The Comparison Of Multimodal Imaging Techniques In Diagnosis Of Breast Tomosynthesis BI-RADS 4A Lesions

Part One Retrospective analysis of benign and malignant lesions ofbreast tomosynthesis BI-RADS 4A and study on thediagnostic value of ultrasonographyObjective: Breast Imaging Reporting and Data System(BI-RADS)4A represents a type of lesion with a low likelihood of malignancy but requiring intervention(malignancy probability ranging from 2% to 10%).This study was to explore the characteristics of benign and malignant lesions through retrospective analysis of the comparison between digital breast tomosynthesis(DBT)BI-RADS 4A lesions and pathological results,and to explore the diagnostic value of ultrasonography(US)for DBT BI-RADS 4A lesions.Methods: A total of 1857 cases with breast lesions diagnosed and treated in our hospital from April 2019 to December 2021 were enrolled.All of these patients underwent both DBT and US preoperative.All DBT images were described and evaluated by radiologists with more than 20 years of experience in breast imaging.US was performed by a sonographer with more than 3 years of experience in breast US.If suspicious lesions were found,they would be described and diagnosed according to BI-RADS.Chi-square test was used to compare the characteristics between benign and malignant lesions.A receiver operating characteristic(ROC)curve analysis was used to evaluate the diagnostic performance of US with the biopsy or surgical pathological results as the gold standard.In addition,the sensitivity,specificity,positive predictive value(PPV)and negative predictive value(NPV)were calculated on the basis of BI-RADS categories,at cut-off values of > 3,which indicated malignancy.The comparisons of sensitivity and specificity were performed by using Mc Nemar test.Results: A total of 1857 lesions were included in this study,of which 161 were malignant,invasive ductal carcinoma was the most common,followed by ductal carcinoma in situ.In 1696 benign lesions,adenosis was the most common,followed by fibroadenoma.The PPV of DBT BI-RADS 4A was8.7%(161/1857).Of the 1696 benign lesions,1302 were masses,218 were microcalcifications,90 were asymmetries,and 86 were architectural distortions.Of the 161 malignant lesions,77 were masses,39 were microcalcifications,33 were asymmetries,and 12 were architectural distortions.There were significant statistical differences in shape and margin between benign and malignant masses(both P<0.001),but there was no statistical difference in density and size(P=0.110;P=0.069).There was significant statistical difference between benign and malignant microcalcifications in the morphology of microcalcification(P<0.001),but there was no statistical difference in distribution(P=0.153).The area under the ROC curves(AUC)of US in diagnosis DBT BI-RADS 4A lesions was0.591(95% CI: 0.548-0.635),and the sensitivity,specificity,PPV and NPV of diagnosis were 82.0%,33.0%,10.4% and 95.1%,respectively.The sensitivity,specificity and PPV of DBT were 100.0%,0.0% and 8.7% respectively.The sensitivity and specificity of us and DBT were statistically significant(both P< 0.001).Conclusion: DBT BI-RADS 4A lesions are mainly benign lesions,and also include a small number of malignant lesions with atypical malignant signs.The shape,margin of mass and morphology of microcalcifications are helpful to differentiate benign from malignant lesions.The diagnostic efficacy of US for DBT BI-RADS 4A lesions is low.Therefore,it is necessary to explore new technologies for further research.Part Two Diagnostic value of contrast-enhanced mammography in thedifferential diagnosis of benign and malignant BI-RADS 4AlesionObjectives: Contrast-enhanced mammography(CEM)is a new promising technique developed in recent years that can evaluate the morphology and angiogenesis associated with breast cancer.This study was to investigate the diagnostic performance of CEM combined with BI-RADS and the Kaiser score in digital breast tomosynthesis(DBT)BI-RADS 4A lesions to potentially reduce unnecessary breast biopsies.Methods: This retrospective study evaluated 106 patients with 109 DBT BI-RADS 4A lesions from June 2019 to June 2021.All CEM cases were independently evaluated by two radiologists with different diagnostic experiences.Two readers were asked to classify each lesion according to the CEM BI-RADS.For the absence of enhancement on CEM,the lesions were downgraded to BI-RADS 3.For lesions with enhancement,the readers were asked to classify all enhancing lesions referring to the Kaiser score for breast MRI.Receiver operating characteristic(ROC)curve analysis was used to evaluate the diagnostic performance.Two readers rated all cases and inter-reader agreement was assessed by Cohen’s kappa coefficients.Results: A total of 109 BI-RADS category 4A lesions classified by DBT in 106 women were included in the study.Pathologically,ninety-five lesions(87.2%)were benign and 14(12.8%)lesions were malignant.Of the 109BI-RADS category 4A lesions,20 showed no enhancement,68 lesions presented as enhancing masses,and 21 presented as non-mass enhancement.The AUC of CEM to distinguish malignant from benign lesions ranged between 0.770(95% CI :0.670-0.871)and 0.805(95% CI : 0.715-0.894)with BI-RADS and 0.880(95% CI : 0.816-0.944)and 0.906(95% CI :0.850-0.961)with Kaiser score.The AUC of Kaiser score reading was higher than BI-RADS for two readers(P<0.001).CEM with Kaiser score(readers 1and 2)had higher specificity than CEM with BI-RADS(P<0.001).CEM combined with Kaiser score could have potentially obviated 75.8% to 82.1%unnecessary benign biopsies.The inter-reader agreement was substantial(kappa: 0.727)for Kaiser score.Conclusions: Both the CEM BI-RADS and Kaiser score may be used to distinguish DBT BI-RADS 4A lesions,but diagnostic efficiency of Kaiser score is superior to BI-RADS,reducing unnecessary biopsy in benign lesions.Part Three Diagnostic value of contrast-enhanced mammography andMRI for BI-RADS 4A lesionsObjectives: Contrast-enhanced mammography(CEM)has the same diagnostic efficacy as MRI in the diagnosis of breast cancer,and even higher specificity.However,the diagnostic value of CEM and MRI in digital breast tomosynthesis(DBT)BI-RADS 4A lesions is currently unclear.This study aimed to investigate the diagnostic performance of CEM and MRI in DBT BI-RADS 4A lesions.Methods: This retrospective study evaluated 42 patients who were diagnosed as BI-RADS 4A after DBT examination in our hospital from June2019 to June 2021,and had both CEM and enhanced MRI examinations before puncture or operation.The CEM and MRI images were independently evaluated by two experienced radiologists without knowing the pathological results.They reached an agreement in case of disagreement.BI-RADS 1-3was considered benign,while BI-RADS 4-5 was considered malignant.The diagnostic efficacy was evaluated using ROC curve analysis.Cohen’s kappa was used to statistically analyze the consistency of CEM and MRI enhancement curve and Kaiser score.Results: Among 42 cases,12 cases were malignant(28.6%,12/42)and30 cases were benign(71.4%,30/42).Thirty-five cases had enhancement in both CEM and MRI,and the coincidence rate of enhancement curve between CEM and MRI was 82.9%.Cohen’s Kappa coefficient of enhancement curves of CEM and MRI was 0.743.The AUC of CEM and MRI according to BI-RADS were 0.763(95% CI: 0.606-0.880)and 0.724(95% CI: 0.564-0.850),respectively,with no significant difference between the two(P=0.523).The sensitivity and specificity of CEM and were 83.3% and 40.0%,91.7%and 20.0% respectively.There was no significant difference in sensitivity and specificity between CEM and MRI(P=1.000,P=0.077).The AUC of CEM and MRI combined with Kaiser score to differentiate benign and malignant breast lesions were 0.804(95% CI: 0.635-0.918)and 0.770(95% CI:0.597-0.895),respectively,and there was no significant difference between the two(P=0.609).The agreement of Kaiser score by CEM and MRI was substantial(kappa: 0.717).Conclusions: Both CEM and MRI have relatively high diagnostic value in DBT BI-RADS 4A lesions,and there is no significant difference in the diagnostic efficacy between these two methods in our study.CEM is an effective alternative when MRI is not applicable or there are contraindications.Part Four Diagnostic value of radiomics based on contrast-enhancedmammography for BI-RADS 4A lesionsObjectives: Multiple studies have shown that radiomics based on contrast-enhanced mammography(CEM)exhibits high accuracy in the differentiation of benign and malignant breast lesions,but there have been no relevant reports on the diagnostic value of digital breast tomosynthesis(DBT)BI-RADS 4A lesions.Therefore,this study aimed to investigate the diagnostic performance of radiomics based on CEM in predicting DBT BI-RADS 4A lesions to potentially reduce unnecessary breast biopsies.Methods: This retrospective study evaluated 99 patients diagnosed with DBT BI-RADS 4A lesions from June 2019 to June 2022.All CEM cases were evaluated by two radiologists with different diagnostic experience according to the CEM BI-RADS,and a BI-RADS classification was given.Patients were randomly divided into training set(69 cases)and test set(30 cases)in a 7:3ratio.The ROI of all cases were manually segmented,and the radiomics features were extracted from the original image and preprocessing image.The radiomics features included three categories of first order features,morphological features,and texture features.F test was used for feature screening to construct the radiomics signature,and the radiomics signature score(Rad-score)was calculated for each patient.The diagnostic performance of the radiomics model,the BI-RADS classification and the combined model for BI-RADS 4A lesions was evaluated using the receiver operating characteristic(ROC)curve and the area under curve(AUC).The differences in AUC were compared using the De Long test.Results: In both training and test sets,the Rad-score of patients with benign lesions was higher than that of patients with malignant lesions(both P<0.05).The AUC of the radiomics model in the training set and the test set based on CC views,MLO views,and combined views were 0.935(95%CI:0.854-0.982)、0.823(95%CI: 0.725-0.913)、0.947(95%CI: 0.864-0.987)and 0.812(95%CI: 0.629-0.931)、0.716(95%CI: 0.523-0.865)、0.937(95%CI: 0.785-0.993),respectively.In the training set,the AUC of the radiomics model based on the combined views was higher than that of the MLO(P=0.042),and the remaining differences were not statistically significant(all P>0.05).The AUC values of the radiomics model,the BI-RADS classification and the combined model for predicting breast BI-RADS 4A lesions in the training set and the test set were 0.947(95% CI:0.864-0.987),0.843(95% CI: 0.736-0.920),0.995(95% CI: 0.939-1.000),and0.937(95% CI: 0.785-0.993),0.821(95% CI: 0.638-0.936),and 0.994(95%CI: 0.874-1.000),respectively.The AUC of the combined model in both the training set(P=0.002)and the test set(P=0.046)was higher than that of the BI-RADS classification;there was no significant difference compared to the radiomics model(P=0.053,P=0.140).There was no statistically significant difference between the AUC values of the radiomics model and the BI-RADS classification in the training set(P=0.088)and the test set(P=0.249)Conclusions: The radiomics model has high diagnostic value in differential diagnosis of benign and malignant BI-RADS 4A lesions.The radiomics model based on the combination of CC and MLO views has a higher AUC than that of the MLO view in the training set.Compared with BI-RADS classification,the combined model of BI-RADS classification and the radiomics model has a higher diagnostic efficiency.