| Objectives:1 To acquire characteristic labels based on breast MRI and establish an radiomic prediction model for differential diagnosis of benign and malignant breast lesions,evaluate their diagnostic efficiency and optimize the best prediction model.2 To analyze Traditional Chinese Medicine(TCM)theory deeply by using MRI and texture features,explore the etiology and pathogenesis of breast cancer,explore the texture feature labels related to TCM syndrome differentiation of breast cancer,and provide new ideas for the modernization research of TCM.Materials and methods:1 587 patients with breast lesions(including 404 cases of breast cancer and 183 cases of benign lesions)diagnosed by biopsy or pathology from September 2014 to may 2017 were retrospectively collected and divided into training group and test group according to the ratio of 7:3.After manual delineation of the maximum ROI of the lesions,63 quantitative texture features reflecting the shape,size,intensity and texture of the tumor were automatically extracted from the plain and enhanced images respectively.Then,the training group data were reduced in dimension,and the predictive models based on plain and enhanced MRI scanning were constructed by logistic regression analysis method.The test group data were substituted into the models for testing,and the diagnostic accuracy of the two models were obtained respectively.ROC was used to evaluate the diagnostic efficiency of the two models.Rad score of the two models was superimposed by generalized linear regression method,and the weights of the two models were compared and evaluated in the form of visual nomogram.2 This study retrospectively collected 404 cases of breast cancer diagnosed by pathology from September 2014 to May 2017 in the Cancer Hospital of China Medical University.Based on the theory of TCM,all cases were grouped into the following two forms:(1)Liver Meridian of Foot-Jueyin and Stomach Meridian of Foot-Yangming;(2)The Kidney Meridian of Foot-Shaoyin,Stomach Meridian of Foot-Yangming and Spleen Meridian of Foot-Taiyin.During the grouping process,the location of breast cancer lesions indicated by sagittal T2WI and transverse DCE-MRI were dynamically combined for meridian division.The clinical data,MRI signs and texture features of patients in each group were collected,including tumor size,location,tumor-like enhancement shape,tumor-like enhancement edge,tumor-like enhance-ment type,TIC,molecular typing,Ktrans maximum,Ktrans average,Kep maximum,Kep average,Ve maximum,Ve average,and texture feature parameters Cluster shade,Cluster promise,Long run low grey level emphasis,and Uniformity;(3)According to the classical theory of zang-fu organs of TCM,"liver is born in the left and lung is hidden in the right"for grouping analysis.Finally,to analyze whether there are statistical differences in clinical data,MRI signs and texture features of breast cancer of different types.Results:1 Based on the texture feature analysis of MR plain images,the characteristics were obtained as follows:Variance、Quantile 5、Correlation、Inverse Difference Moment、Haralick correlation and Long run low grey level emphasis.The accurate rate of diagnosis of benign and malignant lesions was 71.2%,the best diagnostic threshold was 0.658,the sensitivity was0.740,the specificity was 0.636,and ROC was 0.742.Based on the texture feature analysis of MR enhanced sequence images,the feature parameters with the significance of distinguishing as follows:Unity,Cluster Shade,Cluster Promise and Long run low grey level emphasis.The accurate rate was 83.1%,the best diagnostic threshold was 0.688,the sensitivity was 0.820,the specificity was 0.855,and ROC was 0.887.2 This study shows that there is no statistical difference between the Liver Meridian of Foot-Jueyin and the Stomach Meridian of Foot-Yangming in terms of breast MRI,perfusion parameters and texture features.In the grouping of the Spleen Meridian of Foot-Taiyin,Kidney Meridian of Foot-Shaoyin and Stomach Meridian of Foot-Yangming,there were no statistical differences in the imaging manifestations of breast cancer lesions,such as tumor size,enhanced morphology,type and perfusion parameters.Based on the analysis of MRI texture characteristic parameters,there were statistical differences between the three groups:Cluster Shad and Cluster Promise(p=0.05,p=0.025),among which cluster shad was higher in Spleen Meridian of Foot-Taiyin and Stomach Meridian of Foot-Yangming,cluster promise was higher in Stomach Meridian of Foot-Yangming.3 The DCE-MRI perfusion parameters Ktrans(maximum,average),Kep(maximum,average)and texture feature Quantile 5 of the left and right lesions of the body are different.The perfusion parameters of the left lesion are higher than that of the right lesion,indicating that malignant tumors have heterogeneity in microscopic histopathology.Conclusions:1 Texture analysis of MRI can effectively distinguish between benign and malignant breast lesions.The diagnostic efficiency of the radiomic model constructed with enhanced MRI is better than the flat scan model.Texture analysis based on enhanced MRI is expected to be a basic tool for clinical diagnosis.2 This study was guided by the theory of TCM and analyzed with the parameters of MRI texture:(1)The occurrence and transmission of breast cancer cannot completely separate the liver and stomach from each other.(2)The difference of texture characteristic parameters cluster shade and cluster promise in the three groups of meridian classification is statistically significant.The texture parameters of the lesions of the Stomach Meridian of Foot-Yangming and Spleen Meridian of Foot-Taiyin are larger in deviation and higher in heterogeneity.(3) There are some differences in the texture characteristics of DCE-MRI perfusion parameters between the left and right lesions in the body.The perfusion parameters of left lesions are higher than that of right lesions,suggesting that the microvascular permeability or blood flow perfusion of left lesions may be higher than that of right lesions.3 MRI can provide diagnostic information from the aspects of morphological structure and functional changes.Radiomics can combine macroscopic images with microscopic information to provide objective basis for TCM syndrome differentiation and treatment,which is conducive to promoting the research process of TCM modernization. |