Analysis Of The Diagnostic Efficacy Of Digital Breast Tomosynthesis And First-order Texture Features For Breast Mass-like Lesions

Objective:To compare and analyze the detection ability and accuracy of full-field digital mammography(DBT)and digital breast tomography(DBT)in the detection of breast mass lesions,to study the diagnostic efficacy of DBT in the diagnosis of breast mass-like lesions,and to explore the diagnostic value of DBT radiomics in the evaluation of breast mass-like benign and malignant diseases.Methods:Clinical data of patients with breast lump-like lesions who received FFDM and DBT examinations in the First Affiliated Hospital of Xinjiang Medical University from September 2021 to December 2022 were collected.Two diagnostic radiologists of breast imaging with different years of experience were invited to analyze the image characteristics of lump-like lesions in mammography and tomographic images,respectively.The classification and diagnosis of BI-RADS were performed according to the ACR2013 guidelines.The two groups were divided according to the positive criteria of BI-RADS 4B and above indicating malignancy,and the negative criteria of BI-RADS 4A and below indicating benign.Finally,pathological diagnosis was used as the gold standard to compare the difference in the detection rate and correct diagnosis rate of lumps by Chi-square test.P<0.05 was considered statistically significant.Receiver operating characteristic curve(ROC)was used to evaluate diagnostic efficacy.In addition,the DBT images obtained were manually delineated in CC and MLO positions the region of interest of the lump-like lesions interest,ROI),the features of the outlined images were extracted using pyradiomics software package,the feature dimension was reduced by Logistic algorithm,and meaningful image omics features were screened.The performance of the model was evaluated using five quantitative indicators,including accuracy,area under the curve,sensitivity,specificity and truncation value.Results:In this study,44 patients with a total of 50 lesions were collected,ranging in age from 35 to 75years old,with an average of 48.5±8.89 years old.All patients had complete pathological data,including 16 malignant lesions and 34 benign lesions.Compared with the detection rate and correct diagnosis rate of DBT in lumps,DBT could improve the detection rate and correct diagnosis rate of lumps,and the detection rate of benign lesions was more significant.Compared with the three examination methods,the difference was statistically significant(x~2=6.341,P<0.05).For the detection rate of breast lesions in dense and non-dense glands,DBT,FFDM and DBT combined with FFDM had statistical significance in the detection of mass in dense gonads(x~2=6.318,P<0.05).For the comparison of sensitivity and specificity in the diagnosis of benign and malignant lesions,the highest sensitivity and specificity of DM+DBT were 93.75%and 97.06%,respectively.The area under ROC curve of DM+DBT,DBT and DM in the diagnosis of benign and malignant lesions were 0.954,0.847 and 0.697,respectively.The differences among the three were statistically significant.For the screening of the imaging features,the feature models at both the MLO and CC levels had the ability to distinguish benign and malignant lesions(AUC>0.65).Conclusion:The detection rate and coincidence rate of breast lump-like lesions of DBT is better than FFDM,and the accuracy of DBT for the display of the tumor margin and the classification and diagnosis of BI-RADS is better than FFDM.In addition,the imaging features based on DBT have certain significance for the differentiation of benign and malignant breast lesions.