Joint K‐space and Image‐space Parallel Imaging (KIPI) for accelerated chemical exchange saturation transfer acquisition

Abstract

Purpose To develop an auto‐calibrated technique by joint K‐space and Image‐space Parallel Imaging (KIPI) for accelerated CEST acquisition. Theory and Methods The KIPI method selects a calibration frame with a low acceleration factor (AF) and auto‐calibration signals (ACS) acquired, from which the coil sensitivity profiles and artifact correction maps are calculated after restoring the k‐space by GRAPPA. Then the other frames with high AF and without ACS can be reconstructed by SENSE and artifact suppression. The signal leakage due to the T2‐decay filtering in k‐space compromises the SENSE reconstruction, which can be corrected by the artifact suppression algorithm of KIPI. The 2D and 3D imaging experiments were done on the phantom, healthy volunteer, and brain tumor patient with a 3T scanner. Results The proposed KIPI method was evaluated by retrospectively undersampled data with variable AFs and compared against existing parallel imaging methods (SENSE/auto, GRAPPA, and ESPIRiT). KIPI enabled CEST frames with random AFs to achieve similar image quality, eliminated the strong aliasing artifacts, and generated significantly smaller errors than the other methods (p < 0.01). The KIPI method permitted an AF up to 12‐fold in both phase‐encoding and slice‐encoding directions for 3D CEST source images, achieving an overall 8.2‐fold speedup in scan time. Conclusion KIPI is a novel auto‐calibrated parallel imaging method that enables variable AFs for different CEST frames, achieves a significant reduction in scan time, and does not compromise the accuracy of CEST maps.

Numerical fitting of Extrapolated semisolid Magnetization transfer Reference (NEMR) signals: Improved detection of ischemic stroke

“…The long scan time may preclude the method from being used for stroke patients at early time points. Fast imaging methods should be used to alleviate this issue in future work 58–60 . Second, the current study included only single‐slice CEST data, which need to be extended to multiple slices for routine clinical use.…”

“…Fast imaging methods should be used to alleviate this issue in future work. [58][59][60] Second, the current study included only single-slice CEST data, which need to be extended to multiple slices for routine clinical use. Volumetric acquisition sequences should be adopted in future investigations.…”

Acquisition sequences and reconstruction methods for fast chemical exchange saturation transfer imaging

“…The pulse-train CEST scheme has been shown to yield higher SNR and sensitivity than the pulsed steady-state strategy, although the latter may provide sharper images. 64 Another saturation scheme is a hybrid form of the aforementioned two approaches that has a long saturation module and multiple short saturation modules, 28,[68][69][70] as shown in Figure 1D. For multislice CEST imaging, Sun et al 28 first proposed to unevenly segment the RF saturation into two categories, including a long primary saturation module to introduce the steady-state CEST contrast, and multiple short secondary saturation pulses to maintain the CEST contrast.…”

“…As an extension of the vSENSE approach, an autocalibrating reconstruction method by joint k-space and image-space parallel imaging (KIPI) was recently proposed. 70 In this method, the already accelerated frame at a low acceleration factor is first reconstructed by GRAPPA, and then the sensitivity map is obtained to unfold the other highly undersampled frames with SENSE. In essence, KIPI combines the robustness of autocalibrated GRAPPA and the theoretical accuracy of SENSE.…”

Accelerating CEST imaging using a model‐based deep neural network with synthetic training data

“…11 Several image reconstruction strategies have been developed to enable rapid CEST data acquisition, which recovers images from undersampled k-space measurements by exploiting redundancies in the data. One of the most successful techniques is parallel imaging, [12][13][14][15][16] which exploits the redundancy provided by multiple receive coils and turns the ill-posed problem into an overdetermined or determined one. An alternative way is utilizing the sparse prior in the underlying image, which is the foundation of compressed sensing (CS).…”