Table 1.
Patient and lesion histopathology characteristics.
Fig 1.
Pipeline for extracting the proposed T2w CAD features.
a) Steps involved in measuring Lesion-to-muscle signal intensity ratio (LMSIR). b) steps involved to predict LMSIR from T2w features.
Fig 2.
Bland-Altman plot of the agreement between measured and predicted LMSIR as a CAD feature of relative T2w lesion intensity.
Fig 3.
Receiver operating characteristic (ROC) and Area Under the Curves (AUC)—95% confidence-interval (CI) for each feature space.
Cross marks indicate optimal operating point (OP) and 95% CI of Sensitivity and Specificity.
Table 2.
CAD performance according to feature groups.
Sensitivity and Specificity are reported at the optimal operating point of group 1 in ROC space.
Table 3.
Bonferroni corrected p-values of AUC performance differences across groups.
Fig 4.
Invasive ductal carcinoma in a 36-year old high risk woman appeared as a focal non-mass, with persistent enhancement (BI-RADS score 4).
a) T1w 4th post-contrast DCE-MRI. b) T2w MIR. c) zoomed progressive enhancement at each of the four post-contrast scans. d) zoomed lesion T2w appearance.
Table 4.
Explanation of CAD classification output for lesion in Fig 4.
Fig 5.
A benign fibroepithelial lesion with fibrocystic changes and hyperplasia in a 35-year old high risk woman. The lesion appeared as a circumscribed round mass, with smooth margins, rapid early enhancement and mild washout in delayed enhancement (BI-RADS score 4).
a) T1w 4th post-contrast DCE-MRI. b) T2w MIR. c) zoomed progressive enhancement at each of the four post-contrast scans. d) zoomed lesion T2w appearance.
Table 5.
Explanation of CAD classification output for lesion in Fig 5.