Background

More than half of patients with Crohn’s disease (CD) require at least one surgery for symptom management; however, approximately half of the patients may experience postoperative anastomotic recurrence (PAR).

Objectives

This study aims to develop and validate a preoperative computed tomography enterography (CTE)-based radiomics signature to predict early PAR in CD.

Design

A total of 186 patients with CD (training cohort, n = 134; test cohort, n = 52) who underwent preoperative CTE and surgery between January 2014 and June 2020 were included in this retrospective multi-centre study.

Methods

106 radiomic features were initially extracted from intestinal lesions and peri-intestinal mesenteric fat, respectively; significant radiomic features were selected from them and then used to develop intestinal or mesenteric radiomics signatures, using the least absolute shrinkage and selection operator and a Cox regression model. A radiomics-based nomogram incorporating these signatures with clinical-radiological factors was created for comparison with a model based on clinical-radiological features alone.

Results

68 of 134 patients in training cohort and 16 of 52 patients in test cohort suffered from PAR. The intestinal radiomic signature (hazard ratio [HR]: 2.17; 95% confidence interval [CI]: 1.32–3.58; P = 0.002) and mesenteric radiomic signature (HR: 2.19; 95% CI: 1.14–4.19; P = 0.018) were independent risk factors for PAR in the training cohort as per a multivariate analysis. The radiomics-based nomogram (C-index: 0.710; 95% CI: 0.672–0.748) yielded superior predictive performance than the clinical-radiological model (C-index, 0.607; 95% CI: 0.582–0.632) in the test cohort. Decision curve analysis demonstrated that the radiomics-based nomogram outperformed the clinical-radiological model in terms of clinical usefulness.

Conclusions

Preoperative mesenteric and intestinal CTE radiomics signatures are potential non-invasive predictors of PAR in postoperative patients with CD.

中文翻译：

---

基于术前计算机断层扫描肠造影的放射组学特征：克罗恩病患者术后吻合口复发的潜在预测因子

背景

超过一半的克罗恩病 (CD) 患者需要至少进行一次手术来控制症状；然而，大约一半的患者可能会出现术后吻合口复发 (PAR)。

目标

本研究旨在开发和验证基于术前计算机断层扫描肠造影 (CTE) 的放射组学特征，以预测 CD 的早期 PAR。

设计

这项回顾性多中心研究纳入了 2014 年 1 月至 2020 年 6 月期间接受术前 CTE 和手术的 186 名 CD 患者（训练队列，n = 134；测试队列，n = 52）。

方法

最初分别从肠道病变和肠周肠系膜脂肪中提取了 106 个放射组学特征；从中选择重要的放射组学特征，然后使用最小绝对收缩和选择算子和 Cox 回归模型，用于开发肠道或肠系膜放射组学特征。创建了一个基于放射组学的列线图，将这些特征与临床放射学因素结合起来，用于与仅基于临床放射学特征的模型进行比较。

结果

训练组 134 名患者中的 68 名和测试组 52 名患者中的 16 名患有 PAR。肠道放射组学特征（风险比 [HR]：2.17；95% 置信区间 [CI]：1.32-3.58；P  = 0.002）和肠系膜放射组学特征（HR：2.19；95% CI：1.14-4.19；P  = 0.018）根据多变量分析，是训练队列中 PAR 的独立危险因素。基于放射组学的列线图（C 指数：0.710；95% CI：0.672–0.748）在测试队列中产生了优于临床放射学模型（C 指数，0.607；95% CI：0.582–0.632）的预测性能。决策曲线分析表明，基于放射组学的列线图在临床实用性方面优于临床放射学模型。

结论

术前肠系膜和肠道 CTE 放射组学特征是 CD 术后患者 PAR 的潜在非侵入性预测因子。