Abstract: Background and purpose: The technique of dynamic contrast-enhanced MRI (DCE-MRI) is widely applied in differential diagnosis between benign and malignant tumor and therapeutic estimation of neoadjuvant chemotherapy in clinic. However, there is no standard quantitative measurement method. This study aimed to assess the variability of different region of interest (ROI) selections for tumor bed of breast cancer using DCE-MRI, and to ascertain the optimal ROI delineation. Methods: We retrospectively analyzed DCE-MRI of 30 patients diagnosed with breast cancer by pathology. The ROIs were delineated by 2 different observers using iCAD software with 4 methods, including whole tumor (Whole), the slice containing the most enhancing voxels (SliceMax), 3 slices centered in SliceMax (Partial) and the 5% most enhancing contiguous voxels within SliceMax (5Max), to generate the volume transfer constant (K^trans), the extracellular volume fraction (V_e) and rate constant (K_ep). And the reproducibilities of the measurements were assessed using the Bland-Altman method. Results: In the analysis of ROIs delineation, the K^trans, V_e and K_ep reported by different observers were 1.26±0.54 vs 1.25±0.53, 0.75±0.23 vs 0.73±0.22 and 1.93±1.46 vs 1.95±1.51 (P>0.05) using the method of Whole, and 1.28±0.43 vs 1.26±0.43, 0.74±0.21 vs 0.80±0.27, 1.95±1.53 vs 1.93±1.43 (P>0.05) using the method of Partial, and 1.30±0.33 vs 1.32±0.33, 0.77±0.20 vs 0.73±0.24, 1.82±1.53 vs 1.87±1.45 (P>0.05) using the method of SliceMax, and 1.31±0.35 vs 1.35±0.33, 0.77±0.20 vs 0.98±0.25, 1.97±1.36 vs 1.73±1.55 using the method of 5Max (P<0.05). Using the methods of ROI delineation except 5Max, there was no significant difference between K^trans, V_e and K_ep reported by different observers. The bias vs limits of agreement were 0.002 vs -0.013 to 0.012, -0.003 vs -0.023 to 0.017, 0.006 vs -0.018 to 0.029, -0.035 vs -0.054 to 0.018 measured with Whole method, SliceMax, Partial and 5Max respectively using the Bland-Altman method. Conclusion: It may be reliable to measure functional parameters of primary tumors in breast cancer using DCE-MRI according to the methods of Whole, Partial and SliceMax.

Key words: Breast cancer, Tumor bed, Dynamic contrast-enhanced MRI, Region of interest