A Model Study Based On CT Finding To Predict Malignant Risk Of Ovarian Mass

Objective: To discuss a logistic regression model which based on CT finding alone to predict the risk of malignancy in ovarian masses,in order to find an available and non-injure way to distinguish the nature of ovarian masses before surgery.Methods: We collected 305 patients with ovarian cancer proved by pathology from January 1,2011 and April 30,2016 and 100 patients with ovarian benign lesions between January 1,2015 and April 30,2016 were recruited randomly as control in our study from West China second hospital of Sichuan University.According to classify malignant cases in chronological order and random way to divided benign cases.All patients were divided into experimental group and validation group.CT and ultrasound(US)imaging were assessed.The logistic regression model was established by taking the pathological diagnosis as the gold standard by extracting the CT image of the ovarian mass in the experimental group,and the model was validated with the validation group.Meanwhile referring to risk of malignancy index(RMI3/4),we designed a new model which based on CT imaging(RMI3-CT/RMI4-CT).Diagnosis efficacy was evaluated among LR-CT,RMI3-CT,RMI4-CT,RMI3 and RMI4 using receiver operating characteristic(ROC)curve.Results: 1.LR-CT is related to the imaging of solid component,presence of septation,necrosis,omental and/or mesentery metastasis andextent of enhancement.2.The sensitivity of CT is better than US for diagnosis of ovarian mass,an excellent agreement between FIGO and CT regarding the grading of these adnexal masses was obtained(? = 0.827).3.The RMI scoring systems based on CT are not effectively than based on US.the RMI4 scoring system based on CT and US is not better than RMI3.But ovarian vascular pediele sign is helpful for differentiating the origin of a pelvic mass and for confirming the origin of the ovary side on contrast-enhanced CT scan.4.The cutoff value to predict ovarian malignancy risk of LR-CT in ROC curve is 0.8153(81.53%).The area under the ROC curve of LR-CT(0.947)was better than the other four models whether in experimental group or validation group,There was no statistical difference among all models(P >0.05),the models based on CT imaging(LR-CT,RMI3-CT,RMI4-CT)are available as RMI3/ RMI4 to identify the nature of ovarian mass and LR-CT is the best one.5.Compare with other models,although the diagnosis efficiency of LR-CT is mild,the false positive rate and misdiagnosis rate in early stage is still relatively high.Conclusion: The logistic regression model based on CT imaging is an available method in predict the nature of the ovarian masses preoperative,it maybe becoming an available way in clinic prospective diagnosis.