Neha Bhooshan1, Maryellen Giger1, Li Lan1, Angelica Marquez2, Hui Li1, Gillian Newstead1
1University of Chicago, Chicago, IL, United States; 2Loyola University, Chicago, IL, United States
This studys purpose was to investigate the automated analysis of T2-weighted MR images in distinguishing malignant and benign breast lesions. Using 86 benign and 110 malignant lesions, our CADx scheme automatically performed lesion segmentation, feature extraction, and classification. T2 morphological features yielded an AUC of 0.78 0.03 while T1 kinetic and morphological features achieved an AUC of 0.83 0.03. When considering all features, two T2 features, three T1 features and one geometric feature were selected, giving an AUC of 0.85 0.03. T2 MRI has the potential to improve the performance of CADx in distinguishing malignant and benign breast lesions.