The volumetric IDEAL IQ[1] (GE Healthcare, Waukesha, WI) sequence provides reliable, quantitative fat/water separation and R2* characterization. The need to repeatedly sample the same k-space lines, however, constrains the spatial resolution that can be achieved in a clinically feasible scan time, as illustrated in the comparative description in Table 1. We present advances in a non-Cartesian approach that solves the spatial resolution limitation while providing several quantitative parameters. The method is demonstrated in 3T bilateral breast MR imaging, where it ultimately can serve as an acquisition platform for high performance DCE-MRI compatible with compressed sensing methodologies. We previously presented a 3D radial trajectory–Vastly undersampled Isotropic Projection (VIPR)[2] with T1-weighting and an out and back acquisition that provides 2 echoes along 2 unique radial lines in each TR. Alternate TRs are time-shifted to effectively produce an approximate p/2 phase shift for fat between each of four uniquely sampled adjacent k-space radial lines[3]. The reconstructed radial echo images are utilized by the IDEAL fat/water decomposition algorithm to provide the final images—VIPR IDEAL. The radial sampling pattern allows for a unique set of radial lines to be acquired in each echo (as opposed to the same lines acquired in Cartesian sampling), which when combined with the efficient out and back trajectory and radial undersampling, leads to a significant improvement in resolution.

Table 2 demonstrates the image acquisition parameters for each of the 3 sequences. To achieve the performance advantages in resolution and scan time listed in Table 2 for VIPR IDEAL relative to the Cartesian IDEAL scan, the PILS effect is exploited through the use of a 16-channel breast coil. Imaging at 3T allows for faster sampling of the spectral dimension as compared to 1.5T, which translates to a shorter TR for higher performance.

One normal and one patient volunteer were imaged using a IRB-approved and HIPAA compliant protocol. The normal volunteer was imaged on a 3T system (Optima 750w, GE Healthcare) with both an 8-channel breast coil (GE Healthcare, Waukesha WI) and a 16-channel breast coil (NeoCoil, Pewaukee, WI) using IDEAL IQ and VIPR IDEAL as shown in Figure 1. The patient volunteer was imaged on 3T system (Discovery 750, GE Healthcare) with a 16-channel breast coil (Sentinelle, Invivo, Gainsville, FL) (Figures 2 and 3) using IDEAL IQ, T1-W FSPGR Two-point Dixon (standard of care Dixon DCE) and VIPR IDEAL as described in Tables 1 and 2.

A comparison of the methods in Figure 2 demonstrates equivalent resolution in the in-plane, axial dimension between VIPR IDEAL, Cartesian IDEAL IQ and T1-W FSPGR two-point Dixon. Figure 1(F) and 3(C) shows that isotropic VIPR IDEAL provides significantly better depiction of fibroglandular detail when reconstructed into orthogonal planes. The substantial benefit of a larger array of coils to minimize undersampling aliasing is clearly apparent in the comparison of the 8 and 16-channel sagittal VIPR images. Robust fat/water separation with VIPR IDEAL is shown throughout.

Though demonstrated in volumetric T1-W imaging, robust fat/water separation with VIPR IDEAL and Cartesian IDEAL IQ supplies a Proton Density Fat Fraction map (labeled PDFF in Table 1) that can be exploited in a DCE compressed sensing reconstruction. Moreover, the B0 maps provided by IDEAL also allow data from multiple echoes to be correctly combined.

The nearly six-fold voxel reduction with VIPR causes some loss of SNR in comparison to Cartesian IDEAL, though this could be mitigated by averaging adjacent slices. Similarly, T1-W FSPGR two-point Dixon scan time combined with its voxel size put it on par with VIPR IDEAL for SNR and performance. Finally, the VIPR sampling patterns can be interleaved in a manner compatible with compressed sensing and reconstructed over varying time intervals with incoherent artifacts, making it suitable for iterative reconstruction algorithm.

We have successfully demonstrated the benefits of adding the sampling characteristics of VIPR in conjunction with IDEAL to obtain high spatial resolution, flexible echo selection and important quantitative properties. Doubling the receiver channel configuration produced a substantially greater performance increase for VIPR relative to the Cartesian methods. Lastly, VIPR IDEAL is a promising acquisition foundation for further study with a DCE Breast MRI protocol.