Functional neuroimaging of the human spinal cord has gained increasing interest in the past few years (e.g. [1]). Recently, an echo-planar imaging approach has been presented that covers both regions in the same acquisition [2] and can be used to investigate the functional connectivity between these two regions [3]. However, with standard resolutions and imaging volumes, the repetition time (TR) usually exceeds 3 s which is not optimal for a connectivity analysis. Multi-band acceleration (e.g. [4]) can speed up acquisitions, however, suffers from noise amplification for some of the neck coils available. Here, a variant is presented that uses conventional (single-band) imaging for the spinal cord slices to avoid noise amplification but multi-band acceleration for the brain slices to obtain a significantly shortened TR.

The basic pulse sequence and geometric arrangement of the echo-planar imaging approach used is presented in Fig. 1. It involves two slice groups covering the brain and the spinal cord, respectively, with independent geometry and timing parameters (field-of-view, slice thickness, bandwidth, echo time) as well as coil elements and a dynamic shim update between the two slice groups [2] (see Fig. 1b). For the brain slices, blipped-CAIPI [4] is used with a multi-band RF excitation and additional gradient blips in the slice direction in order to induce an apparent shift in the image that depends on the band’s distance from the isocenter and improves the performance of multi-band imaging considerably [4]. The image reconstruction was split into two steps with the brain slices being calculated on-the-fly during the acquisition and the spinal cord slices reconstructed retrospectively. The corresponding modification of the reconstruction protocol parameters (bandwidth, number of slices, etc.) was performed automatically with a script.

Measurements were performed on a 3 T whole-body MR system (Magnetom TIM Trio) using a 12-channel head and a four channel neck coil. Water phantoms and healthy volunteers from which informed consent was obtained prior to the examination, were investigated. Imaging parameters for the brain / spinal cord slice group were a voxel size of 2.0×2.0×2.0 mm³ / 1.0×1.0×5.0 mm³, a slice gap 1.0 mm / 0.0 mm, a bandwidth per pixel of 1532 Hz / 1082 Hz, an echo time of 30 ms / 35 ms, and parallel imaging with an acceleration factor of two (both groups). Without multi-band acceleration, a minimum TR of 3040 ms was obtained, and a flip angle of 90° could be used. Using multi-band acceleration with a factor of two for the 32 brain slices, a minimum TR of 1970 ms could be achieved, and a flip angle of 70° was used. For test purposes, single slice groups without and with multi-band acceleration were acquired with the different coils.

Images of a single slice group acquired with the 12-channel head and the neck coil without and with multi-band acceleration are presented in Fig. 2. Both coils yield good results without multi-band acceleration. However, with multi-band acceleration the neck coil images suffer from severe noise amplification while the images obtained with the 12-channel head coil are reasonable. Thus, applying multi-band acceleration to the spinal cord slice group is not recommendable and a “partial” multi-band acceleration for the brain slice group only has been used for subsequent measurements.

Results of combined acquisitions with the two slice groups for the brain and the spinal cord are presented in Fig. 3 and 4 for phantoms and in vivo, respectively. A dynamic shim update between the two slice groups was performed and slice-specific z-shim gradient pulses were applied to optimize the image quality in the brain and spinal cord slice group, respectively. While the image quality in the spinal cord is retained and only minor artefacts are observed in the brain when multi-band acceleration is used, the TR could be decreased by about 35%.

With the shorter TR achievable with the partial multi-band acceleration, the temporal resolution of combined acquisitions of the brain and the cervical spinal cord can be improved considerably without affecting the image quality in the spinal cord. The brain slices are more prone to artefacts related to multi-band acceleration, most likely because of residual field inhomogeneities and their position far from the isocenter. In conclusion, partial multi-band acceleration may help to improve measurements of the functional connectivity between the human brain and the cervical spinal cord.