Qualitative Classification Of Shear Wave Elastography And Texture Features Of Conventional Ultrasound In The Assessment Of Regional Lymph Nodes Status For Breast Cancer

Objectives:To explore the clinical application and diagnostic performance of qualitative classification of shear wave elastography(SWE)and texture features of conventional ultrasound for pre-operative evaluation of lymph node(LN).Methods:1.A total of 121 axillary lymph nodes from 118 patients were enrolled,who were scheduled for breast cancer surgery and core needle biopsy of ALNs.Conventional US and SWE were performed before biopsy in all enrolled patients.The scoring standard of conventional US used to evaluate axillary lymph nodes status was based on morphological changes,including:the length of the shortest axis,the L/S ratio,the cortical thickness and status of fatty hilum.And each ALN was evaluated by a qualitative classification method which could classify the SWE images into four color patterns:Color pattern 1:homogeneous pattern;Color pattern 2:filling defect within lymph node;Color pattern 3:homogeneous within lymph node with a localized colored area at the marginal sinus;Color pattern 4:filling defect within lymph node with a localized colored area at the margin of lymph node.The conventional US assessment and quantitative SWE parameters were recorded.The diagnostic performances of SWE parameters,conventional US,and combinations of US and SWE parameters were then compared.2.From April 2017 to December 2017,a total of 134 patients who underwent sentinel lymph node biopsy were included.All were examined by conventional ultrasonography(US)before biopsy.According to the pathological results,patients were divided into sentinel node metastasis group and non-metastasis group.Texture features of breast lesions and surrounding tissue were extracted by four matrix methods,including gray-level co-occurence matrix,gray-level run-length matrix,gray-level size zone matrix and neighbourhood gray-tone difference matrix.The method of mutInfFS and Random Forest were used to select the features.Subsequently,an ultrasonic radiomics model was established to predict the metastasis of breast cancer sentinel lymph nodes and its diagnostic performance was then analyzed.Results(1)The average score of conventional US with benign and malignant axillary lymph node was 0.54±0.79 and 2.58±0.869,respectively,with a statistically significant difference(p<0.05).The AUC value of conventional US to assess lymph node status was 0.942.The optimal cutoff value of this scoring standard was 1.5 points,and showed 91.7%sensitivity and 90.9%NPV,when combined with 82%specificity and 83.3%PPV.Compared with pathology results,the kappa value of scoring standard of conventional US assessment for axillary lymph node was 0.736.(2)The mean value of Emean,Emin,Emax,SD and Eratio were significantly higher in metastasis axillary lymph nodes than in benign axillary lymph nodes:Emean,50.74±35.02versu16.85 ± 6.44(kPa);Emin,42.47±32.33versu14.82 ± 6.06(kPa);Emax,55.69±36.92versu18.64±7.08(kPa);SD,3.76±3.14versu1.16±0.93(kPa);and Eratio,5.72±4.78versu 1.55±0.62(all,p<0.05).Among all the quantitative SWE parameters,the AUC value of Emean was highest,with no statistically significant difference(p>0.05).Compared with pathology results,the kappa value of Emean and Emax was 0.868 and 0.835,respectively.(3)Benign ALNs were presented as color pattern 1,while metastatic ALNs were usually presented as color pattern 2 to 4.There were statistically significant differences in qualitative SWE patterns between metastatic and benign ALNs(p<0.05).The AUC value of qualitative SWE patterns was 0.983,which was higher than that of conventional US and quantitative SWE parameters.The optimal cutoff value of qualitative SWE was Color pattern 2 with 96.7%sensitivity,100%specificity,100%NPV and 96.8%PPV.Compared with pathology,the kappa value of qualitative SWE assessment for axillary lymph node was 0.967.(4)The combination of conventional US and qualitative SWE classification obtained a AUC value of 0.998,which was higher than that of the combination of conventional US and all quantitative parameters,with 98.3%sensitivity,100%specificity,100%NPV and 98.4%PPV..(5)In all,43 texture features were extracted from both breast lesion and surrounding tissue.(6)For conventional US images of breast lesions,model with 9 features yielded the highest AUC of 0.742 in the training set and 0.712 in the validation set.For conventional US images of breast lesions and surrounding tissue,model with 15 features yielded the highest AUC of 0.793 in the training set and 0.684 in the validation set.Conclusions1.The qualitative SWE classification of ALN proposed in our study may be used to differentiate metastatic ALNs from benign ALNs,especially for differential diagnosis of metastatic ALNs and benign reactive ALNs.Thus it could achieve more accurate diagnoses and avoid unnecessary biopsy.2.Texture features extracted from conventional US images of breast cancer are helpful for predicting sentinel lymph nodes metastasis,providing a non-invasive approach to assess sentinel lymph nodes status in clinical practice.