Preoperative Diagnostic Performance Of Dynamic Contrast-enhanced Magnetic Resonance Imaging Combined With Diffusion-weighted Imaging In Breast Lesions

Objective Joint analysis of clinical, radiological and pathological featuresof missed, false-negative and false-positive lesions was performed to explorethe causes of misdiagnosis.Materials and methods A consecutive cohort of175patients whounderwent preoperative breast MRI with222pathologically confirmedlesions were eligible for investigation. Two breast imaging radiologistsperformed the MRI interpretation according to BI-RADS atlas. Comparisonof MRI diagnostic performance between different lesion types was takenplace and diagnostic value of individual radiographic feature was explored.Logistic regression models were also established to differentiate benign andmalignant lesions.Results Of all222beast lesions (161benign and61malignant),6benignlesions were missed on MRI.2DCIS and2invasive carcinoma werefalse-negative.14lesions were overestimated. The sensitivity, specificity andaccuracy of DCE-MRI combined with DWI were93.4%(57/61),91.3%(147/161) and91.9%(204/222). The overall accuracy of mass and NMLElesions was similar, while there were more features identifiable formalignancy in mass than NMLE lesions. ADC value worked great for masses with AUC of0.932, but poor for NMLE lesions with AUC of0.591.Differences between false-positive and true-positive lesions were limited,while false-positive and true-positive lesions differed significantly. Benignmasses with irregular or speculated margin, rapid initial rise, wash-out TIC orreduced ADC value tended to be overestimated as malignancy.Underestimated lesions were mainly DCIS.Conclusion Accuracy of DCE-MRI combined with DWI was high indifferentiating benign and malignant lesions, and similar in masses andNMLE lesions. Features including irregular or speculated margin, rapidinitial rise, wash-out TIC or reduced ADC value of masses were powerful fordifferentiation of malignancy and benignity, and potential cause offalse-positive diagnosis. Underestimated lesions were mainly DCIS.Radiologists tended to overestimate lesions and avoided underestimation atthe same time, while the overall accuracy was almost unaffected. Objective To study the MRI features of breast phyllodes tumor and tocorrelate them to different pathological types.Materials and methods clinical and imaging findings of16patients withpathologically confirmed PT were retrospectively reviewed. All of these16patients had pre-operative MRI, which were interpreted by2experiencedradiologists according to BI-RADS classification. Inter-group comparisonswere performed between benign and malignant PT based on pathologicalfindings.Results The pathologic findings were benign, borderline, and malignant in10,4, and2cases. Both2malignant cases had previous breast tumor excision.PT showed isointensity on non-contrast enhanced T1WI and hyperintensityon STIR, while large lesions showed heterogenous internal signal intensity.Non-benign PT tended to be featured with larger diameter, confluentmultinodular, type III TIC and lower ADC. MRI detected100%PT lesionswith a diagnostic accuracy of62.5%. BI-RADS classification results showedlow correlation with pathological results.Conclusion MRI features of breast PT are diverse. Breast MRI has certainadvantages in preoperative diagnosis of PT, while the final diagnosis shouldonly be based upon carefully histopathological examination of the wholelesion.