Multi-sequence MRI radiomics for predicting clinical stage of cervical squamous cell carcinoma
10.13929/j.issn.1672-8475.2024.10.009
- VernacularTitle:多序列MRI影像组学预测宫颈鳞状细胞癌临床分期
- Author:
Dan ZHAO
1
;
Zixin SHI
;
Yaying SU
;
Jiaojiao LI
;
Shujun CUI
Author Information
1. 河北北方学院研究生学院,河北 张家口 075000
- Keywords:
uterine cervical neoplasms;
carcinoma,squamous cell;
magnetic resonance imaging;
neoplasm staging;
radiomics
- From:
Chinese Journal of Interventional Imaging and Therapy
2024;21(10):607-612
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the value of multi-sequence MRI radiomics for predicting clinical stage of cervical squamous cell carcinoma(CSCC).Methods Totally 190 patients with single CSCC confirmed by pathology were retrospectively collected.Among them,67 cases with International Federation of Gynecology and Obstetrics(FIGO)stage ⅠB—ⅡA were classified into early stage group,while 123 cases with FIGO ⅡB—ⅢC were enrolled in middle-late stage group.The patients were divided into training set(n=114,including 40 cases in early stage subgroup and 74 cases in middle-late stage subgroup)and test set(n=76,including 27 cases in early stage subgroup and 49 cases in middle-late stage subgroup)at the ratio of 6∶4.Single factor and logistic analyses were used to screen clinical relevant factors,and a clinical model was constructed.The best radiomics features of lesions were extracted and selected based on pre-treatment pelvic MR T2WI,diffusion-weighted imaging(DWI),dynamic contrast enhancement(DCE)-T1WI and all the three,respectively,and the radiomics models were constructed,including T2WI,DWI,DCE-TWI and combined sequences models,then a clinical-radiomics model was established based on clinical model and combined sequences model.The predictive efficacy of each model was evaluated by receiver operating characteristic curves,and the area under the curves(AUC)were calculated.The integrated discrimination improvement(IDI)index was also calculated to compare the diagnostic efficacy of each model in training set,and decision curve analysis(DCA)was used to evaluate their clinical value.Results Squamous cell carcinoma associated antigen in middle-late stage subgroup was higher than that in early stage subgroup in both training and test sets(both P＜0.05),which was used to establish the clinical model.The AUC of clinical,T2WI,DWI,DCE-TWI,combined sequences and clinical-radiomics models for predicting clinical stage of CSCC was 0.66,0.71,0.78,0.81,0.88 and 0.89 in training set,respectively,which was 0.62,0.64,0.72,0.73,0.77 and 0.76 in test set,respectively.In training set,the predictive efficacy of clinical-radiomics model was higher than that of combined sequences model(IDI=0.19,P＜0.05),both higher than that of the rest models(IDI=0.19-0.47,all P＜0.05).When the thresholds were 0.02-1.00 and 0.05-1.00,combined sequences and clinical-radiomics models had higher clinical net benefits in training set.Conclusion Multi-sequence MRI radiomics could effectively predict clinical stage of CSCC,and combining clinical data could improve its diagnostic efficacy.
