Dynamic contrast enhanced MRI (DCE-MRI) signal depends, among other factors, on tissue relaxation time (T₁) and transmit RF (B₁) homogeneity. In standard DCE processing, T₁ and B₁ are assumed to be constant and DCE pharmaco-kinetic (PK) parameters derived as such have demonstrated good correlation to cancer therapy response [1]. However, with breast-MRI, significant B₁ variations are typically observed [2]. Moreover, recent work in prostate-DCE has demonstrated that use of T₁ mapping with B₁ correction can improve lesion conspicuity [3]. In this work, we investigated impact of incorporating T₁ and/or B₁ maps on PK parameters in breast cancer patients and effect of these PK maps on assessing therapy response in longitudinal data.

Patient database: Six breast cancer patients were scanned on a GE 3T MR750 scanner using an 8-channel Breast coil. Of these, two patients (#P1: complete responder, #P2: partial responder) were scanned at 3 time-points of therapy cycle. An IRB approved all studies. Imaging: a. VIBRANT-TRICKS with TE/TR = 3.7/7ms, FA=12º, 0.68mm×0.68mm×1.4mm resolution, 48 bolus phases with 9.3s resolution; b. Variable Flip Angle (VFA) based T₁ map estimation using 3D SPGR: TE/TR=2.1/5.3ms, 1.36mm×1.36mm×1.4mm resolution and five FAs (2º, 3º, 5º, 10º and 15º). c. Bloch-Siegert [4] based B₁ map acquisition using body receive coil with 2D-GRE and TE/TR = 13.5/30ms, FA = 20º, 2.73mm×2.73mm×7mm resolution. DCE Motion Correction: Performed on all cases as described in [5]. DCE data analysis: In-house tool developed in Insight Toolkit (ITK) was adapted for PK analysis [1]. Bloch Siegert based B₁ data was processed to obtain spatially varying scale factor for FA correction. For VFA data, FA was corrected at each voxel using B₁ scale maps. Corrected VFA data was processed to obtain T₁ map. No significant geometrical distortions were observed between B₁ map, VFA data and DCE data and hence only an identity transform was used for interpolation of B₁/T₁ data to DCE data. T₁ map and B₁ FA scaling map were used to correct signal to concentration mapping of DCE data [3]. A two-parameter Toft’s model was fitted to DCE concentration data using a population based AIF [6] to obtain K_trans and ve estimates.

DCE Processing schemes: We considered four different processing schemes: Proc#1: DCE only with T₁ fixed (blood = 1600 ms, tissue = 1444 ms) and considered as typical processing [1]. Proc#2: DCE with T₁ map only (no B₁ map applied to T₁ as well). Proc#3: DCE with B₁ map and T₁ fixed (blood = 1600 ms, tissue = 1444 ms). Proc#4: DCE using B₁ map and T₁ map (exactly as described above in DCE data analysis). PK map analysis: A trained radiologist marked breast lesions on peak enhanced DCE image and K_trans map. Only those voxels with coefficient of determination (R²) > 0.5 for PK model fit and following limits were retained: 0 < K_trans < 5 (min^-1) , 0 < ve < 1 map. Statistical analysis: Analysis restricted to lesions only. We computed effect size (Cohen’s d) for each processing scheme and map within each patient data with Proc#1 as reference. For two sets of longitudinal data (P#1 and #2), we performed repeated measures ANOVA for K_trans and ve maps. Cohen’s d was also computed across time–points for each processing scheme and each map to understand efficacy of a given processing scheme in highlighting therapy changes in longitudinal data.

T₁ alone and T₁ with B₁ correction PK modeling improves depiction of tumor heterogeneity on both K_trans and ve maps (Figure 1). In both longitudinal datasets, trend for map value changes in processing schemes was consistent (Figure 2A and 2C). For longitudinal data analysis, we notice that Proc#3 provides largest effect change from baseline for both K_trans and ve maps. (Figure 3, 4, and F-ratio observed in ANOVA (Figure 5)). However, we noticed that trends were different in one longitudinal dataset (P #1) for ve map across different processing schemes (Figure 3B). While Proc#2 and Proc#4 suggest a steady rise in ve value (more plausible with complete responder status) [1], proc #1 and proc#3 suggest an initial rise, followed by decrease in ve value. Since the primary utility of DCE is to characterize tumor heterogeneity and study impact of therapy on tumor micro-vasculature (i.e. longitudinal trend-lines), results (Fig.1, Fig 3B) suggests that it is more important to utilize DCE-MRI with T₁ and B₁ mapping for studying breast cancer in therapy response cycle.In breast tumor imaging, a DCE-MRI protocol incorporating T₁ and B₁ mapping can be more reliable in reflecting tumor heterogeneity and predicting therapy response.