| Part one Evaluation of the value of lymph node metastasis in breast cancer based on MRI features of lymph node morphologyObjective:To investigate the diagnostic value of MRI features based on lymph node morphology in preoperative lymph node metastasis of breast cancer.Methods:Continuously collected 586 patients with breast cancer in our hospital.Patients who underwent radiotherapy and chemotherapy,axillary lymph nodes and breast masses without MRI or puncture pathological findings before MRI were excluded.All patients underwent MRI plain scan plus DWI before operation.After or after the pathological examination results are gold standards.MRI images were jointly evaluated by two imaging physicians to diagnose whether lymph nodes were metastasized.If the diagnosis results were inconsistent,the images were evaluated together and consensus was reached.The accuracy,sensitivity,specificity,positive predictive value and negative predictive value of MRI in the diagnosis of lymph node metastasis of breast cancer were compared and analyzed.To verify the diagnostic efficacy of MRI on lymph node status.Result:The pathological results of lymph nodes showed a total of 300metastatic lymph nodes and 286 non-metastatic lymph nodes.A total of 250metastatic lymph nodes were diagnosed on MRI images,and 336 lymph nodes were not metastasized.Combined with the imaging features of lymph node morphology,structure and size in comparison with MRI images.The benign and malignant accuracy of lymph nodes was 78.8%,the sensitivity was 71.0%,the specificity was 87.1%,the positive predictive value was 85.2%,the negative predictive value was 74.1%,Theχ~2 test P value<0.05,both of which are statistically significant.Part two Application of extra tree for prediction of breast cancer lymph node metastasis via multi-parameter analysisObjective:The Extra Tree method was used to study the risk factors associated with axillary lymph node metastasis in breast cancer,in order to improve the diagnostic efficiency of MRI on lymph nodes,and to establish a predictive model of preoperative breast cancer axillary lymph node metastasis.Methods:Continuously and retrospectively analyzed 387 cases of breast cancer confirmed by surgery or puncture pathology from March 2015 to March 2017 in our hospital.All patients underwent radiotherapy or chemotherapy before MRI.All patients had axillary lymph nodes with surgical or puncture pathological results;MRI images were read by two radiologists.When the two opinions were not uniform,the agreement was reached and the MRI features of lymph nodes and breast masses were qualitatively and quantitatively described.All data were used separately.Extra Tree in integrated learning method for feature screening and multi-factor logistic regression analysis,compare the diagnostic efficacy of the two,select the research method with the best diagnostic efficacy,and screen out the risk factors or predictors of lymph node metastasis in breast cancer.Related predictive models.Result:The diagnostic model established by the Extra Tree method has better diagnostic performance than the logistic regression model.When using the Extra Tree method for data feature selection,breast mass location,lymph node liquefaction necrosis,breast skin thickening invasion,lymph node shortest diameter,MRI breast mass BI-RADS classification,mass edge,age,lymph node TIC curve contribution rate High,above mean,which is a risk factor for axillary lymph node metastasis of breast mass.The area under the ROC curve(AUC value)of the correlation prediction model was 0.86,the accuracy rate was 0.84,the sensitivity was 0.85,and the specificity was 0.73;Nipple invasion,short aspect ratio of breast mass,lymph node structure disappearance,lymph node ADC value,breast mass ADC value,breast mass enhancement mode,lymph node short diameter,breast mass TIC curve,lymph node morphology,breast mass morphology,chest wall muscle receptor Invasion,lymph node fusion,early enhancement rate of breast masses have a lower contribution rate,and the correlation with axillary lymph node metastasis of breast mass is relatively low.The area under the ROC curve(AUC value)established by multivariate logistic regression was 0.80,the accuracy was 0.80,the sensitivity was 0.80,and the specificity was 0.73.Conclusion:1.The characteristics of lymph node morphology based on MRI imaging have a certain diagnostic value for the evaluation of lymph node metastasis of breast cancer,but its diagnostic efficacy needs to be improved.2.The results show that the predictive model of breast cancer MRI lymph node metastasis established by Extra Tree method is a useful tool for evaluating lymph node metastasis in breast cancer patients,can provide the contribution rate of each risk factor,and its comprehensive diagnostic efficiency improves the treatment of axillary lymph node in clinical breast cancer patients.And prognosis provides an effective reference. |
PDF Full Text Request