High-performance gradients for advanced neuroimaging

The power required to drive a gradient coil scales approximately with the fifth power of the diameter of the gradient coil.1,2 As such, the most efficient manner to achieve high maximum gradient amplitude (Gmax) is to design a gradient coil with a smaller coil diameter. To achieve Gmax = 300 mT/m—a performance level associated with novel scientific and clinical discovery3—a hypothetical high-performance gradient coil powered by dedicated gradient power amplifiers would require 8 MVA of peak power per axis. This equates to of 24 MVA of peak power required for X, Y and Z gradient axes.

For comparison, a conventional 60 to 70 cm patient bore MR scanner requires between 1 to 2 MVA per axis, producing Gmax = 40 to 80 mT/m. In a complete gradient system of X, Y and Z axes, amplifiers must provide 3-6 MVA in peak power. Such a significant increase in power required from this hypothetical high-performance gradient system presents practical challenges for users wishing to adopt such systems, as the cost of infrastructure upgrades and power supplies may not be achievable.

Smaller diameter gradient coils require less power and can achieve higher Gmax with the same gradient drivers used in clinical MR scanners. Recently, dedicated brain imaging gradient coils have been developed that achieve 200-300 mT/m Gmax with 1 to 3 MVA of peak power per axis.4-9 Further increasing gradient peak power to 6 to 7 MVA per axis, dedicated brain imaging gradients have achieved 500 to 600 mT/m10-12, but these systems require peak power far beyond that found in typical clinical MR environments.

The GE HealthCare SIGNA™ MAGNUS‡ (Mesoscale diffusion with Advanced Gradients for Neuro Ultrafast Scanning) 3.0T MR system received U.S. FDA clearance in November 2024.4 SIGNA MAGNUS has Gmax = 300 mT/m and a maximum gradient slew rate SRmax of 750 T/m/s with a 930 A and 2150 V (2 MVA) gradient power amplifier. This article presents data to highlight the strengths and advantages of a high-performance, dedicated brain 3.0T MR system that operates with nominal power supply requirements similar to conventional MR systems.

Table 1 summarizes the key performance features and system differences between the hypothetical high-performance gradient coil, SIGNA MAGNUS and a conventional whole-body MR system. We will endeavor to highlight how these gradient systems perform in the context of key neuroimaging applications, such as diffusion-weighted echo planar imaging. We also introduce different gradient capabilities that influence imaging speed, resolution and physiological safety. Finally, we shall outline how specific types of brain imaging applications are enabled or enhanced with advanced gradient systems such as SIGNA MAGNUS.

Table 1.

Comparison of gradient characteristics for the conventional whole-body (CWB), the hypothetical high-performance whole-body (HPWB) and SIGNA MAGNUS systems.

Gmax and SRmax indicate the maximum gradient amplitude and slew rate (respectively) that can be achieved, unencumbered by physiologic limits, such as PNS. For SIGNA MAGNUS, the patient bore is tapered, with a diameter of 37 cm at the head and 74 cm at the arms/torso regions, while HPWB systems utilize conventional straight cylindrical geometries with uniform patient bores.

Echo planar imaging (EPI) and physiologic limitations

For imaging the brain microstructure, EPI13-15 is an essential tool for fast image acquisitions. In EPI, multiple echoes are acquired by a sequence of repeated trapezoidal gradient waveforms (Figures 1a and 1b). In a system with a high SRmax (i.e., 750 T/m/s) the ramp times for the trapezoidal waveforms can be substantially decreased, allowing significant shortening of the EPI echo spacing (ESP). A shorter ESP time results in increased EPI bandwidth per pixel in the phase encoding direction, which then minimizes geometric distortion and signal loss due to dephasing.14, 16-18

In Figure 2, the maximum allowable slew rate, SRpeak, for whole-body EPI acquisition is limited by peripheral nerve stimulation (PNS)19 to a value well below the system maximum SRmax. For a typical brain imaging protocol, such as with 20 to 24 cm field-of-view (FOV) and 1.5 mm isotropic resolution, this results in ESP = 720 to 800 µs from long EPI readout gradient rise times (Figure 1). In contrast, the head-only SIGNA MAGNUS system benefits from much higher PNS thresholds, permitting the use of the full SRmax of 750 T/m/s without PNS. At the SRpeak = SRmax of SIGNA MAGNUS, the ESP is reduced by almost a factor of two to 340 to 370 µs, effectively doubling the EPI phase encoding pixel bandwidth. As seen in Figure 1, this results in marked improvement in EPI image quality, with reduced geometric distortion and T*-related blurring.8,9,20,21

 

Figure 1.

Effect of high slew rates on EPI. (A) EPI echo train showing long echo spacing time (ESP) between echoes for a whole-body MR system with SRmax ≤ 200T/m/s limited by PNS to SRpeak=120 T/m/s. The long ESP is due to long ramp times for the trapezoidal gradient pulses. (B) EPI echo train for the SIGNA MAGNUS MR system with SRmax ≥ 700 T/m/s. The short ESP enabled by the ability to utilize the full slew rate (i.e., SRpeak = SRmax) while staying within PNS limits also reduces acquisition time. Corresponding EPI images, acquired in (C) whole-body and (D) head-only gradient MR systems. The distortion and signal loss in the pre-frontal cortex (arrows) in the whole-body system are quite evident.

Physiologic limits

As noted in Figure 2, the physiologic threshold for PNS is a fundamental limitation on gradient performance. A readout gradient amplitude of Greadout ≈ 45 mT/m is required to achieve an EPI acquisition with a FOV of 20 to 24 cm. The PNS limits for an MR system are dictated primarily by the design (i.e., size and performance) of the gradient coil and the body area being scanned. For whole-body gradient systems, the PNS threshold limits the allowable slew rate to SRpeak ≈ 100 to 120 T/m/s. Increasing the amplitude of Greadout increases the slope of the gradient waveform, further reducing the achievable SRPNS to well below the 100 T/m/s PNS threshold. In contrast, the PNS threshold for SIGNA MAGNUS, which images only the head, is four to five times higher than that of whole-body gradients due to system design limits of whole-body systems for scanning the body area. As shown in Figure 2, the PNS threshold limit for SIGNA MAGNUS lies above the line for 45 mT/m, indicating that it can operate with no PNS restrictions at this amplitude. As such, for typical EPI brain imaging protocols, the full 750 T/m/s SRmax can be used in the imaging protocol.

Figure 2.

PNS thresholds for whole-body gradients and head-only gradients (SIGNA MAGNUS). The PNS threshold for whole-body gradients is substantially lower (orange) than that for SIGNA MAGNUS (green), demonstrating that whole-body gradient performance is artificially constrained below its maximum in order to remain in the safe operating ranges dictated by PNS thresholds.

Physiologic limitations of PNS and cardiac stimulation (CS) reduce the degree to which high Gmax and SRmax can be used on a hypothetical high-performance gradient system. As shown in Figure 2, all whole-body gradient systems have allowable slew rates far below their system maximum, i.e., SRpeak « SRmax. Maximizing one gradient parameter, SRmax or Gmax, severely limits the other. Consequently, the accessible parameter space (Figure 2, orange area) is severely reduced. This is in comparison to SIGNA MAGNUS, where the accessible imaging parameter space is vast compared to that for whole-body systems (Figure 2, green area).

These limitations have clear implications in applications that require high Gmax, such as diffusion-weighted imaging. In addition to concerns of PNS due to fast gradient switching, there is also an additional safety concern of unintended cardiac stimulation.22 IEC60601- 2-33 states that the maximum E-field (dB/dt) generated by a gradient coil cannot exceed the threshold of

where ts,eff is the effective gradient pulse ramp time. For the same gradient amplitudes, the larger diameter whole-body gradient systems generate higher dB/dt in the patient bore and underscore the need to derate gradient performance in whole-body MR systems to remain within safe operational limits.

Diffusion imaging—ultra-high b-value diffusion

Image quality for diffusion imaging greatly benefits from higher Gmax and SRpeak. The maximum b-value that can be achieved is a function of Gmax and also SRpeak, such that the overall TE is kept as short as possible to have a high image SNR. The PNS-limited slew rate impacts the gradient ramp or rise time, ξ, as shown in Figure 3. ξ contributes to the diffusion encoding pulse width, δ, and the diffusion time, ∆.

If ξ = g/SRPNS , then the waveform ramp time will be limited by PNS and:

Figure 3.

Components of a pulsed gradient spin echo (PGSE) diffusion encoding waveform and EPI readout.

Minimizing δ is desirable, as longer diffusion pulse widths introduce errors due to finite pulse widths that can lead to an underestimation of the diffusion coefficient, D. In addition, sufficiently high b-values for diffusion-weighted imaging are important to sensitizing the signal response to restricted diffusion within impermeable membranes or cellular structures. As shown in Figure 4, higher Gmax and SRpeak allow a substantial reduction in diffusion TE and a corresponding increase in diffusion image SNR.

Figure 4.

Simulations of achievable diffusion TE times as a function of b-value for conventional MR (80 mT/m, 200 T/m/s), HPWB MR (200 mT/m, 200 T/m/s) and SIGNA MAGNUS (300 mT/m, 750 T/m/s). The plots do not take into account necessary derating of Gpeak or SRpeak for whole-body MR systems due to cardiac stimulation concerns. The relative SNR was computed using white matter T2=75 ms.

As b-values increase beyond the typical range of 500 to 5,000 s/mm2, the reduction in TE with high-performance gradients results in SNR gains of up to two times with SIGNA MAGNUS. The increase in SNR allows for either shorter scan times (fewer averages) or increased spatial resolution. With SIGNA MAGNUS, spatial resolution of down to 0.8 mm isotropic has been achieved.

If we assume that there is a SNR threshold that enables ultra-high b-value diffusion imaging, we can make an approximation, as in the dashed red line in Figure 5. Below that line, the SNR would be insufficient for a given acquisition protocol. This sets a threshold for the maximum b-value that can be utilized with adequate SNR. Figure 5 demonstrates that the SIGNA MAGNUS system is able to sustain far higher b-values than whole-body gradient systems. Further, using a diffusion imaging protocol with 40 directions and 2.2 mm isotropic resolution (light blue curve), a conventional whole-body MR scanner intersects the Signal Response = 1.0 at b = 10,000 s/mm2 while a HPWB MR system has an upper limit at b = 23,000 s/mm2. If we assume Signal Response of 1.0 provides an adequate SNR, then further evaluation of Figure 5 also demonstrates that SIGNA MAGNUS produces diffusion imaging with adequate SNR at b-values up to 35,000 s/mm2.

 

Figure 5.

Simulations showing the signal response

as a function of b-values accounting for both changes in TE time and signal attenuation from diffusivity for a whole-body MR (CWB), high-performance whole-body MR (HPWB) and head-only gradient (SIGNA MAGNUS). The arrows indicate the approximate b-value achievable for a given SNR threshold.

Figure 6.

Components of OGSE diffusion encoding waveform and EPI readout for assessing time-dependent diffusion.

Time-dependent diffusivity—using oscillating gradient spin echo encoding (OGSE)

Time-dependent diffusion has been previously demonstrated in pre-clinical, small-bore MR systems equipped with small-diameter high-performance gradients.23, 24 The early work on time-dependent diffusion using OGSE established the potential for detecting tumor microstructure changes from different length scale responses at different OGSE frequencies, ⨍. Sensitivity to length scales as small as µm enables characterization of tumor microstructure on the basis of cellularity or cell size, characteristics that are similarly used to distinguish tissue biopsy samples in histopathology assessments.

Distinct from PGSE (Figure 3), OGSE (Figure 6) requires simultaneous high slew rates (SRpeak) and high Gmax to achieve both high b-values (b = 500–1,000 s/mm2) and high diffusion encoding frequencies (⨍ = 80–120 Hz, equivalent to short diffusion times or pulse widths, δ). For OGSE, the total diffusion encoding time, tenc, must also be kept at a minimum to achieve reasonable TE times for adequate image SNR. These multiple requirements have made routine utilization of OGSE on human whole-body MR scanners elusive due to limitations in gradient performance. SIGNA MAGNUS provides gradient performance comparable to that of pre-clinical MR, enabling routine use of OGSE in patients.

Zhu, et al.25 demonstrated the ability to differentiate between low-grade gliomas and high-grade glioblastoma by assessing apparent diffusion coefficient (ADC) differences at ⨍ = 0 (approximated by a PGSE acquisition), and ADC at ⨍ = 100 Hz using OGSE acquired in an early prototype of the SIGNA MAGNUS system (Gmax = 200 mT/m and SRmax = 500 T/m/s).4 As noted in Figure 7, there is increased differentiation of ADC(⨍) at higher frequencies. OGSE acquisitions at ⨍ = 80 to 120 Hz allows characterization of length scales or cell sizes down to 4 µm, a requirement for characterization of tumor tissue based on cellularity. An excellent review of the engineering requirements for OGSE can be found in Zhu, et al.26

Figure 7.

Estimation of ADC as a function of OGSE frequencies, ⨍, for different length scales of spherical cells with impermeable membranes. The light red regions indicate the parameter space accessible by CWB MR (i.e., up to 30 Hz), while the green region indicates the accessible parameter space for SIGNA MAGNUS. With SIGNA MAGNUS, a higher OGSE frequency (up to 140 Hz) is achieved, improving sensitivity to smaller length scales.

For OGSE, the b-value is substantially reduced compared to PGSE as

where ⨍ = 1/(2∆) is the diffusion encoding frequency for cosine-modulated OGSE diffusion encoding, ξ is the gradient pulse rise time, and N is the number of diffusion encoding cycles (Figure 6). If we assume that ξ = g/SRpeak at g = Gmax, due to PNS limitations (Figure 2), for a head gradient system (SIGNA MAGNUS), ξ ≈ 1.2 ms (SRpeak = 250 T/m/s at Gmax = 300 mT/m).

In terms of image SNR, the higher Gmax enables shorter tenc for OGSE, which leads to shorter TE and higher SNR. As shown in Figure 8, for a b = 800 s/mm2, ⨍ = 100 Hz acquisition, the TE time achieved with Gmax = 300 mT/m is about one-half that with Gmax = 200 mT/m, ignoring PNS limits. If PNS limits are applied, ⨍ = 100 Hz is not achievable for whole-body gradient systems. Achievable OGSE encoding frequency, ⨍, can be increased by reducing the gradient ramp time, ξ. It is important to note that in whole-body systems, this limits the maximum gradient amplitude, g, to far below the system Gmax. To compensate for the loss of amplitude, N must be substantially increased, resulting in a fall-off in image SNR due to the long TE time. We therefore conclude that a head-only MR system with high Gmax and high SRpeak represents the ideal MR system configuration to perform OGSE scans in a clinical setting and with clinically practical acquisition times.

Figure 8.

Figure from Zhu, et al. showing the differences in TE times achievable with different Gmax.26

Table 2 lists the achievable OGSE imaging parameters for CWB and a dedicated head MR system (SIGNA MAGNUS). This table identifies the achievable parameters for OGSE. It exemplifies the reason why OGSE, though previously demonstrated in pre-clinical animal MR systems, has not been readily translatable to clinical use. The advent of dedicated head gradients, such as SIGNA MAGNUS, overcomes PNS restrictions of whole-body MR with the ability to utilize four to five times higher slew rates, all while conforming to safe PNS limits. For a 2.2 mm isotropic OGSE acquisition, images at TE =120 ms have adequate SNR, but images at TE > 120 ms are unusable due to severely challenged SNR.

Table 2.

The achievable maximum OGSE b-values and corresponding TE times with OGSE on different gradient platforms. The scan parameters shown include limitations due to PNS. For the whole-body MR systems, TE times exceed 200 ms for ⨍ ≥ 60 Hz, rendering imaging impractical due to insufficient SNR.

Figure 9.

(A) TE=87 ms, (B) TE=123 ms and (C) TE > 200 ms OGSE images acquired with b800 s/mm2 and f=100 Hz at different TE times acquired on the SIGNA MAGNUS system.
(A) TE times up to 120 ms are acceptable but (C) images at TE>200 ms have signal at or below the noise floor and are unusable (e.g., no imaging data). Images were acquired with 2.2 mm isotropic resolution.

Discussion and summary

This article outlines the unique advantages of a dedicated head gradient coil compared to a conventional whole-body and a hypothetical high-performance, whole-body gradient MR system. Whole-body MR systems, regardless of gradient performance, are still constrained by PNS. Head-only gradient systems, such as SIGNA MAGNUS, have PNS thresholds about four to five times that of any whole-body gradient system, while also staying within cardiac stimulation limits. As such, these systems can fully utilize the full system slew rate for EPI acquisitions, i.e., SRpeak = SRmax. With the improved EPI acquisition, signal dephasing and geometric distortion are markedly reduced.20

The much higher gradient slew rate afforded by SIGNA MAGNUS allows functional activation to be measured in regions of high magnetic susceptibility, such as the entorhinal cortex, hippocampus, temporal lobes and pre-frontal cortex, which has been previously noted to have severe geometric distortion and signal loss in fMRI/EPI studies with whole-body gradient systems.

The combination of high Gmax and high SRpeak, while staying within the PNS limits, allows routine clinical assessment of time-dependent diffusivity for the first time. In addition to the work already demonstrated in brain tumors by Zhu, et al., OGSE has the potential of probing microstructure changes along axonal tracts in cases of neurological disorders, neurodegenerative diseases and stroke. Novel diffusion encoding (such as OGSE) is only achievable with simultaneous high Gmax and SRpeak. The gradient performance of SIGNA MAGNUS allows TE < 100 ms at b = 800 s/mm2 and ⨍ = 100 Hz. This further improvement of gradient performance results in increased SNR and increased sensitivity due to shorter TE and higher b-values, potentially allowing more robust assessment of traumatic brain injuries using a new imaging biomarker.

The smaller gradient diameter in SIGNA MAGNUS raises the PNS threshold to allow the use of much faster gradient rise times than permitted in whole-body gradient MR systems. In addition, there is substantial improvement in image quality and capability that allows new imaging biomarkers to be introduced in a clinical setting. Head gradients allow high Gmax to be achieved with greater power efficiency and substantially less overall power compared to whole-body gradient designs. SIGNA MAGNUS utilizes only 2 MVA of peak power per axis. By using a single gradient power amplifier (GPA) system, SIGNA MAGNUS does not require any synchronization of parallel GPAs to produce high Gmax and high SRmax, which avoids additional system-level complications that may introduce additional errors in gradient amplitude fidelity and real-time compensation of eddy currents. 

‡ 510(k) cleared with the FDA. Not yet CE Marked. Not available for sale in all regions.

References

1. Hidalgo-Tobon S.S. 2010. “Theory of gradient coil design methods for magnetic resonance imaging.” Concepts in Magnetic Resonance Part A 36A: 223–242. https://doi.org/10.1002/cmr.a.20163.
2. Lee, Seung-Kyun, Matt A. Bernstein. 2022. “Systematic Dimensional Analysis of the Scaling Relationship for Gradient and Shim Coil Design Parameters.” Magnetic Resonance in Medicine 88 (4): 1901–1911. https://doi.org/10.1002/mrm.29316.
3. Behjroze, Vachha, Susie Y. Huang. 2021. “MRI with ultrahigh field strength and high-performance gradients: challenges and opportunities for clinical neuroimaging at 7 T and beyond.” European Radiology Experimental 5 (35). https://doi.org/10.1186/s41747-021-00216-2.
4. Foo, Thomas K.F., Ek Tsoon Tan, Mark E. Vermilyea, et al. 2020. “Highly efficient head-only magnetic field insert gradient coil for achieving simultaneous high gradient amplitude and slew rate at 3.0T (MAGNUS) for brain microstructure imaging.” Magnetic Resonance in Medicine 83 (6): 2356–2369. https://doi.org/10.1002/mrm.28087.
5. Weiger, Markus, Johan Overweg, Manuela Barbara Rösler, et al. 2017. “A high-performance gradient insert for rapid and short-T2 imaging at full duty cycle.” Magnetic Resonance in Medicine 79 (6): 3256–3266. https://doi.org/10.1002/mrm.26954.
6. Abad, Nastaren, Chitresh Bhushan, Luca Marinelli L, et al. 2024. “Cross Site Reproducibility for standard and ultra-high b-value diffusion imaging in high-performance gradient (MAGNUS 3T) MRI systems.” Proceedings of the International Society of Magnetic Resonance in Medicine 32: 2516. https://www.ismrm.org/24/pf/D-131.htm.
7. Zhang, Jing, Nastaren Abad, Suchandrima Banerjee, Alexander MacKay, Thomas K.F. Foo. 2024. “Myelin Water Imaging on MAGNUS (high-performance Mesoscale Anatomy Gradient for Neuroimaging with Ultrafast Scanning) system.” Proceedings of the International Society of Magnetic Resonance in Medicine 32: 2519. https://archive.ismrm.org/2024/2519.html.
8. Feinberg, David A., Alexander J.S. Beckett, An T. Vu, et al. 2023. “Next-generation MRI scanner designed for ultra-high-resolution human brain imaging at 7 Tesla.” Nature Methods 20: 2048–2057. https://doi.org/10.1038/s41592-023-02068-7.
9. Vu, An T., Alexander J.S. Beckett, Salvatore T. Torrisi, et al. 2022. “Evaluation of High Resolution Diffusion MRI on the Next-Generation 7T scanner.” Proceedings of the International Society of Magnetic Resonance in Medicine 30: 5036. https://archive.ismrm.org/2022/5036.html.
10. Huang, Susie Y., Thomas Witzel, Boris Keil, et al. 2021. “Connectome 2.0: Developing the next-generation ultra-high gradient strength human MRI scanner for bridging studies of the micro-, meso- and macro-connectome.” NeuroImage 243: 118530. https://doi.org/10.1016/j.neuroimage.2021.118530. 
11. Wu, Dan, Liyi Kang, Haotian Li, et al. 2024. “Developing an AI-empowered head-only ultra-high-performance gradient MRI system for high spatiotemporal neuroimaging.” NeuroImage 290: 120553. https://doi.org/10.1016/j.neuroimage.2024.120553.
12. Ramos-Llordén, Gabriel, Hong-Hsi Lee, Mathias Davids, et al. 2025. “Ultra-high gradient connectomics and microstructure MRI scanner for imaging of human brain circuits across scales.” Nature Biomedical Engineering 1–16. https://doi.org/10.1038/s41551-025-01457-x.
13. Mansfield, Peter, Ronald Coxon, Jonathan Hykin. 1995. “Echo-volumar imaging (EVI) of the brain at 3.0 T: First Normal Volunteer and Functional Imaging Results.” Journal of Computer Assisted Tomography 19 (6): 847-52. https://doi.org/10.1097/00004728-199511000-00002.
14. Jezzard, Peter, Robert S. Balaban. 1995. “Correction for geometric distortion in echo planar images from B0 field variations.” Magnetic Resonance in Medicine 34 (1): 65–73. https://doi.org/10.1002/mrm.1910340111.
15. Chen, Nan-kuei, Alice M. Wyrwicz. 1999. “Correction for EPI distortions using multi-echo gradient-echo imaging.” Magnetic Resonance in Medicine 41 (6): 1206–1213. https://doi.org/10.1002/(SICI)1522-2594(199906)41:6<1206::AID-MRM17>3.0.CO;2-L.
16. Jezzard, Peter, Stuart Clare. 1999. “Sources of distortion in functional MRI data.” Human Brain Mapping 8: 80–85. https://doi.org/10.1002/(SICI)1097-0193(1999)8:2/3<80::AID-HBM2>3.0.CO;2-C.
17. Andersson, Jesper L.R., Chloe Hutton, John Ashburner, Robert Turner, Karl Friston. 2001. “Modeling Geometric Deformations in EPI Time Series.” NeuroImage 13 (5): 903–919. https://doi.org/10.1006/nimg.2001.0746. 
18. Zeng, Huairen, Todd Constable. 2002. “Image distortion correction in EPI: Comparison of field mapping with point spread function mapping.” Magnetic Resonance in Medicine 48 (1): 137–146. https://doi.org/10.1002/mrm.10200.
19. Reilly, J.P. 1998. “Maximum pulsed electromagnetic field limits based on peripheral nerve stimulation: application to IEEE/ANSI C95.1 electromagnetic field standards.” IEEE Transactions on Biomedical Engineering 45 (1): 137–141. https://doi.org/10.1109/10.650371.
20. Tan, Ek T., Seung-Kyun Lee, Paul T. Weavers, et al. 2016. “High slew-rate head-only gradient for improving distortion in echo planar imaging: Preliminary experience.” Journal of Magnetic Resonance Imaging 44 (3): 653–664. https://doi.org/10.1002/jmri.25210.
21. Feinberg, David A., An T. Vu, Alexander Beckett. 2018. “Pushing the limits of ultra-high resolution human brain imaging with SMS-EPI demonstrated for columnar level fMRI.” NeuroImage 164: 155–163. https://doi.org/10.1016/j.neuroimage.2017.02.020.
22. Klein, Valerie, Mathias Davids, Lothar R. Schad, Lawrence L. Wald LL, Bastien Guérin. 2020. “Investigating cardiac stimulation limits of MRI gradient coils using electromagnetic and electrophysiological simulations in human and canine body models.” Magnetic Resonance in Medicine 85 (2): 1047–1061. https://doi.org/10.1002/mrm.28472.
23. Gore, John C., Junzhong Xu, Daniel C. Colvin, Thomas E. Yankeelov, Edward C. Parsons, Mark D. Does. 2010. “Characterization of tissue structure at varying length scales using temporal diffusion spectroscopy.” NMR in Biomedicine 23 (7): 745–756. https://doi.org/10.1002/nbm.1531.
24. Does, Mark D., Edward C. Parsons, John C. Gore. 2003. “Oscillating gradient measurements of water diffusion in normal and globally ischemic rat brain.” Magnetic Resonance in Medicine 49 (2): 206–215. https://doi.org/10.1002/mrm.10385.
25. Zhu, Ante, Robert Shih, Raymond Y. Huang, et al. 2023. “Revealing tumor microstructure with oscillating diffusion encoding MRI in pre-surgical and post-treatment glioma patients.” Magnetic Resonance in Medicine 90 (5): 1789–1801. https://doi.org/10.1002/mrm.29758.
26. Zhu, Ante, Eric S. Michael, Hua Li, et al. 2025. “Engineering clinical translation of OGSE diffusion MRI.” Magnetic Resonance in Medicine 94 (3): 913-936. https://doi.org/10.1002/mrm.30510.