Objective: To assess the effect of changing the region of interest (ROI) on early enhancement (EE) in dynamic breast magnetic resonance (MR) imaging.
Methods: We evaluated retrospectively 102 breast lesions: 54 lesions (33 malignancies and 21 benignancies) studied with 2D and 48 lesions (30 and 18, respectively) with 3D gradient-echo dynamic technique (contrast dose 0.1 mmol/kg). Each lesion was postprocessed using 3 different regions of interest (ROIs): small circular ROI on maximal enhancement (SCR); large circular ROI within the lesion (LCR); and irregular ROI by manual contouring (IRR). EE was classified as benign (< or = 50%), uncertain (51-89%), or malignant (> or = 90%).
Results: With 2D, the uncertain EEs were 17% for both SCR and LCR, 33% for IRR (P = 0.008); with 3D, the uncertain EEs were 4%, 15%, and 13%, respectively (SCR versus LCR, P = 0.063). More uncertain EEs were obtained with 2D (17-33%) than with 3D (4-15%), significantly for SCR (P = 0.043) and IRR (P = 0.013). Considering uncertain EEs as positive, sensitivity was 100% for SCR, 91% for LCR, and 82% for IRR (SCR versus IRR, P = 0.031) with 2D, 100%, 97%, and 87%, respectively, with 3D technique, without significant differences; specificity ranged from 71% to 90% with 2D and 61% to 83% with 3D, without significant differences.
Conclusion: The type of ROI influences the EE in dynamic breast MR. Using 3D technique with small ROI located on the area of maximal enhancement gives the best results in terms of certainty of the level of EE together with top levels of sensitivity.