Contribution of Radiomics Features from DCE-MRI and DWI in Differentiating Benign from Malignant Lesions in Suspicious Breast (MRI BI-RADS Category 4) Findings

We aimed to find a promising tool to improve the diagnostic efficiency of suspious breast lesions classified in BI-RADS Category 4 from malignant lesions in order to avoid unnecessary biopsy,surgery,even psychological pressure. 33 patients (all female, 27y-82y) were included in our retrospective study and all underwent pre-operative breast DCE-MRI and DWI using a 3.0T MRI (SIEMENS Magntom Verio 3.0T).The radiomics features were acquired by Omni-Kinetics software (GE Healthcare).Non-parametric test and ROC curve were used in statistical analysis.The results implied that the radiomics parameters,especially skewness, kurtosis,IDM and inertia in Ktrans and ADC had great potential.