Dynamic susceptibility contrast MRI (DSC-MRI) is a highly valuable tool for investigating cancer, stroke and traumatic injury. Many studies have shown the need for fast temporal sampling (< 2 seconds) to accurately measure the passage of a contrast agent bolus [1]. The purpose of this study is to combine 3D spiral-based acquisition strategies with GRAPPA parallel imaging to increase the temporal resolution of DSC-MRI.

The Distributed Spirals k-space trajectory (DS) is shown in Figure 1 [2]. A cross-section of the volume in the kz-kr direction shows a uniform sampling pattern which is used to interpolate missing data using GRAPPA [3]. The proposed method is a modification of the method presented in [4], which uses fully sampled data sets through time to calibrate GRAPPA weights for individually undersampled segments of the trajectory.

The fully sampled volume is segmented into 4 bins with interleaved trajectories (ABCD), each bin supporting one fourth the fully sampled field of view. During dynamic data acquisition, the four bins are acquired sequentially, (ABCDABCD, etc); each acquired bin can be used to interpolate missing data for the other 3 bins, creating a fully sampled volume at each time point. As the proposed method acquires bins sequentially, sliding window and GRAPPA reconstructions can be performed on the same data set.

A perfusion phantom was constructed following the method presented in [5]. The rate and amount of contrast agent injection was controlled using a power injector. Data were collected on a 3T Philips Ingenia scanner. Acquiring a single bin took 1.0 seconds; a total of 250 bins were collected. Segmenting the k-space trajectory and calculating GRAPPA weights is performed once, requiring 2.7 minutes on a 32 core, Intel Xeon 3.1GHz processor. Applying GRAPPA weights to undersampled data requires 1.8 seconds, and is performed on each undersampled bin. Data were reconstructed using sliding window, zero-filling unacquired data, and the proposed method.

As shown in Figure 2, the GRAPPA reconstruction reduces aliasing artifacts when compared to the zero-filled image. Due to the nature of the DS trajectory, aliasing is less evident than for (e.g.) Cartesian k-space trajectories. Temporal blurring caused by a sliding window reconstruction is evident in the signal time course for the measured arterial input function (Fig 2A), which causes an artificial widening of the signal time course. The proposed method is a promising method for increasing temporal resolution for 3D DSC-MRI perfusion studies.