Upper-limb muscle responses to epidural, subdural and intraspinal stimulation of the cervical spinal cord

Abstract

Objective: Electrical stimulation of the spinal cord has potential applications following spinal cord injury for reanimating paralysed limbs and promoting neuroplastic changes that may facilitate motor rehabilitation. Here we systematically compare the efficacy, selectivity and frequency-dependence of different stimulation methods in the cervical enlargement of anaesthetized monkeys.

Approach: Stimulating electrodes were positioned at multiple epidural and subdural sites on both dorsal and ventral surfaces, as well as at different depths within the spinal cord. Motor responses were recorded from arm, forearm and hand muscles.

Main results: Stimulation efficacy increased from dorsal to ventral stimulation sites, with the exception of ventral epidural electrodes which had the highest recruitment thresholds. Compared to epidural and intraspinal methods, responses to subdural stimulation were more selective but also more similar between adjacent sites. Trains of stimuli delivered to ventral sites elicited consistent responses at all frequencies whereas from dorsal sites we observed a mixture of short-latency facilitation and long-latency suppression. Finally, paired stimuli delivered to dorsal surface and intraspinal sites exhibited symmetric facilitatory interactions at interstimulus intervals between 2–5 ms whereas on the ventral side interactions tended to be suppressive for near-simultaneous stimuli.

Significance: We interpret these results in the context of differential activation of afferent and efferent roots and intraspinal circuit elements. In particular, we propose that distinct direct and indirect actions of spinal cord stimulation on motoneurons may be advantageous for different applications, and this should be taken into consideration when designing neuroprostheses for upper-limb function.

Figures

Figure 1. Stimulation methods, sites and example intensity series.

(a) Cross-section of the spinal cord to show locations of epidural, subdural and intraspinal stimulation. (b) View of the spinal cord surface exposed by laminectomy in each animal. Circles indicate locations of stimulation sites. For dorsal approaches, locations of spinous processes (C5 to C7) are indicated. For ventral approaches, the locations of spinal roots are indicated. Dashed lines indicate approximate midline of the spinal cord. (c) Top row shows example EMG responses to single dorsal subdural stimuli (site x in panel b) at intensities between 20 to 200 μA. Bottom row shows mean, rectified response to 20 stimuli. Red traces indicate a significant effect of stimulation. (d) Example responses to ventral subdural stimuli (site y) at intensities between 5 to 50 μA. (e) Example responses to ventral subdural stimuli (site y) at intensities between 20 to 200 μA. (f) Response functions for example stimulation sites, normalized to the saturation level. Filled circles indicate significant responses. Lines show cumulative normal distribution fit.

Figure 2. Efficacy of muscle activation by different spinal cord stimulation methods.

(a) Summary of significant muscle responses for different intensities of stimulation delivered to an example dorsal subdural stimulation site (site x in figure 1(b)). Also shown is the recruitment curve (below) obtained by averaging across muscles, and response pattern (right) obtained by averaging across intensities. Muscle groups are shaded according to most preferred (black), intermediate (dark grey) and least preferred (light grey). (b) Average recruitment curves for different stimulation methods. Shading indicates standard error of mean. (c) Mean threshold for muscle recruitment by different stimulation methods. (d) Proportion of significant muscle responses averaged over the range 20–200 μA for different stimulation methods. (e) Average intensity range over which graded muscle responses could be obtained (from 5% to 95% of saturation level) for different stimulation methods. Bars indicate standard error of mean. (f) Mean response function fit for each method of stimulation, normalized to the level at which the response saturated.

Figure 3. Selectivity of muscle responses to different stimulation methods.

(a) Average recruitment curves for most preferred, intermediate and least preferred groups of muscles (arm, forearm and intrinsic hand). Dashed line indicates best-fit cumulative normal distribution. (b) Mean parametric selectivity, defined as ratio of across group variance to within group variance. (c) Mean non-parametric selectivity index, defined as normalized difference in efficacy between most preferred and least preferred muscle groups. (d)–(f) Equivalent analyses for forearm muscles grouped according to flexors and extensors. Bars indicate standard error of mean.

Figure 4. Dependence of similarity between muscle response patterns on electrode separation for different stimulation methods.

(a) Example similarity matrices for different stimulation methods, showing the correlation coefficient between response patterns evoked from pairs of stimulation sites. For surface stimulation, similarity is calculated between medial (M) and lateral (L) sites at different rostro-caudal locations separated by up to 25 mm. For intraspinal stimulation, similarity is calculated between all electrode positions within a single array penetration separated by up to 1.5 mm. (b) Average similarity for surface stimulation as a function of increasing rostro-caudal separation. Dashed lines indicate fit of linear regression with three explanatory variables: separation, side (dorsal/ventral) and method (epidural/subdural). (c) Average similarity for intraspinal stimulation as a function of increasing depth separation between electrodes within the linear array (open circles). Also shown is similarity between comparable electrode sites across different electrode penetrations (filled squares; mean electrode track separation 7 mm, range 2.5–15 mm).

Figure 5. Muscle responses to stimulation trains.

(a) Example EMG response in abductor pollicis brevis to trains of 15 stimuli delivered to a dorsal subdural site at 10, 50 and 100 Hz. Note the difference in time-base between traces such that stimuli appear equally spaced. (b) Example EMG response in flexor carpi radialis to trains of stimuli delivered to a ventral subdural site. (c) Time-course of modelled facilitatory (red) and suppressive (blue) influences of a preceding stimulus on subsequent responses. Green line indicates the resultant effect which changes from facilitatory (left of vertical dashed line) to suppressive (right of vertical dashed line). Responses below the horizontal dashed line are set to zero. (d) The strength (amplitude multiplied by time constant) of facilitation and suppression for different stimulation sites determined by model fit. Bars indicate standard error of mean.

Figure 6. Average train responses and model fits for different stimulation methods.

(a) Colour maps showing the magnitude of response to each pulse in trains of different frequencies, normalized by the response to the first stimulus. Each plot is the average of all the muscle-stimulation site combinations for which there was a significant response to the first stimulus. (b) Average model fit for the same muscle-stimulation site combinations.

Figure 7. Interactions between paired subdural and intraspinal stimulation.

(a) Example EMG recordings (left) and average rectified responses (right) from first dorsal interosseous muscle, in response to individual dorsal subdural and dorsal intraspinal stimuli as well as paired stimuli with 13 different time intervals. Grey bar represents the time window for assessing nonlinear EMG response. Grey lines indicate response predicted by linear summation of responses to individual stimuli. Black lines show actual responses that do not differ significantly from the linear prediction. Red lines show actual responses that are significantly greater than the linear prediction. (b) The proportion of supra- and sub-linear interactions between dorsal subdural and intraspinal sites for different interstimulus intervals. (c) Example responses from flexor carpi radialis to paired ventral subdural and ventral intraspinal stimuli. Red line shows a response that is significantly less than the linear prediction. (d) The proportion of supra- and sub-linear interactions between ventral subdural and intraspinal sites for different interstimulus intervals. Note that panels b and d are colour-coded according to depth of intraspinal electrode relative to the surface that was stimulated (blue: superficial, red: intermediate, green: deep). (e) Average proportion of significant interactions (supra- and sub-linear) across all time intervals and sides (dorsal and ventral) divided according to the depth of intraspinal electrode relative to the surface that was stimulated. Bars indicate standard error of mean.