Whole-body MRI is of interest in the assessment of many diseases such as obesity, diabetes, cancer and circulatory malfunctions.^1-4 Continuously Moving Table (CMT) MRI is a high-throughput imaging technique that offers a powerful alternative for rapid whole-body MR examination. In CMT MRI, data are acquired with continuous motion of the patient table, allowing scanning of multiple locations in the body in a single sweep. Together with a continuous golden angle (GA) radial acquisition, CMT MRI image the whole-body in 1-2 min.⁵ One downside of the high speed is lower signal-to-noise ratio. Using an RF receive coil with multiple channels can improve SNR if the channels are combined appropriately. Here we describe the development of a single-pass, calibrationless, multi-channel whole-body CMT MRI technique with significantly improved image quality and no additional scan time requirement.

Experiments were performed on a Philips Achieva 3T (Philips Healthcare, Best, Netherlands) scanner equipped with two-channel parallel transmit capability. A 16-channel Torso-XL surface coil (Invivo, Gainesville, Florida) was used with an X-tend tabletop (X-tend ApS, Hornslet, Denmark) to enable the acquisition of multichannel CMT data. Figure 1 shows the general setup for the X-tend tabletop. The anterior 8-channel section of the Torso-XL hung from the top of the scanner bore in a fabric sling. The sling was allowed to hang low enough to slide against the subject’s body in order to maximize SNR. The posterior portion of the coil was placed in a rolling “coil wagon” sandwiched between two layers of the X-tend tabletop. As the X-tend table moved through the scanner bore, the coil wagon was held stationary at isocenter by straps attached to the scanner bore covers. The straps limited the range of z direction table movement to 1.6 m.

An adult male volunteer was scanned at 3 T under an IRB-approved protocol. The volunteer entered the magnet feet first, supine with arms at his side. Imaging was performed with full zFOV = 1.6 m, table speed = 2 cm/s, in-plane FOV = 40 cm x 40 cm transverse, in-plane voxel size: 1.56 mm x 1.56 mm, TR/TE: 3.7/1.35 ms, flip angle: 15°, radial samples: 256, excited slice thickness: 12 mm, readout bandwidth: 854 Hz/pixel, 21,622 radial projections, total scan time: 80 s. Pre-scan optimization consisting of center frequency and RF drive scale determination was performed in the abdomen at the position of the umbilicus. After the preparation phase, the table moved such that the head was slightly inferior to the z position at isocenter, followed by the CMT scan with the 16-channel array covering the 1.6 m FOV.

Slice-wise complex image reconstructions were performed using the Trajectory Optimized Non-Uniform FFT (TRON) algorithm,⁶ a GPU-accelerated reconstruction customized for GA radial acquisitions. Reconstructed slice centers were separated by 1.56 mm (21 radial projections), matching the in-plane resolution, and were 18.9 mm (256 radial projections) thick. Reconstruction of the 953 slices of 512 x 512 images required 11.3 s for the 16-channel data and 9.7 s for the body coil data on a Xeon workstation with dual Nvidia GeForce GTX TITAN X cards. A Gaussian gridding kernel with width of 3.0 was used with a grid oversampling factor of 1.25. As part of the reconstruction, coil combination was performed in image space using the adaptive method of Walsh et al.⁷ because it preserves the relative complex phase and slightly improves the signal-to-noise level.

Figure 2 shows three axial slices from the volume from both the built-in quadrature body coil and Torso-XL data. The multichannel image has significantly reduced artifacts, and SNR improved from 12.9 with the body coil to 19.6 with the 16-channel coil. Uniformity is similar in both images, with notably better sensitivity with the 16-channel coil to the spinal column. The improvement over using the built-in body coil is clear in even this simple adaptive combination of the channels, suggesting that simply having the receive coils closer to the subject and a larger number of receive channels is a great benefit. A notable downside is that the additional hardware inside the bore reduces the useful bore radius and limits the potential for applications with large, e.g. obese, subjects. Hardware improvements could include a coil designed specifically for CMT MRI, with the majority of elements concentrated at the isocenter and with thin profile or even built into the scanner. Software improvements could include moving to an iterative reconstruction with phase constraints.Significantly improved image quality and SNR in whole-body CMT MRI was achieved without time penalty using a multichannel acquisition coupled to a calibrationless reconstruction.