Written by He Zhu^1,2 and Rudy Rizzo^1
1Skope MRT, ²Apodibot Medical

During the run-up of the Skope company from Prof. Pruessmann’s research group at ETH Zurich, several publications [1-4] were focused on single shot spiral imaging from 3T to 7T. This can be understood as spiral imaging, particularly versions of long readouts in single shot variants, can most take advantage of the technologies that became the core of Skope’s products. To better articulate this view, we briefly compare spiral to EPI trajectories.

Spiral and EPI imaging 

First, we acknowledge EPI trajectories have been used extensively with nominal, or predicted trajectories, to produce acceptable images in applications such as fMRI and DWI because trajectory and eddy current related artifacts can be better identified and addressed with tailored algorithms. Second, parallel imaging reconstruction methods are more readily available for Cartesian imaging. However, Skope products provide trajectory and eddy current measurements and reconstruction with arbitrary trajectory that include SENSE acceleration and B0 correction. The reward for producing images in this fashion is excellent image quality as shown in the publications mentioned above. This can be attributed to two major advantages of spiral imaging: 1) it naturally obviates sharp turns of readout gradients so it induces less eddy currents; 2) it relies on receivers’ sensitivity profiles in two directions for acceleration so the noise amplification is less, i.e. 4-times acceleration in EPI imaging comes from 4-fold undersampling all along the phase encoding direction while 4-times acceleration in spiral imaging is equivalent to 2×2-fold cartesian undersampling.

With evidence demonstrated at 3T and 7T, single shot spiral imaging was chosen as a showcase example when Skope’s Clip-On Camera system and companion reconstruction software (skope-i) were installed at Tsinghua University in late 2025 on an uMR Jupiter 5T whole body scanner by United Imaging Healthcare. To be noted, the Clip-on Camera system from Skope may be used for concurrent field monitoring as for NeuroCam products at 3T and 7T. In these latter products, the field probes are embedded into the head coil enclosure, occupying spaces between receiver loops and electronic components. MRI data and camera data are then acquired simultaneously, significantly simplifying the workflow. Currently, a similar product is yet to emerge for 5T, so the Clip-On field probes were used in this proof-of-concept study as located in a foam holder that is placed inside the vendor’s head coil to measure trajectories and field dynamics in a non-concurrent experiment (Figure 1). The MRI data was acquired in a separate scan either using a phantom or with a healthy subject’s brain. Another product from Skope, the Dynamic Field Camera (DFC) is designed specifically for this purpose, so our current usage is often informally referred to as “Clip-On as DFC”.

Figure 1: Skope’s Clip-On Camera system used in a 5T whole body scanner by United Imaging Healthcare

Balancing spiral imaging parameters at 5T 

A single shot spiral sequence at 3T (with a 32-channel head coil) featured the following parameters: FOV=23cm, in-plane resolution=1.3mm, slice thickness=2mm, SENSE acceleration factor = 4, readout duration = 32ms and TE/TR = 34ms/5s. Other details of the sequence can be found in Reference [1]. At 7T (with a 32-channel head coil), the corresponding parameters were elevated to: FOV=23cm, in-plane resolution= 0.8mm, slice thickness=1 or 2mm, SENSE acceleration factor =4, readout duration = 53ms and TE/TR = 25ms/3.3s. Other details of the sequence can be found in Reference [2]. At 5T (with a 34-channel head coil), we started with the following parameters: FOV=22cm, in-plane resolution=1mm, slice thickness=3mm, SENSE acceleration factor = 4, TE/TR = 20ms/3.6s and readout duration = 40ms. Total acquisition time of the single shot spiral sequence and dynamic field measurement (i.e., linear terms – equal to the spiral trajectory – and higher order terms) was 8.3s with camera sync function included. Parameters of the spiral trajectory design were chosen approximately at the middle point values between 3T and 7T.

One issue worth mentioning is regarding signal averaging. The 3T sequence was designed from a 4-shot interleaved spiral sequence and 1 of the 4 shots was used to reconstruct final images. However, this one interleave was average 4 times in phantom studies and in vivo. At 7T, a similar design was used but without averaging so it was truly single shot imaging. At 5T, we aimed for the latter design, thus we selected a thicker slice of 3mm to ensure adequate SNR.

Workflow

We used the Pulseq platform to generate single shot spiral and EPI sequences. A companion low resolution GRE sequence was also used to generate B0 and coil sensitivity maps for reconstruction. The Clip-On Camera system was operated by skope-fx that (1) controlled the camera hardware, (2) recorded trajectories and field dynamics and (3) stored them in a compact file of H5 format.

Figure 2 shows detailed comparisons of Matlab generated nominal trajectory (red) and skope-fx measured trajectory (blue). Scanner’s raw data were exported in ISMRMRD format by scanner manufacturer’s software. All data were imported in Matlab and reconstructed by Skope’s image reconstruction software skope-i of which core algorithmics implement High-Order CG SENSE from Reference [3]. The reconstruction time, with 20 iterations, was approximately 33 seconds per slice on a laptop with an Intel Core ultra 9 CPU. GPU processing was unavailable on this unit. The Pulseq platform in this workflow stands out as the framework that enables open-source sequence testing, and it is now supported by most of MRI vendors. Its convenience and accessibility in research for sequence development mostly outweighs its lack of run-time parameter control and online image reconstruction.

Figure 2: Trajectory comparison. (a): A complete nominal trajectory (red) and a complete skope-measured trajectory (blue) are displayed together; (b): a central zoom-in plot shows the measured trajectory lags behind the nominal trajectory from beginning; (c): upper-left quadrant of (a) shows the measured trajectory lags behind nominal trajectory first and catches up; (d) lower-right quadrant of (a) shows the measured trajectory lags behind nominal trajectory progressively further during the entirety of the trajectory. The difference in (c) and (d) shows the non-linearity of the gradient system appears spatially anisotropic. 

Image quality

Figure 3 compares single-shot spiral phantom and brain images reconstructed with nominal trajectories versus field camera measured trajectories. Note that the nominal reconstruction included iteratively and manually tuned trajectory gradient delay to yield proper contrast. Major improvements are seen in terms of increased image sharpness and absence of undersampling artifacts when the reconstruction process includes the measured trajectory, measured hardware delays and measured high-order field drifts.

Figure 3: A comparison of single shot spiral images of a healthy brain and a phantom reconstructed with nominal trajectory (left column), measured trajectory (middle column) and their difference (right column). 

Figure 4 shows in vivo images acquired by the single shot spiral sequence with a 4-fold acceleration in (a) and an EPI sequence with a 4-fold acceleration in (b). The EPI sequence was also implemented in the pulseq framework and similarly reconstructed using skope-i and measured trajectories and higher-order dynamic fields. The EPI sequence in this comparison was designed to have matching FOV and resolution to the spiral sequence. The resulting EPI images are sharp and intrinsically aliasing free (i.e., no phase correction lines nor algorithm were needed). With an undersampling factor of 4, a train of 52 echoes resulted in a TE of 30ms. Compared to the TE of 20ms in the spiral sequence, the EPI images in Fig.4b appear to be darker due to stronger T2* decay. As MRI for advanced neuroscience moves to higher B0 fields, faster T2* decay will require shorter TE to produce proper contrasts. That requirement can be more conveniently met with a spiral sequence than with an EPI design due to its intrinsic location of echo center. Note that the selected TE of 20ms was not the minimum possible for the spiral design.

Figure 4: Brain images acquired by single shot spiral with 4-fold acceleration (a, top) and EPI (b, bottom) sequences.

Some takeaways of MRI at 5T

The 5T system by United Imaging Healthcare at Tsinghua University is equipped with an 8-channel body coil for transmit. The images shown above were all excited using this body coil. As commercial 7T scanners do not come with a body coil for transmit, users need to arrange for separate transmit solutions for individual anatomical regions. At 5T, users have the convenience of choosing lightweight receiver coils as on 3T. The excitation profile of this body coil appears to be flatter than that of a common head-only transmit coil on 7T. In fact, the excitation pulse used in the tested sequence reported only mild darkening volumes, indication of relatively good B1+ homogeneity. Compared to 3T, the SNR gain seems significant as the parameters and images above indicate we reached single shot imaging without averaging. Note that the relaxation parameters at 5T are actively being investigated by early adopters of this scanner, thus future findings may affect readout durations of single shot spiral imaging as we further optimize its parameters.

In conclusion, technologies in Skope’s products enable single shot spiral imaging to surpass EPI in terms of SNR and image quality. These advantages are rooted in underlying principles of physics and engineering. To reiterate them, combining parallel imaging with spiral allows reconstruction algorithm to use coil sensitivity profiles in 2 directions that results in higher usable acceleration factors and more robust reconstruction; gradient coils induce less eddy currents while executing spiral waveforms as they do not make sharp turns in k-space. In the context of fMRI, single shot spiral at 7T can produce sub-millimeter functional maps [4] that resolve activation patterns across cortical layers in an emerging field termed layer fMRI. We are confident Skope’s products may facilitate next-gen neuroscience technologies that enhance knowledge of the brain, healthy and diseased.

ACKNOWLEDGEMENT

We are grateful to Prof. Hua Guo and his graduate students Yuhang He, Fan Liu and Wen Zhong at the Center for Biomedical Imaging Research at Tsinghua University for this collaboration opportunity. 

References

[1] Wilm, B. et al., “Single-shot spiral imaging enabled by an expanded encoding model: Demonstration in diffusion MRI,“ Magnetic Resonance in Medicine, 2016, DOI: 10.1002/mrm.26493.

[2] Engel, M. et al., “Single-shot spiral imaging at 7T,“ Magnetic Resonance in Medicine, 2017, DOI: 10.1002/mrm.27176.

[3] Pruessmann, K. et al., “Advances in sensitivity encoding with arbitrary k-space trajectories,“ Magnetic Resonance in Medicine, 2001, DOI: 10.1002/mrm.1241.

[4] Kasper, L. et al., “Advances in spiral fMRI: A high-resolution study with single-shot acquisition,“ Neuroimage, 2022, DOI: 10.1016/j.neuroimage.2021.118738.