The Value Of 2D And 3D Radiomics Based On Conventional CT Scan In Predicting The Metastasis Of Ovarian Malignant Tumors

Objective:To establish an effective,non-invasive model for predicting ovarian malignant tumor metastasis,and to explore the value of plain CT imaging in predicting ovarian malignant tumor metastasis,and to compare the efficacy of 2D and 3D prediction models.Materials and methods:A retrospective search was conducted on 116 patients with ovarian malignant tumor confirmed by surgery and pathology and preoperative CT examination from three medical institutions of Jiangxi Maternal and Child Health Care Hospital,Jiangxi Provincial People’s Hospital and the First Affiliated Hospital of Nanchang University from May 2020 to August 2022,including 80 patients with metastasis and36 patients without metastasis.ITK-SNAP software was used to segment the lesions on CT images and extract features.Patients were randomly divided into training group and verification group with a ratio of 7:3,with 82 cases and 34 cases,respectively.Use max-relevance and min-redundancy(m RMR),least absolute shrinkage and selection operator,Multivariate logistic regression was used to construct a prediction model of ovarian malignant tumor metastasis combining clinical factors and imaging features.The area under curve(AUC)and decision curve were used to evaluate the efficiency of the prediction model.The Delong test was used to compare the performance of 2D and 3D prediction models.Results:After eliminating miscellaneous and irrelevant features,the 2D and 3D models ultimately retained 9 optimal features respectively,and patients with and without metastasis were grouped.Ascites associated with metastasis and tumor marker CA-125 were integrated to construct the prediction model.In the 2D model training cohort,the imaging nomogram model(AUC=0.93)had better predictive power than the clinical model alone(AUC=0.85)and the imaging nomogram model(AUC=0.88).The AUC values of the imaging model,clinical model and imaging model were 0.92,0.89 and 0.84,respectively.The imaging model still showed better predictive ability.The AUC of the training group and the test group were 0.95 and0.98,respectively.The Hosmer-Lemeshow test of 2D and 3D Normograph models showed a good coincidence.In the decision curve,the net benefit of Normograph model is obviously higher than other models.De Long’s test showed no difference in the predictive performance between 2D and 3D image omics models(p>0.05).Conclusions:2D and 3D imaging nomogram models based on CT imaging and clinical features have good efficacy in predicting the metastatic state of ovarian malignant tumors.The predictive efficacy of 2D and 3D models is equivalent to the clinical practical value.