Abstract
Abstract Purpose: To assess the value of kinetic and architectural diagnostic criteria on dynamic MR breast imaging, and to construct a scoring system for lesion characterization. Material and Methods: A total of 92 women with 109 histopathologically verified breast lesions were included in this retrospective study. The patients were examined by a 1.5 T system using a dedicated double breast coil. A dynamic examination with one precontrast and seven postcontrast series was performed, using a T1-weighted 3D FLASH sequence. Thirty lesions (15 malignant and 15 benign) were randomly chosen for the validation set, and the remaining 79 lesions (62 malignant and 17 benign) formed the estimation set, in which multivariate analysis was performed in order to select the most important features. These parameters were then used for constructing the scoring system, which was tested on the validation set. The scoring system was compared with the routine standard evaluation that used all established diagnostic criteria. ROC curves were generated to assess the diagnostic accuracy of different approaches. Results: In the multivariate analysis of the 79 lesions, time-to-peak enhancement and the descriptor of margins were found to be the most important independent factors for distinguishing benign from malignant lesions, and formed the basis of the scoring system. The areas under the ROC curves for the standard evaluation, and the scoring system were 0.813 and 0.880 in the 30 lesions. Conclusion: Time-to-peak enhancement and the descriptor of margins appear to be the most important diagnostic criteria for mass lesions in dynamic breast MR imaging.