Peripheral nerve stimulation is a well documented adverse effect of MRI, resulting from transient electric fields induced in tissue from the rapid ramping of the gradients. Magnetic stimulation in tissue was first demonstrated ex-vivo by Oberg¹. The magnetic strength-duration (SD) curve for human nerves in vivo was first shown by McRobbie and Foster² using a topical coil applied to the forearm. Since then many authors have studied PNS in whole body gradient systems, and with a remarkable consistency in stimulus parameters: rheobase and chronaxie^3-7. These results form the basis of the gradient limits in the IEC60601-2-33 standard. It has commonly been assumed that magnetic stimulation is analogous to direct electrical stimulation, and equivalent electrical circuits, particularly the Spatially Extended Non-linear Node model⁸ have been invoked. However recent publications have cast renewed doubt upon aspects of the SENN as applied to magnetic stimuli. In particular the findings that chronaxies determined with magnetic situation may be an order of magnitude longer than for electrode stimulation^9,10. This abstract reviews some of the evidence and offers a hypothetical solution.The literature on magnetic stimulation was reviewed, including that related to transcranial magnetic stimulation (TMS) and in-vitro studies. Existing data², where available, was reanalysed. Topical magnetic stimulation data of the human forearm were fitted using both hyperbolic (chronaxie) and expontential (time constant) fits. Published results relating to to the strength-duration curve from MRI gradient experiments and topical magnetic stimulation were compared against the predictions of the SENN. Waveform dependence was also examined. Predictions of the SENN relating to the above were considered.

Figure 1 shows reanalysed data from topical stimulation with a hyperbolic and exponential fit. Note the difference in calculated rheobases from each. The SENN predicts an A-fibre (motor nerve) membrane time constant (equivalent to the chronaxie) of the order of 0.12 ms⁸. Experiments with MR gradient systems^3-7 constantly give a PNS chronaxie of the order of 0.5 ms (range 0.36-1.054 ms). Few topical studies have examined the chronaxie for PNS, but the earliest² reported a value of 0.47 ms, corroborated closely (mean 0.56 ms, standard deviation (SD) 0.12 ms; mean 0.65 ms, SD 0.05 ms) by recent studies^9,10.

The SENN predicts a rheobase of around 6 V/m for a 20 μm fibre with a dB/dt rheobase for whole body stimulation of 62.5 T/s⁸. Experimental values for dB/dt are all around 20 T/s with E estimated around 2 V/m.

The SENN strongly predicts a stimulus waveform dependence that gives rise to lower thresholds for mono-phasic pulses compared with biphasic pulses. Whilst this may be true for electrical stimulation, even the earliest magnetic stimulation results^1,2 (figure 2) unequivocally contradict this.

From its inception until recently, the SENN model has relied upon data from magnetic stimulation for its validation. Unfortunately the experimental evidence does not fit the predictions well. Reilly and Diamant¹¹ cite "median experimental values of τ_c (the time constant) reported in the range 146-152 μs." Almost all the available evidence contracts this. The waveform discrepancy, i.e. biphasic having lower thresholds than mono-phasic was published prior to the 1989 review which established the SENN for magnetic stimulation⁸.

The SENN forms the basis of several aspects of our MR safety standards, but it inadequately describes well-known magnetic stimulation phenomena. Fortunately, MR manufacturers are able to use experimental PNS data derived from their scanners to set the appropriate stimulation limits. However, sufficient doubt exists to question the cardiac limits which remain theoretical, and the choice of model may affect the derived rheobase value.

The TMS literature has reported that transverse stimulation can occur with even uniform time-varying magnetic fields¹². Indeed a close inspection of Oberg's original work suggests that transverse stimulation was occurring. This mode of stimulation is discounted in the SENN approach, which requires an E-field gradient between nodes. However invoking transverse stimulation may offer the means to reconcile both the chronaxie and the waveform discrepancies, by the accumulation of charge across the membrane - with different temporal characteristics.

The SENN does not adequately predict the behaviour of magnetic stimulation in terms of its time constant and waveform dependence. This has implications for the setting of gradient limits in MRI. Stimulation of the nerve by induced electric fields perpendicular to the nerve axis may remove the inconsistencies. A better model of magnetic stimulation is required.