The Value Of Magnetic Resonance Imaging In Differential Diagnosis For Breast Non-mass Enhancement Lesions

Objective:To evaluate the interobserver variability between two radiologists with different experiences for morphological assessment of non-mass enhancement(NME)lesions on the basis of contrast-enhanced MRI(CE-MRI).To assess the validity and reliability of morphological characteristics,diffusion-weighted imaging,dynamic contrast-enhanced magnetic resonance imaging,and radiomic signatures constructed based on CE-MRI images in differentiating between benign and malignant NME lesions.Methods:The breast lesions presented as NME on CE-MRI images between December 2017 and November 2021 and confirmed by pathology from three institutions were enrolled retrospectively.Step 1,405 NME lesions were enrolled and divided into different subgroups according to levels of background parenchymal enhancement(BPE)and the maximal diameters.Cohen’s Kappa coefficient was calculated to compare the agreement of morphological evaluation of the two observers in the overall and different subgroups.Step 2,firstly,184 NME lesions were set as the training cohort to construct morphological diagnostic model using binary logistic regression analysis with pathological results as the reference standard.The apparent diffusion coefficient(ADC)model(ADC+morphology)and the time-intensity curves(TIC)model(TIC+morphology)were then established using binary logistic regression.The models mentioned above were compared for sensitivity,specificity,and area under the curve(AUC)in the training and the validation cohort(77 NME lesions).Step 3,247 NME lesions were set as the training cohort to construct conventional clinical MRI model,radiomics model,and clinical MRI-radiomics combined model using multivariate logistic regression analysis.Radiomic features were extracted from one post-contrast phase(around 90s after contrast injection)of breast dynamic CEMRI(DCE-MRI)images.A validation cohort including 72 NME lesions was used.Step 4,183 malignant and 84 benign NME lesions classified as BI-RADS category 4 initially were enrolled and reclassified according to the clinical MRI-radiomics combined model and diffusion-weighted imaging.The malignancy risk(NS values)converted from nomogram score and the minimal ADC values were calculated.The change in categories was compared using the McNemar test.Finally,the diagnostic performance of the two methods in differentiating benign and malignant BI-RADS category 4 NME lesions,ductal carcinoma in situ and invasive cancer was evaluated.Results:Overall,interobserver agreement in internal enhancement and distribution characteristics was 0.55 and 0.61 respectively.For NME lesions with moderate or marked BPE,interobserver agreements in both internal enhancement and distribution characteristics were lower than NME leisons with minimum or mild BPE.Of these,the morphological assessment in small size subgroup showed the least level of agreement.In the differential diagnosis of benign and malignant lesions,for the TIC/ADC model in the training cohort,sensitivities were 0.924(95%CI,0.921～0.928)/0.814(95%CI,0.809～0.819),specificities were 0.615(95%CI,0.591～0.640)/0.615(95%CI,0.591～0.640),and AUCs were 0.811(95%CI,0.727～0.894)/0.769(95%CI,0.681～0.856),respectively.The AUC of the TICADC combined model was significantly higher than ADC model alone,while comparable with the TIC model(P=0.494).The AUC of TIC model and ADC model was 0.799(95%CI,0.674～0.924)and 0.635(95%CI,0.485～0.785).The combined model,which contained factors including TIC types and radiomics signatures,showed good discrimination,with an acceptable sensitivity of 0.869(95%CI,0.816～0.916),improved specificity of 0.839(95%CI,0.750～0.929).The nomogram was applied to the validation cohort,reaching good discrimination,with a sensitivity of 0.820(95%CI,0.700～0.920),specificity of 0.864(95%CI,0.682～1.000).The NS values showed higher diagnostic efficacy compared with the minimal ADC values.The NS values and minimal ADC values did not achieve high diagnostic accuracy in discriminating between ductal carcinoma in situ(DCIS)and invasive cancer(AUC<0.7).However,the diagnostic performance of the NS-ADC combined model was higher than the NS values alone,while comparable to the minimal ADC values.Conclusions:Interobserver agreement achieved moderate and good level in internal enhancement and distribution characteristics assessment respectively,however,was affected by lesion size and BPE to some extent.Based on the morphologic analyses,the performance of the TIC model was found to be superior than the ADC model for differentiating between benign and malignant NME lesions.Additional radiomics signatures into a conventional clinical model could help to differentiate benign from malignant NMEs and improve specificity without an additional DWI sequence.Compared to DWI,the clinical-radiomics combined model could improve the diagnostic performance in discriminating the BI-RADS 4 NME lesions.However,DWI may play a role in promoting the diagnostic performance in discriminating DCIS from invasive cancer.Part Ⅰ:Diagnostic Performance Assessment and Interobserver Variability for Non-mass Breast Lesions Based on Contrast-enhanced MRI:A Multicenter Study.Objective:To evaluate the interobserver variability between two radiologists with different experiences for morphological assessment of non-mass enhancement(NME)lesions on the basis of contrast-enhanced MRI(CE-MRI).To investigate the effects of background parenchymal enhancement(BPE)and lesion size on the interobserver variability.To calculate and assess the positive predictive value(PPV)of morphological characteristics for malignancy of NME lesions.Methods:This retrospective study enrolled 405 NME lesions confirmed by pathology from three institutions.Lesions were divided into three subroups according to the degree of BPE on CE-MRI images at early phase:minimal-mild subgroup,moderate subgroup,and marked subgroup.Lesions were further divided into three subgroups according to the maximal diameter:small lesions subgroup(≤2cm),medium lesions subgroup(>2cm and ≤5cm),large lesions subgroup(>5cm).Cohen’s Kappa coefficient was calculated to compare the interobserver variability between two radiologists for morphological assessment(including internal enhancement and distribution characteristics)overall and within subgroups.PPVs of morphological characteristics for differentiating benign and malignant lesions were calculated based on the assessment consensus using the histopathology results as the reference standard.Results:Overall,interobserver agreement in internal enhancement and distribution characteristics was 0.55 and 0.61 respectively.For NME lesions with moderate or marked BPE,interobserver agreements in both internal enhancement and distribution characteristics were lower than NME leisons with minimum or mild BPE.Of these,the morphological assessment in small size subgroup showed the least level of agreement(k=0.39 for distribution and k=0.29 for internal enhancement).Internal enhancement presented as heterogenous or clumped and distribution presented as segmental or focal/regional were confusable for the two observers.The combined and pairwise analysis of internal enhancement and distribution characteristics showed the highest PPV was 75.9%when segmental distribution and heterogenous enhancement combined.Conclusions:Interobserver agreement achieved moderate and good level in internal enhancement and distribution characteristics assessment respectively,however,was affected by lesion size and BPE to some extent.It is noted that heterogenous and clumped enhancement,segmental and focal or regional distribution were confusable for the two observers.This suggested that the interpretation and application of BI-RADS descriptors and literature reports should be cautious.Part Ⅱ:Contrasts between diffusion-weighted imaging and dynamic contrast-enhanced MR in diagnosing malignancies of breast non-mass enhancement lesions based on morphologic assessmentObjective:To compare the diagnostic performance of DCE curves and DWI in discriminating benign and malignant NME lesions on the basis of morphologic characteristics assessment on contrast-enhanced(CE)-MRI images.Methods:In this retrospective study,255 patients with 261 NME lesions comfirmed by pathology were divided into training and validation cohorts using random-stratified grouping based on a 7:3 ratio.Of these,180 patients with 184 NME lesions were set as the training cohort to construct morphological diagnostic model using binary logistic regression analysis with pathological results as the reference standard.The apparent diffusion coefficient(ADC)model(ADC+morphology)and the time-intensity curves(TIC)model(TIC+morphology)were then established using binary logistic regression.The models mentioned above were compared for sensitivity,specificity,and area under the curve(AUC)in the training and the validation cohort(75 patients with 77 NME lesions).Results:For the TIC/ADC model in the training cohort,sensitivities were 0.924(95%CI,0.921～0.928)/0.814(95%CI,0.809～0.819),specificities were 0.615(95%CI,0.591～0.640)/0.615(95%CI,0.591～0.640),and AUCs were 0.811(95%CI,0.727～0.894)/0.769(95%CI,0.681～0.856).The AUC of the TIC-ADC combined model was significantly higher than ADC model alone,while comparable with the TIC model(P=0.494).In the validation cohort,the AUCs of TIC/ADC model were 0.799(95%CI,0.674～0.924)/0.635(95%CI,0.485～0.785).Conclusions:Based on the morphologic analyses,the performance of the TIC model was found to be superior than the ADC model for differentiating between benign and malignant NME lesions.Part Ⅲ:Additional Value of Radiomics-Based Signature for Differetiating Benign and Malignant Non-Mass Enhancements on DCEMRIObjective:To assess the additional value of a radiomics-based signature for distinguishing between benign and malignant non-mass enhancement lesions(NMEs)on dynamic contrast-enhanced breast magnetic resonance imaging(breast DCE-MRI).Methods:In this retrospective study,232 patients with 247 histopathologically confirmed NMEs(malignant:191;benign:56)were enrolled from December 2017 to October 2020 as a primary cohort to develop the discriminative models.Radiomic features were extracted from one post-contrast phase(around 90s after contrast injection)of breast DCE-MRI images.The least absolute shrinkage and selection operator(LASSO)regression model was adapted to select features and construct the radiomics-based signature.Based on clinical and routine MR features,radiomics features,and combined information,three discriminative models were built using multivariable logistic regression analyses.In addition,an independent cohort of 72 patients with 72 NMEs(50 malignant and 22 benign)was collected from November 2020 to April 2021 for the validation of the three discriminative models.Finally,the combined model was assessed using nomogram and decision curve analyses.Results:The routine MR model with two selected features of the time-intensity curve(TIC)type and MR-reported axillary lymph node(ALN)status showed a high sensitivity of 0.942(95%CI,0.906～0.974)and low specificity of 0.589(95%CI,0.464～0.714).The radiomics model with six selected features was significantly correlated with malignancy(P<0.001 for both primary and validation cohorts).Finally,the individual combined model,which contained factors including TIC types and radiomics signatures,showed good discrimination,with an acceptable sensitivity of 0.869(95%CI,0.816～0.916),improved specificity of 0.839(95%CI,0.750～0.929).The nomogram was applied to the validation cohort,reaching good discrimination,with a sensitivity of 0.820(95%CI,0.700～0.920),specificity of 0.864(95%CI,0.682～1.000).The combined model was clinically helpful,as demonstrated by decision curve analysis.Conclusions:Our study added radiomics signatures into a conventional clinical model and developed a radiomics nomogram including radiomics signatures and TIC types.This radiomics model could be used to differentiate benign from malignant NMEs on breast MRI.Part Ⅳ:Assessing the Value of Diffusion-weighted Imaging in reclassifying the Malignancy of BI-RADS 4 Non-mass-enhancement Lesions and differentiating ductal carcinoma in situ and invasive cancerObjective:Compared to the clinical-radiomics combined model constructed from Part III,to assess the value of DWI in discriminating BI-RADS 4 non-mass enhancement(NME)lesions,ductal carcinoma in situ(DCIS)and invasive carcinoma.Methods:This retrospective study enrolled 364 NME lesions(343 patients).Of these,183 malignant and 84 benign lesions classified as BI-RADS 4 NMEs by the initial diagnosis were re-classified based on the clinical-radiomics combined model and DWI respectively.The malignancy risk(NS values)converted from nomogram score of the clinical-combined model and the minimal ADC values from DWI were calculated and compared.The percentage of false-positives were estimated in comparison with the original classification.Receiver operating characteristic(ROC)curve analysis was performed to determine the diagnostic value of the NS and minimal ADC values in distinguishing benign and malignant NMEs,DCIS and invasive breast carcinoma.Results:The diagnostic value of the NS values(AUC:0.843;95%CI,0.789～0.896)for discriminating the 267 NME breast lesions categorized as BI-RADS 4 was significantly higher than the minimal ADC values(AUC:0.662;95%CI,0.590～0.735).The NS values showed higher sensitivity,specificity,and accuracy compared with the minimal ADC values(sensitivity:80.3%vs.65.6%;specificity:79.8%vs.65.5%;and accuracy:80.1%vs.65.5%).The NS values and minimal ADC values did not achieve high diagnostic accuracy in discriminating between ductal carcinoma in situ(DCIS)and invasive cancer(AUC<0.7).However,the diagnostic performance of the NS-ADC combined model(AUC,0.731;95%CI,0.655～0.806)was higher than the NS values alone,while comparable to the minimal ADC values.Conclusions:The clinical-radiomics combined model from the CE-MRI could improve the diagnostic performance in discriminating the BI-RADS 4 NME lesions without an additional DWI sequence.However,DWI may play a role in promoting the diagnostic performance in discriminating DCIS from invasive cancer.