Building and application of breast magnetic resonance imaging BI-RADS 4 lesions for predicting the risk of malignancy using a nomogram diagnostic model
Objective:To retrospectively analyze the clinical and imaging data of the cases diagnosed as BI-RADS category 4 lesions by MRI,screen out the influencing factors for breast malignancy,establish a nomogram diagnostic model for predicting breast malignancy risk,and evaluate its diagnostic value.Methods:We collected case data of breast MRI diagnosis as BI-RADS category 4 lesions in Zhongnan Hospital of Wuhan University from January 2017 to December 2020.The data was divided into mass type and non-mass enhancement(NME)type,then randomly divided into experimental group and vali-dation group in a ratio of 3∶1 in each type.The clinical and imaging characteristics of these cases were analyzed and compared,using logistic regression analysis to screen for independent risk factors of ma-lignant lesions,and the ROC curve was drawn to analyze the value,construct a nomogram diagnostic model,and reclassify cases.Results:In this study,a total of 418 cases met the inclusion criteria.Age greater than 39.5 years old,marginal spiculation,maximum intensity projection(MIP)positivity,ef-flux type in time-intensity curve(TIC),and apparent diffusion coefficient(ADC)<1.063×10-3 mm2/s were independent risk factors for malignancy in mass type lesions.The area under the ROC curve(AUC)value was 0.968(95%CI:0.943-0.994);The C-index of the malignant column chart di-agnostic model in experimental group and validation group were 0.962(95%CI:0.930-0.994)and 0.922(95%CI:0.846～0.998),respectively.22.1%(52/235)of lesions were reclassified and down-graded to category 3,and 11.9%(28/235)of lesions were upgraded to category 5.In NME-type le-sions,calcification,ADC<1.036×10-3 mm2/s,MIP positivity,and lesion distribution type(focal,multi-regional,diffuse,segment)were independent risk factors for malignancy,while skin redness and swelling were protective factors.The AUC value was 0.964(95%CI:0.934-0.995).The C-in-dex of the malignant nomogram diagnostic model experimental group and validation group were 0.956(95%CI:0.918-0.994)and 0.905(95%CI:0.816-0.994),respectively.16.4%(30/183)of NME le-sions were reclassified and downgraded to category 3,and 18.0%(33/183)of NME lesions were up-graded to category 5.Conclusion:The malignant risk prediction nomogram diagnostic model based on clinical and MRI imaging features of breast BI-RADS category 4 lesions has high diagnostic value and can be used for reclassification.
NETL
NSTL
万方数据
