In breast tumor characterization, DCE-MRI is commonly used for understanding the vasculature changes associated with neo-angiogenesis [1]. For DCE-MRI, pre-contrast T₁-relaxation mapping is necessary for converting signal to contrast concentration for pharmaco-kinetic (PK) modeling. However, contrast uptake primarily affects T₁ relaxation rate of water protons only, and not fat protons [2]. Hence, in breast tissue with a high density of fat and water mixed voxels, generating water-only T₁ relaxation maps is necessary for accurate PK modeling [3,4,5]. In this study, we evaluated T₁ relaxation maps obtained using traditional non-fat-suppressed variable flip angle (VFA) data with those obtained from Dixon based fat-water separation VFA method. Results are presented in fat, fibro-glandular tissue (FGT) and breast tumors.Patient database: Six breast cancer patients were scanned on a GE 3T MR750 scanner using an 8-channel breast coil. An appropriate IRB approved all the studies. Imaging: T₁ mapping consisted of two different acquisitions: a. Conventional 3D SPGR non-fat suppressed variable flip angle (T1-VFA) method: TE/TR = 2.1/5.28 ms, five FA= (2, 3, 5, 10, 15)º, matrix size=256×256×112 (1.36mm×1.36mm×1.4mm resolution), axial orientation; b. VIBRANT-FLEX-VFA : Dixon based method to generate fat-only and water-only images with TE/TR = 1.2, 2.3 / 5.3 ms, and flip angles and geometry as described in a above. Fat-only images were available in four cases, while water-only and conventional (T1-VFA) images were available in all six patients. c. Bloch-Siegert [6] based B₁ map acquisition using a body receive coil with 2D-GRE and TE/TR = 13.5/30ms, FA = 20º, matrix size=128×128×22 (2.73mm×2.73mm×7mm resolution). Fat fraction was computed using FA = 2º fat and water images. T₁ mapping: An in-house tool developed within the Insight Toolkit (ITK) framework was used for VFA based T₁ mapping [7]. Bloch-Siegert based B₁ data was processed to obtain spatially varying scaling factor for FA correction. No significant geometrical distortions were observed between B₁ data and VFA data and hence only an identity transform was used for geometric matching of B₁ map to T₁ data. The flip angle corrected VFA data were processed to obtain T₁ map using: linear fitting and non-linear Levenberg–Marquardt fitting. To assess goodness-of-fit for T₁ mapping, coefficient of determination (R²) was computed. Analysis: A trained radiologist marked representative ROIs in both left and right breast for fat (on fat-only, 2º,N=4), FGT (water-only, 15º, N=6) and tumor lesion (water-only, 15º, N=2). Only those voxels with R² > 0.5 for T₁ model fit were retained.Figure 1 demonstrates well-known variation of transmit B₁ in breast imaging at 3T (left breast scaling ~= 1.2, right breast scaling ~= 0.8) [8]. Fat T₁ values from fat-only images are lower (337±50 ms) compared to those with T1-VFA (442±35 ms) (Figure 2) and match with literature values [4,9]. For FGT, T,₁ values were elevated with water–only images (1489±265 ms), compared to T1-VFA images (1238±138 ms) and similar to previous literature values for FGT (1444ms) [3,9]. R² values for water-only images were lower (~0.82±0.08), compared to T1-VFA based images (~0.9±0.06) (Figure 3). For tumor lesions (Figure 4, A and B), in one case where fitting accuracy was higher (R² = 0.82), there was a 10% increase in T₁ with water-only (2780 ms) compared to T1-VFA (2523 ms). However, in other tumor case, with relatively poor fitting (R² = 0.77), we noticed reduction (26%) in the T₁ value with water-only (1638 ms), compared to T₁-VFA (2216 ms). Further investigation indicated that motion across FA volumes was responsible for this apparent reduction in T₁ values (Figure 4C). Lesion mask was drawn on FA =15º water-only image, while other FA images had moved from this reference. As a result, the fat-voxels (fat fraction in lesion = 0.29±0.3) from surrounding region (patient had very fatty breast) were counted in tumor ROI and reduced the tumor T₁ (See Fig 4C). Matching FA = 15º to FA = 2º using dense registration and warping tumor mask accordingly increased T₁ value for tumor to 2038 ms. With water-only images and in FGT and tumor regions, non-linear fitting produced slightly elevated T₁ values compared to linear fitting (Fig 5A), as well as marginally higher R² with non-linear (0.8) compared to linear fitting (0.78, p < 0.01) (Fig 5B). Bland-Altman analysis indicated a mean bias of 17 ms (limits of agreement = +84 to -118 ms) (Fig 5C).We demonstrate that T₁ mapping with Dixon based VFA method and non-linear fitting recovers the T₁ values for tissue in breast by reducing partial volume effects. The Dixon water only T₁ mapping will improve the accuracy of PK modeling in breast tumors.