Single-scan 2D MRI has been generally constrained to acquisitions in high quality magnets. This study introduces a methodology dubbed cross-term spatiotemporal encoding (xSPEN), that delivers such images under much poorer external field conditions. Central to MRI is Mansfield’s echo-planar imaging (EPI) proposition [1], a single-shot approach relying on oscillating gradients to rapidly scan the k-space. During recent years an alternative single-shot method has emerged, based on spatiotemporal encoding (SPEN) concepts [2]. SPEN provides similar acquisition durations, sensitivity and resolution as EPI, but with an enhanced immunity to B₀ inhomogeneities and susceptibility distortions [3, 4]. Still, instances will arise where SPEN’s immunity is insufficient to deliver meaningful data; this study introduces an alternative modality that we denominate cross-term spatiotemporal encoding –xSPEN for short– with good sensitivity and remarkable resilience to field heterogeneities.xSPEN departs from conventional k-space scanning, and relies instead on performing a spatiotemporal encoding of the image being sought. Unlike hitherto proposed SPEN methods, however, this encoding does not utilize only a field gradient along the direction being probed, but adds an ancillary source of inhomogeneous frequency broadening; for instance that imposed by an orthogonal field gradient. Images along the “y” axis are thereby read out in xSPEN by applying a “z” gradient. Figure 1 illustrates one possible implementation of the resulting single-scan xSPEN sequence. This incorporates an oscillating ±G_x readout gradient sampling the k_ro-space of one of the axes, and a constant gradient G_z charged with both the slice selection and the reading out of the second (y-axis) in-plane dimension. Figure 1b illustrates how these oscillating and constant gradients transverse the resulting “hybrid” (k_ro-y) space. All that such sequence needs to transform the resulting signals into xSPEN images, is splitting the resulting FID string into ±G_ro-sampled segments, position these in their correct 2D space coordinates, subject these rearranged arrays to a 1D FT along k_ro, and display in magnitude-mode the resulting matrix.

Figure 2 shows ex-vivo rat brain experiments performed for varying degrees of field inhomogeneities on an Agilent 7T scanner. Compared in this figure are spin-echo multiscan acquisitions acting as “gold-standard”, against single-shot 2D images collected by SE-EPI, SPEN and xSPEN sequences. Notice that while degrading the external magnetic field homogeneity rapidly distorts EPI, and eventually also the SPEN images, xSPEN scans remain insensible to these field distortions. Moreover, other than for a FT along the readout dimension and a magnitude calculation of the ensuing matrix, no special processing, field-mapping or navigator scans are associated to the retrieval of these xSPEN images.

Figure 3 further demonstrates in-vivo single-shot mice acquisitions performed at 7T with three planes. When applied to the selected head region, the multi-scan image acting as reference (Fig. 3a, top) shows clearly the mouse’s eyes, brain and snout. Quality EPI images could only be recorded from the brain’s center; other regions –particularly the eyeballs– appeared severely distorted, and somewhat distorted even in the SPEN experiments (Fig. 3b and 3c). By contrast the single-scan xSPEN images are free from any evident distortions and, apart from SNR considerations, are comparable one-to-one to their multi-scan counterparts. In fact, given that all these experiments were collected without an external respiratory trigger, certain motion artifacts can be noticed in the coronal and sagittal planes of the lengthier multi-scan experiments along their phase-encoded dimensions, which are absent in the single-shot xSPEN scans.

Figure 4 compares xSPEN and SE-EPI head scans acquired on a human volunteer with non-magnetic, metallic dentures. While for clarity only three representative images are shown for each plane, full sets of axial (60 slices), sagittal (40 slices) and coronal images (60 slices) were collected with a 4 mm isotropic resolution. Analysis of these images reveals numerous differences between xSPEN and SE-EPI; most evident among these being the fact that whereas SE-EPI can generally target only the most homogeneous brain regions, xSPEN has no problems to deliver single-shot images for most of the head –regardless of whether containing water or fat. The consequences of the denture are also evident in SE-EPI, which become distorted in proximity to the metallic implants and for which fat suppression becomes inefficient. Likewise, susceptibility effects are clearly visible by SE-EPI’s distorted eyeballs and ears. By contrast, the xSPEN images are remarkably faithful, and affected by relatively small SNR penalties. These and other examples confirm the unusual resilience of this new single-shot xSPEN method to shifts and field heterogeneities.

xSPEN can deliver single-scan MRI with good sensitivity and unprecedented resilience to field inhomogeneities. This ability could enable investigations that have hitherto escaped from MRI’s scope.