Multispectral imaging with three-dimensional rosette trajectories

Two-dimensional intersecting k-space trajectories have previously been demonstrated to allow fast multispectral imaging. Repeated sampling of k-space points leads to destructive interference of the signal coming from the off-resonance spectral peaks; on-resonance data reconstruction yields images of the on-resonance peak, with some of the off-resonance energy being spread as noise in the image. A shift of the k-space data by a given off-resonance frequency brings a second frequency of interest on resonance, allowing the reconstruction of a second spectral peak from the same k-space data. Given the higher signal-to-noise per unit time characteristic of a 3D acquisition, we extended the concept of intersecting trajectories to three dimensions. A 3D, rosette-like pulse sequence was designed and implemented on a clinical 1.5T scanner. An iterative density compensation function was developed to weight the 3D intersecting trajectories before Fourier transformation. Three volunteers were scanned using this sequence and separate fat and water images were reconstructed from the same imaging dataset.