CEST imaging is a new molecular MRI method that allows detection of low-concentration, endogenous or exogenous chemicals with exchangeable protons through the water signal [1]. Notably, an endogenous mobile protein-based CEST MRI type, dubbed amide proton transfer (APT) imaging [2], allows detection of low-concentration species in tissue in vivo, potentially extending achievable MRI contrast to the protein level. The CEST imaging acquisition based on the standard imaging scheme is time consuming because multiple RF saturation frequency offsets are always required, limiting its clinical translation despite of many clinical benefits. In this study, the feasibility of accelerated CEST imaging using combined CS and SENSE technique (CS-SENSE) was demonstrated in healthy volunteers and glioma patients at 3 T.

CS-SENSE reconstruction for the high acceleration has two separate stage (Fig. 1) [3]: (1) CS reconstruction for each coil channel with randomly 2D-undersampled k-space data; (2) SENSE reconstruction for the final unfolded image. The randomly undersampled k-space data with reduced FOV from each coil channel can be reconstructed by solving the linear combination of a least square fitting, total variation, and wavelet sparsity regularization [4]:

$$argmin[\frac{1}{2}||Ax-b||_2^2+\alpha||x||_{TV}+\beta||\psi x||_{1}]$$

where A is the undersampled Fourier encoding matrix, x is the image to be reconstructed, b is the undersampled k-space data, and α, β are two positive parameters. The total variation of x is defined by sum of the magnitudes of the discrete gradient at pixels. ψ is a sparsifying wavelet transform. Next, SENSE algorithm is applied for the final unfolded image using sensitivity maps of each coil channel. The final unfolded image can be reconstructed by using a pseudo-inverse formula of the sensitivity matrix C [5]:

$$f=(C^{T}C)^{-1}C^{T}f^{A}$$

where f^A is a set of aliased images with reduced FOV; f is the original full FOV image. Total acceleration factor R of CS-SENSE (R₁ × R₂) is equal to the product of the acceleration factor of CS (= R₁) and the acceleration factor of SENSE (= R₂).

All healthy volunteers (N=2) and patients (N=6) were scanned on a Philips 3 T MRI. CEST imaging sequence consisted of a series of four block RF saturation pulses (200 ms duration each). The frequency sweep corresponded to a full Z-spectrum with 52 frequency offsets, 14 to -8 ppm at intervals of 0.5 ppm. The APT-weighted (APTw) signal was obtained by subtracting the MTR (=1 - S_sat/S₀ where S_sat = saturated and S₀ = unsaturated ) at -3.5 ppm upfield, with respect to water, from that at +3.5 ppm. The correlation coefficients (CC) and the normalized mean square error (NMSE) between the CS-SENSE reconstructed and the reference (SENSE only) image were calculated within whole brain regions.Fig. 2 shows APTw images of the healthy volunteer brain with four RF saturation power levels. CC decreased while the NMSE increased as the CS acceleration factor increased due to elevated noise and aliasing-related artifact resulting from sparse k-space sampling. Fig. 3a-b shows the average Z-spectrum and corresponded MTR asymmetry results from the normal appearing white matter in the reference image. Fig. 3c-e shows correlation and Bland-Altman plots to assess the bias and limits of agreement between CS-SENSE reconstructed and reference images. All CS-SENSE reconstructed maps showed high correlations with the reference, and the Bland-Altman analysis biases were significantly small except for acceleration factors of R = 4×2. The mean differences are 0.00055%, -0.004%, and -0.012% for R = 1.3×2, R = 2×2, and R = 4×2, respectively. Fig. 4a shows one example of the conventional MR images, APTw images and the corresponding error images for a patient with glioblastoma. The mean difference between reference and CS-SENSE (R = 2×2) were -0.0043 % for MTR_asym(3.5 ppm). In addition, the APTw signal of the glioma was significantly higher than that of the normal tissue (p<0.001) as shown in Fig. 4b.Healthy volunteer and tumor patient data have shown that the acceleration factor of R = 4 (2×2) can be achieved without compromising CEST image quality. Therefore, accelerated CEST imaging based on CS-SENSE could be integrated easily into clinical protocols and used for a wide range of clinical applications. The accelerated CEST approach based on CS-SENSE scheme can be extended in several directions. First, there are various ways to design incoherent sampling trajectories such as radial and spiral samplings. Second, joint sparsity in multi-coil images can be applied to improve reconstruction performance by exploiting inter-correlation between the coils to reduce the overall number of measurements. Third, it can be combined with model-based compressed sensing using applying a dictionary, learned from the data model as a sparsity transform.