Multi-contrast weighted imaging methods with a single scan[1,2] provide efficient means for viewing and characterizing the atherosclerotic plaques. However, it would be desirable to further reduce the scan time without decreasing image quality. Thus, we sought to extend previous studies by establishing a new black- and gray- blood dual contrast strategy and reducing scan time with compressed sensing [3,4,5].

Compressed sensing based black- and gray-blood dual contrast magnetic resonance imaging (CS-BLGDuC) strategy includes a relaxation enhanced compressed sensing three-dimensional motion-sensitizing driven equilibrium prepared 3D rapid gradient echo sequence(RECS-3D MERGE)[6] and under-sampling k-spaces for both black blood and gray blood image(Fig.1). Black blood k-space and gray blood k-space were filled with signal acquired at different time within one scan. Based on previous study [6], the pseudo-centric phase encoding order was used (bottom right of Fig 1), the corresponding acceleration factor (AF) was 3, and the delay time was 800ms.

For 3D MRI, after one-dimensional inverse Fourier transformation along $kx$, the signal at each $x$ position is obtained and can be expressed as: $$s=ϕm+n$$ where $s$ is the signal vector, $ϕ$ is the undersampled Fourier operator obtained by a measurement matrix. A Monte-Carlo method is used to construct the measurement matrix [7]. $m$ is the unknown image estimation and $n$ is a vector representing and independent and identically distributed additive white Gaussian noise. According to the CS reconstruction algorism of MRI image proposed by Lustig [7], it is probable that $m$ can be exactly reconstructed if $m$ is sparse in a transform domain by solving the $ℓ1$-norm optimization problem: $$minimize∥ϕm−s∥2+λ1∥ψm∥1+λ2TV(m)$$ where $ψ$ is the sparsifying transform, $TV$ is the total variation of $m$, $λ1$ is the regularization weight for the sparsifying transform and $λ2$ is the $TV$ regularization term.

16 subjects including 6 healthy volunteers and 10 patients with carotid stenosis were carried out. Institutional review board approval and informed consents were obtained. Images were acquired on a GE Signa TM 3T scanner (GE Medical Systems, Milwaukee, WI) with an eight-channel carotid coil. Conventional QIR and TOF were also scanned to compare the image quality, including CNR and plaque depiction. Detailed scan parameters are provided in Table 1.

The scan time of CS-BLGDuC were dramatically reduced compared with that of conventional 3D TOF and QIR. No Statistically significant differences in CNR between QIR images and black blood images acquired by CS-BLGDuC were detected (P= 0.05). Those CS-BLGDuC based black blood images could provide sufficient blood suppression, clear vessel wall and good contrast between vessel wall and lumen. While gray blood images have lumen with uniform signal intensity and the contrast between the arterial wall and lumen is not as high as that of black-blood images (Fig 2).

Fig.3 shows images obtained in a patient with carotid arterial stenosis using CS-BLGDuC, 3D TOF and 2D QIR-FSE. Superficial vascular calcification was well visualized (solid arrows) and clearly distinguished from arterial lumen in Figure 3a. In comparison, calcifications were less conspicuous with 3D TOF image due to the suboptimal delineation of the inner and outer boundaries (dashed arrows) of the carotid arteries in Figure 3b. In addition, we realized that although 2D QIR-FSE images could provide satisfied blood suppression and vessel wall visualization, clear superficial vascular morphology and distribution were hard to get.

It is feasible for the CS-BLGDuC strategy to depict superficial vascular calcifications and improve scan efficiency without decreasing image quality. Potentially, the CS-BLGDuC strategy could be valuable for 3D carotid plaque imaging.