total of 89 patients receiving both DCE-MRI and mammography were analyzed,
including 56 malignant and 33 benign lesions. The 3D tumor mask on MRI was
generated using computer algorithms. A total of 99 texture and histogram
features were extracted from three DCE parameters maps. The suspicious area on
mammography was outlined using MRI findings as guidance, and a similar
radiomics method was applied to extract features from the mass and the margin.
Random forest was applied to select features for building diagnostic models.
The overall accuracy was 0.80 for MRI, 0.75 for mammography, and improved to
0.85 when combined.