The Predictive Diagnostic Value of a Fusion Model Based on Clinical and Radiomics Features for Lymph Vascular Space Invasion in Cervical Cancer
Objective To explore the predictive diagnostic value of a fusion model based on clinical and radiomics features in cervical cancer with lymph vascular space invasion(LVSI)disease.Methods With the retrospective analysis of 192 cervical cancer patients visited the First People's Hospital in Kashgar from November 2017 to June 2022,all patients underwent MRI plain scan and dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)examination.With the collection of general patient information,preoperative laboratory data,and MRI quantitative parameters for univariate analysis,statistically significant indicators were included in Logistic multiple regression analysis to establish a clinical model;With the acquisition of late arterial phase(LAP)images and the labeling of late arterial phase(LAP)images,the radiomics characteristics[sphericity,standardized grayscale non-uniformity,clustered shadows,first-order feature kurtosis,first-order feature mean,high grayscale advantages]were extracted for binary logistic regression analysis to construct an radiomics model;Using Logistic regression to establish a fusion model based on clinical and imaging models,the receiver operating characteristic(ROC)curves were plotted separately,and the predictive performance of the three models was analyzed and evaluated.Results Among the 192 cervical cancer patients,there were 85 cases with LVSI(+)and 111 cases with LVSI(-).The area under the ROC curve(AUC)of LVSI status of cervical cancer predicted by clinical model was 0.736 in the training group and 0.768 in the validation group.The AUC of LVSI status in cervical cancer predicted by radiomics model was 0.709 in the training group and 0.682 in the verification group.The AUC of LVSI status in cervical cancer predicted by fusion model was 0.828 in training group and 0.795 in verification group.Conclusion The fusion model based on clinical and radiomics features can accurately predict the LVSI status of cervical cancer patients,with a higher preditive value than the other 2 models,and can provide a basis for subsequent adoption of non-invasive surgical treatment strategies.
cervical carcinomaImaging omicsDCE-MRIlymph vascular space
NETL
NSTL
万方数据
