Clinical Study On Differentiating Benign And Malignant Ovarian Tumors By Multi-modal MRI Radiomics

Objective:1.The imaging findings of benign and malignant ovarian tumors diagnosed by traditional MRI were analyzed retrospectively,and its value in qualitative diagnosis of benign and malignant ovarian tumors was discussed.2.Multi-modal MRI radiomics were used to find the optimal parameters and establish a predictive model to explore its value in the qualitative diagnosis of benign and malignant ovarian tumors.3.Based on the above research,to explore whether multi-modal MRI radiomics are superior to traditional MRI in differential diagnosis of ovarian tumors.Methods:1.From July 2018 to October 2019,pelvic MRI examination was performed in the second Hospital of Jilin University,and then surgical resection and pathological examination were performed.A total of 93 patients with 110 lesions were included.According to the pathological results,110 lesions were divided into benign group(n =62)and malignant group(n = 48).The MRI findings of ovarian tumors were qualitatively analyzed,including location,shape,boundary,intra-lesion components,DWI signal intensity,pelvic effusion and enhancement degree.SPSS 26 statistical software was used to analyze the effectiveness of traditional MRI in the differential diagnosis of benign and malignant ovarian tumors.2.From July 2018 to October 2019,pelvic MRI examination was performed in the second Hospital of Jilin University,and then surgical resection and pathological examination were performed.A total of 93 patients with 110 lesions were included.The DWI sequence,axial fat compression T2 WI sequence,T1 LAVA enhancement sequence and sagittal T2 WI sequence images of 110 ovarian tumor lesions were introduced into ITK-SNAP software by imaging group method.The multi-modal sequence lesions were delineated layer by layer to generate three-dimensional ROI images.The ROI was analyzed by A.K software,and the relevant imaging parameters of lesions were extracted.Using R 3.5.1 and Python 3.5.6 software,the extracted parameters were divided into training group(n = 76)and test group(n = 34)according to the proportion of 7:3.Then the dimension of histological parameters was reduced,the best feature parameters were obtained,and the prediction model was established,and the differential diagnosis efficiency of the model was verified by statistical methods.Results:1.Traditional MRI imaging diagnosis method can distinguish between benign and malignant ovarian tumors.The results showed that there were significant differences in lesion shape(P=0.000),lesion boundary(P=0.000),pelvic effusion(P=0.000),intralesion component(P=0.000),DWI signal intensity(P=0.002)and enhancement degree(P = 0.000),but there was no significant difference in focus distribution(P=0.757).The AUC value(area under the curve)of traditional MRI imaging in the differential diagnosis of benign and malignant ovarian tumors was 0.835,the sensitivity was 0.919,and the specificity was 0.750.2.Based on DWI combined with axial fat pressing T2 WI,T1 LAVA enhancement and sagittal T2 WI sequence,the AUC values of the multimodal imaging prediction model in the test group were 0.839(95% CI 0.712-0.948),sensitivity and specificity were 0.684 and 0.933 respectively.Conclusion:1.Traditional MRI examination can make qualitative diagnosis of benign and malignant ovarian tumors.benign tumors generally show regular lesions,clear boundaries,uniform components,no pelvic effusion,iso-low signal intensity on DWI,no enhancement,and some benign tumors show unclear boundaries,obvious limitation of DWI and obvious enhancement.Malignant tumors are generally characterized by irregular lesions,unclear boundaries,complex components in the lesions,pelvic effusion,iso-high signal intensity on DWI,enhanced enhancement,and only a few malignant tumors have clear boundaries,regular morphology and mild enhancement.Therefore,the traditional MRI imaging examination can make a preliminary qualitative diagnosis of ovarian tumors before operation and provide clinical basis for the treatment of patients.2.The multimodal prediction model based on DWI combined with axial fat compression T2 WI,T1 LAVA enhancement and sagittal T2 WI sequence has good ability in differential diagnosis of benign and malignant ovarian tumors,provides the most valuable preoperative information for clinical diagnosis and treatment,and guides clinicians to make personalized diagnosis and treatment plans for patients.3.The predictive model established by multimodal MRI imaging has a good effectiveness in the differential diagnosis of benign and malignant ovarian tumors before operation,and the diagnostic effectiveness is higher than that of traditional MRI imaging.Imaging will be further developed in the future,has a good prospect in the diagnosis of ovarian tumors,and will become a right-hand man in clinical diagnosis of ovarian tumors.