Introduction

Gradient-induced peripheral nerve stimulation remains a major performance limitation in MRI. Imaging sequences are limited to experimentally-determined thresholds that reduce the likelihood of uncomfortable stimulation in patients. Traditionally, only EPI-like pulse trains (consisting of bipolar trapezoids) or sinusoidal waveforms are tested in these experiments[1] and the threshold limits set using that data. However, this worst-case pulse sequence may lead to overly conservative limitations on other gradient waveforms. There is evidence that short waveforms have higher stimulation thresholds[2]. Common short waveforms found in pulse sequences such as spoilers, crushers, and diffusion pulses could possibly be optimized for higher stimulation thresholds. In this study, 20 subjects were exposed to three types of common short gradient pulse waveforms and the observed stimulation thresholds were compared to those of a long bipolar trapezoid train.

Methods

20 subjects were placed within a prototype head-only gradient coil. Subjects were moved into the coil until just before their shoulders contacted the face of the coil, and then were run through an initial session to improve consistency of stimulation reporting. Subjects were then exposed to four different gradient waveforms (Figure 1) in randomized order consisting of a 32-cycle (64-lobe) bipolar trapezoid pulse train, a single separated unipolar trapezoid waveform (“spoiler type”), a separated unipolar trapezoid pairs waveform (“crusher type”), and finally a long duration unipolar trapezoid pairs waveform (“diffusion preparation”). The waveforms were each tested at rise times of 0.2 ms, 0.25 ms, and 0.3 ms, and were repeated every 1.5 s. With every repeat the amplitude was raised by 2.5% of the maximum, and subjects were asked to report the onset of stimulation. The full set of tests performed had three repeats for each rise time, each axes combination (X,Y,Z,XY,XZ,YZ), and each waveform. Every test on each axis was randomly ordered. The stimulation data was then averaged over the three repeated tests to compute subject averages. If less than 5 subjects stimulated at that rise time the data was not included in the calculation. If all 20 subjects stimulated, the mean of that data was used. If between 5 to 19 subjects stimulated, a logistic regression was performed to estimate the mean[3]. The means were then used to determine a linear relationship for that axis and waveform. The linear relationship was used to compare the thresholds at the three tested rise times to show the difference between the traditional method (Table 1).

Results

Table 1 summarizes the results. For the bipolar train waveform (Figure 1A) all 20 subjects stimulated a total of 852 times across all axis combinations. The spoiler type waveform (Figure 1B) resulted in 11 subjects stimulating 300 times across all axes, and the crusher type waveform (Figure 1C) resulted solely in stimulations on the X axis, and only 5 subjects stimulated 44 times. Finally, the diffusion preparation pulses (Figure 1D) had only stimulations reported on the XY and XZ axes combinations, with 11 subjects stimulating 235 times. Threshold limits on the XY axis for the spoiler type waveform were an average of 128% higher, and for the diffusion preparation waveform they were an average of 131% higher than the bipolar pulse train. Notably, not enough stimulations were experienced on the crusher waveform to calculate the thresholds. While most stimulations experienced were pressure, buzzing, or tingling, a significant number of subjects (N = 8) experienced stimulation as magnetophosphenes within the diffusion preparation waveform.

Discussion

The results presented here are in general not surprising – the fact that shorter numbers of pulses in a train results in higher thresholds has been observed previously, and in particular as part of the electrostimulation literature[4]. But in the context of specific classes of gradient coils for MRI (i.e. head gradient coils versus body coils), the extent of the threshold increase is very important to measure, particularly in light of the need for high gradient strengths and slew-rates for DTI applications. These results indicate that gradient performance increases of between 20 and 40% may be possible in small pulse numbers as are used for spoiler, crusher, and diffusion gradient pulses.

Acknowledgements

NSERC and the Ontario Research Fund