate agreement for categorical fistula length (κ = 0.42 [95% CI: 0.32 to 0.53]), involvement of the genitalia (κ = 0.45 [95% CI: 0.26 to 0.63]), and presence of an abscess/collection (κ = 0.52 [95% CI: 0.31 to 0.73]). Maximum abscess/collection length had good agreement (ICC = 0.81 [95% CI: 0.41, 1.00]). There was an almost equal split (yes vs. no: 50.7% vs. 49.3%) regarding whether postcontrast T1-weighted images added value compared to T2-weighted images alone across all radiologists and examinations.

Conclusion: Inter-radiologist agreement when interpreting pelvic MRI for perianal CD in children is fair for most imaging features, with fewer features demonstrating moderate or good agreement.

Key points: Question Pelvic magnetic resonance imaging (MRI) is used for diagnosing and monitoring children with perianal Crohn's disease (CD). Limited information is known about inter-radiologist agreement. Findings Agreement between pediatric radiologists when interpreting MRI for perianal CD in children is only fair for most imaging features (κ = 0.21 to 0.35). Clinical relevance Understanding MRI inter-radiologist agreement is crucial to improve the reliability of pelvic MRI in children with perianal Crohn's disease since it may affect patient management (e.g., surgery); further radiologist education and improved imaging feature definitions may help improve inter-radiologist agreement.