Predicting Crohn's disease activity using a radiomics model based on CT enterography
10.3969/j.issn.1002-1671.2025.04.015
- VernacularTitle:基于小肠CT造影影像组学模型预测克罗恩病活动性
- Author:
Guangbing ZHANG
1
;
Luanluan WANG
1
Author Information
1. 南京医科大学附属苏州医院(苏州市立医院)放射科,江苏 苏州 215000
- Publication Type:Journal Article
- Keywords:
Crohn's disease;
activity;
computed tomography enterography;
radiomics;
machine learning
- From:
Journal of Practical Radiology
2025;41(4):608-613
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the value of a radiomics model based on computed tomography enterography(CTE)in the diagnosis of Crohn's disease(CD)activity.Methods A retrospective analysis was conducted on 166 diseased bowel segments from patients with CD.Lesions were classified into remission and active groups based on the simple endoscopic score for Crohn's disease(SES-CD).Two physicians independently delineated the volume of interest(VOI)for the lesions and extracted radiomics features.The optimal feature subset was selected through dimensionality reduction,and the Radiomics score(Radscore)was calculated.The logistic regression-clinical model(LR-CM)and logistic regression-radiomics model(LR-RM)were established based on clinical and radiomics independent risk factors,respectively.Additionally,machine learning models were established using four different machine learning algorithms.An ensemble model(EM)was established based on the prediction results of LR-RM and the four machine learning models.The area under the curve(AUC)for each model was calculated,and the performance differences between the models were compared.Results Seven optimal radiomics features were selected to calculate the Radscore.The AUC of LR-RM in the training and validation sets were 0.920 and 0.908,respectively.The AUC of EM in the training and validation sets were 0.974 and 0.965,respectively,which were significantly higher than the AUC of LR-RM model(P=0.017,P=0.048).Conclusion The radiomics model demonstrates strong diagnostic efficacy in predicting CD activity.The application of ensemble learning techniques significantly enhances the predictive performance of the radiomics model.
