Prediction Of Breast Cancer Was Established By Risk Score Based On Clinical And Ultrasound Images

Objectives To explore the establishment of the best model for the prediction of breast cancer by the risk scoring method based on clinical and conventional ultrasound images.Analyzing and finding the critical reference value of ultrasound and clinical parameter integral for benign and malignant breast tumors.Further optimize BI-RADS classification,to improve the accuracy of breast cancer diagnosis and decrease the unnecessary puncture biopsy or surgery.Methods A retrospective analysis was performed based on the breast ultrasound data from 1092 female patients,during January1,2014 to November 30,2018 in Affiliated Hospital of North China University of Science and Technology.821 cases were confirmed by breast biopsy or pathology.The ultrasound images were standardized by the(Breast Imaging Reporting and Data System,BI-RADS)BI-RADS,which is the 5th edition of 2013 BI-RADS diagnostic standard including part(right or left),position(quadrant),shape(rules and irregular shape),location(parallel to the surface of the skin,not parallel to the surface of the skin),edge shape(fuzzy,Angle,leaf,burr),boundary(clear or not clear),hyperechoic halo,echo,internal echo model(uniform and non-uniform),changes in the surrounding tissue(Involving the cooper ligament,skin thickening,affected retraction,edema),rear echo(enhanced,attenuation,lateral acoustic shadow),whether lymph nodes in axillary were abnormal increased,combined with Adler semi-quantitative grading of blood flow,blood flow distribution pattern(peripheral,internal,peripheral and internal),flow rate and resistance index.Database was established by Excel 2013,and SPSS 22.0 statistical software was used for statistical analysis.Different variables were coded based on the original data of the subjects.Pathological results were taken as the gold standard and used as dependent variables(1= benign,2= malignant).All other factors were used as independent variables,including general information,palpation and characteristics of conventional ultrasound images.All normality measurement data used by(sx ±);the counting data is described by composition ratio;two groups rate and composition ratio were compared with the Chi-Square test.821 subjects were randomly divided into two groups.Significant variables based on univariate analysis were included in multivariate regression analysis,and then logistic regression was used(by Forward LR method)to look for clinical risk parameters and ultrasonic malignant signs.There was statistical significance as P<0.05.The basic model RS1 was established based on the malignant ultrasound images of breast tumors.The screening variables were scored with 10 times of regression coefficient.The new models RS2,RS3 and RS4 are constructed by adding one or more parameters,and the prediction ability and accuracy of the new models are evaluated by comparing with the basic model RS1.The total score of each subject was used as the test variable,and the pathological outcome of patients was used as the state variable.The subject operating curve(ROC)was drawn to evaluate the diagnostic efficacy of the prediction model.Combined with the area under the subject's operating characteristic curve(AUC),the cut-off value was obtained when Youden index was at the maximum,and then analyzed parameter integral critical value and model sensitivity and specificity.External tests were carried out by the verification group,and the consistency test(Kappa value)was used for evaluation.The maximum Kappa value was the best model.Results 1 Risk score model was established based on different parameters:(1)model RS1 was established based on ultrasonic parameters;(2)ultrasonic parameters were combined with age to establish the model RS2;(3)comprehensive ultrasound parameters,age and clinical signs to establish model RS3;(4)combined with ultrasound parameters,age,risk factors and signs to establish model RS4.2 The consistency test of Kappa was conducted between the four scoring models and the pathological results.The Kappa value of the risk factor scoring model RS3 was the highest(Kappa =0.799),which was the best model with a score threshold of 50.8 points.The area under the curve of model RS3 is 0.962.External verification results of AUC was 0.900,while the accuracy,sensitivity and specificity for breast cancer using RS3 were 90.1%,89.1% and 90.9%,and the positive predictive value,negative predictive value,positive likelihood ratio and negative likelihood ratio were 88.3%,91.5%,9.791 and 0.120.3 When respectively compared with BI-RADS grading,RS3 prediction accuracy of breast cancer(90.1%)was obviously higher than the BI-RADS classification(75.1%),and further compared with BI-RADS classification,according to the results of ultrasonic diagnosis of BI-RADS3-5 levels of tumor.The coincidence rat of RS3 by the risk scoring method was 94.8%,88.8%,75.9%,98.1%,100%,respectively.There was also a high coincidence rate in BI-RADS4 grade tumors(grade 4a was 88.8%,grade 4b was75.9%,grade 4c was 98.1%).Conclusions 1 The RS3 risk scoring method combined conventional ultrasound images,age and clinical palpation was the optimal model for predicting breast cancer than others.2 Danger signs and risk factors of RS3 prediction of malignant breast cancer including age > 55,poor palpate mobility,cooper ligament involvement,hyperechoic halo,51 ~ years old,46 ~ years old,microcalcification,border is not smooth,palpation activity is,Which scores were 31.9,30.4,25.8,25.7,23.4,19.9,18.8,16.9,15.2,Respectively,the critical value is 50.8.Among them,poor palpation activity and > age 55 years old were contributed more to the prediction of malignant risk of breast tumor than ultrasonic malignant signs.3 The breast cancer risk scoring model based on the comprehensive epidemiological data,the characteristics of conventional ultrasound images and clinical signs,can not only better identify the patient group with BI-RADS score of 4a or a benign lesion with a low degree of suspected malignancy,but also make up for the low specificity of BI-RADS.RS3 which helps to not only reduce misdiagnosis and missed diagnosis,but also decreased the overdiagnosis of breast cancer and unnecessary puncture or surgery.Figure3;Table22;Reference 108...