Diagnostic value of a combined clinical-radiomics model based on MRI for the assessment of renal fibrosis in chronic kidney disease
10.3760/cma.j.cn112149-20241013-00623
- VernacularTitle:基于MRI的影像组学联合临床指标对慢性肾脏病肾纤维化的诊断价值
- Author:
Chaogang WEI
1
;
Ying ZENG
;
Qing MA
;
Zhicheng JIN
;
Yilin XU
;
Ye ZHU
;
Xiaojing LI
;
Junkang SHEN
;
Zhen JIANG
Author Information
1. 苏州大学附属第二医院影像科，苏州 215004
- Publication Type:Journal Article
- Keywords:
Magnetic resonance imaging;
Chronic kidney disease;
Renal fibrosis;
Radiomics;
Clinical indicator
- From:
Chinese Journal of Radiology
2025;59(10):1163-1169
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To explore the diagnostic value of a clinical-radiomics model based on the T 1 mapping and apparent diffusion coefficient (ADC)-based radiomics, and the clinical indicator for renal fibrosis (RF) caused by chronic kidney disease (CKD). Methods:This cross-sectional study prospectively and consecutively enrolled 122 patients with CKD at the Second Affiliated Hospital of Soochow University from September 2021 to December 2023 who were randomly allocated to a training set ( n=85) or a validation set ( n=37) in an approximate 7∶3 ratio using simple random sampling. Patients underwent T 1 mapping and diffusion-weighted imaging scans. Renal biopsy was performed within 3 days after the MRI scans. Patients were categorized into three groups based on the degree of RF: no RF ( n=25), mild RF ( n=55), and moderate to severe RF ( n=42). To differentiate the presence of RF (no RF vs. any RF) and the severity of RF (mild RF vs. moderate to severe RF), univariate and multivariate logistic regression were used to optimize the independent clinical predictor, which constituted the clinical model. Radiomics features were extracted from regions of interest delineated within the renal parenchyma of the right kidney on T 1 mapping and ADC maps. Features were selected using least absolute shrinkage and selection operator regression to build the radiomics model. A clinical-radiomics model was subsequently constructed by integrating the independent clinical predictors with the selected radiomics features. Model diagnostic performance was evaluated using the area under the receiver operating characteristic curve (AUC). Calibration curve was plotted to assess model calibration, and decision curve analysis was performed to evaluate clinical net benefit. Results:Univariate logistic regression analysis revealed that estimated glomerular filtration rate (eGFR), serum creatinine, and blood urea nitrogen exhibited statistically significant differences ( P0.05) in distinguishing both the presence and severity of RF. Multivariate analysis identified eGFR as an independent clinical predictor for both the presence of RF ( OR=0.939, 95% CI 0.898-0.982, P=0.006) and RF severity ( OR=0.956, 95% CI 0.917-0.997, P=0.037). From the MRI images, 7 radiomics features were selected to build the radiomics model for distinguishing the presence of RF, and 8 features were selected for the model assessing RF severity. These radiomics models were then combined with eGFR to construct the clinical-radiomics models. The clinical-radiomics models demonstrated the highest diagnostic performance, with an AUC of 0.935 (95% CI 0.859-0.977) for RF presence and 0.967 (95% CI 0.891-0.995) for RF severity in the training set, and 0.914 (95% CI 0.774-0.981) and 0.908 (95% CI 0.748-0.981) in the validation set. Calibration curves and decision curve analysis confirmed that the clinical-radiomics models exhibited excellent calibration and provided the highest clinical net benefit for assessing RF in CKD patients. Conclusion:The clinical-radiomics model integrating T 1 mapping and ADC-based radiomics and eGFR can effectively improve the diagnostic performance for RF in CKD patients.
