Motor Cortex Stimulation for Pain and Movement Disorders

Surgical and cortical mapping technique 

Although surgical access to the M1 region can be obtained by means of only a single burr hole, we prefer to make a small craniotomy opening. The midpoint from nasion to inion is determined and an opening centered over a line from ∼1 cm behind this midpoint toward the anterior margin of the tragus is made using a curved scalp incision. In this way, exposure for a craniotomy that measures roughly 5–6 cm in diameter is made and can be adjusted slightly superiorly or inferiorly along the convexity to map face or hand as desired. Typically, a cerebellar retractor is the only instrument necessary to maintain the opening during the procedure. We do not use a Mayfield head holder, because the head does not need to be rigidly held in place, which eliminates the risk of further injury if a patient has a seizure induced by mapping.

Both somatosensory and motor mapping can be used intraoperatively to determine the course of the central sulcus and the M1 region underlying the dura. We use both methods in each case, for internal corroboration. Somatosensory testing consists of placing the four-contact Resume lead (Medtronic, Minneapolis, MN) on the dura in a variety of directions, mostly perpendicular to the suspected precentral gyrus. Median and ulnar nerve somatosensory evoked potentials (SSEPs) are then run using a 20 mA–100 μs monopolar square pulse at a rate of 4.32 Hz. SSEPs are recorded from the Resume lead in both a bipolar (contact 0–1, 1–2, 2–3) and a monopolar (all referenced to the 10–20 location of Fpz) recording montage. The central sulcus is determined as the point at which the N20 response phase reverses (FIG. 1). This process is performed in multiple locations, to map out the central sulcus over the complete craniotomy opening, and short regions just under the bone opening if needed.

View Large Image Download to PowerPoint

FIG. 1. Use of the somatosensory evoked potentials (SSEP) reversal potential method to determine location of M1 extradurally. Contacts 0–3 of a typical four-contact paddle-type electrode are shown, in this example with placement across the underlying central sulcus. The waveforms at the upper left show the SSEP in each contact. The reversal of phase occurs, in this example, between contacts 1 and 2. The inset at the lower left is the intraoperative photo of this technique in use, revealing the relative size of the lead and the craniotomy opening. By moving the lead around in different locations, the path of the sulcus can be mapped out on the dura.

Motor mapping consists of placing an anodal 5-mm stimulation ball probe (Model E1564; Valleylab, Gosport, UK) on the dura over the M1 area referenced to a cathode placed at Fpz. Stimulation consists of trains of five stimuli each, at a rate of five trains per second, a 500-μs pulse width, and a 4-ms interspike interval. Stimulation amplitudes are slowly increased at each location starting at 5 mA and increasing to a maximum of 25 mA. Stimulation can then be stopped when the first electromyographic (EMG) response is noted.

EMG needles are placed in bipolar fashion (separated by 2 cm) in the orbicularis oculi, orbicularis oris, trapezius, deltoid, biceps, triceps, flexor carpi ulnaris, abductor pollicis brevis, first dorsal interosseous, quadriceps, tibialis anterior, and abductor hallucis muscles. Stimulation is performed with a Grass S-88 and two SIU-7 constant current stimulus isolators (Astromed-Grass, West Warwick, RI). Responses are recorded on a Cascade neurological monitoring system (Cadwell, Kennewick, WA). Figure 2 shows an example of the findings in the extensor and abductor pollicis brevis muscles with this technique. In this fashion, two independent types of intraoperative physiology may be used in locating the precentral gyrus.

View Large Image Download to PowerPoint

FIG. 2. Use of the cortical mapping technique for determining location of M1 regions. Three electromyograms (EMG) from muscles in the upper extremity are shown, after stimulation with a ball-tip electrode in three locations on the dura. The EMG at the far left shows activation of the extensor muscle in the forearm; the other two show activation of the abductor pollicis brevis muscle in the hand. This technique corroborates the SSEP method (FIG. 1) and helps determine more precisely the underlying thresholds for individual muscle groups.

Critical for the ability to obtain these neurophysiological data is the accompanying anesthetic technique. We use a complete TIVA (total intravenous anesthetic) protocol, with a continuous dose of propofol combined with either fentanyl or remifentanil. Standard doses are in the range of 75–150 μg/kg/min for propofol and 0.05–0.5 μg/kg/min of fentanyl or remifentanil. No nitrous inhalational agent is used and no muscle relaxants are used.

Patient programming is typically begun within 24 h of electrode implantation. At the initial programming session, all contacts are checked in a monopolar setting using 210 μs and 130 Hz. The voltage is slowly raised to 4.0 V, to look for adverse motor movements and sensory changes. As amplitude is slowly raised, the patient state is evaluated. One primary concern is the possible generation of seizure activity during these tests. To date, we have had two seizures generated during the programming phase. In one patient, a focal motor seizure may have occurred 2 days after initial programming; it was not witnessed by medical personnel, but the description fit well with this type of seizure. This was a patient in whom a seizure was generated during M1 mapping in the operating room. Since then, all our MCS patients in whom a seizure occurred during intraoperative mapping are kept at <2.5 V for the first 2 weeks after surgery. We do not prescribe antiepileptic medications, however. At 18-month follow-up in 1 poststroke pain patient, a seizure was generated while the stimulator voltage was being raised from 4.0 V to 4.3 V. The patient’s voltage has since been kept below 4.0 V, and no further seizure activity has been noted.

In our four PD patients, contacts 1 and 2 were left on, with the stimulator amplitude at 3.0 V, assuming the patient was able to tolerate each of these settings. If a patient felt unable to tolerate these settings, for any reason, then either contact 3 or 0 alone was used instead. In all cases, it was possible to activate two contacts at the initial programming session.

The use of cathodal stimulation may seem odd, given that it is known that, with transcranial motor stimulation, anodal stimulation will generate D-waves and motor evoked potentials at a lower intensity than with cathodal stimulation.69 Due to the design of the present implantable stimulators, however, options are limited to cathodal stimulation only. Also, and more importantly, work by Hanajima et al.70 found that cathodal chronic stimulation activated neurons in the motor cortex at levels lower than did anodal stimulation. If minimal benefit was noted at later sessions, more contacts were added to the monopolar configuration. If this situation failed, then one contact (either contact 0 or contact 3) was placed in the cathodal mode and the other three contacts were set as the anode. Other changes that were investigated when stimulation either did not work, or started to lose efficacy, included changing the pulse width to 410 μs, increasing the amplitude to 4.0 V, or changing the frequency to either 185 Hz or 80 Hz. Table 1 shows the stimulation parameters relative to the surgical implant date.

TABLE 1. Motor Cortex Stimulation (MCS) Settings at Last Follow-Up Visit

	Patient	Time Since Surgery, mo	Left Hemisphere								Right Hemisphere
			Amplitude, V	Pulse Width, μs	Frequency, Hz	Contacts				Case	Amplitude, V	Pulse Width, μs	Frequency, Hz	Contacts				Case
						0	1	2	3					0	1	2	3
	MCS1	18, 6⁎	3.2	240	130	−	−	−	n/a	+	—	—	—	—	—	—	—	—
	MCS2	12	3.5	210	130	−	−	+	+	n/a	3.5	210	130	n/a	−	−	+	n/a
	MCS3	12	3.4	210	100	−	−	−	n/a	+	3.2	210	100	−	−	−	n/a	+
	MCS4	6	OFF	n/a	n/a	n/a	n/a	n/a	n/a	n/a	OFF	n/a	n/a	n/a	n/a	n/a	n/a	n/a

n/a = not applicable; − = cathode; + = anode; — = no electrode was implanted on this particular side for this patient.

⁎ A new implantation 12 mo after the first surgery.

In mixed poststroke patients or pure pain patients, programming was basically similar during the initial phase. Patient programming is typically begun, again, within 24 h of electrode implantation. At the initial programming session, all contacts are checked in a monopolar setting using 210 μs and 130 Hz. The voltage is slowly raised to 4.0 V while the patient is monitored for adverse motor movements and sensory changes. As amplitude is slowly increased, the patient state is evaluated. For upper limb pain care is taken to note the location of the electrode contacts intraoperatively that are over the painful area as determined by motor mapping. The electrode configuration is designed to use the contact that is centered over the area of highest motor response at the lowest stimulation level found intraoperatively. Stimulation amplitudes are started at 2.5 V, but in some cases have gone up to >4.0 V. Pulse widths are started at 210 μs, but in most cases in which changes were needed they have been dropped to 90 μs. In patients for whom MCS for pain seems to have benefit, the frequency has been kept at 130 Hz, but in other cases a variety of frequencies are tried.

Modeling technique 

To understand what effect MCS may be having on both the overall motor system and on processing of pain in general, we have performed studies in computational modeling. The software (UNCuS) for this approach was created and refined by one of the authors [J.E.A.] and is described elsewhere.71, 72 It allows us to simulate general biophysics of each neuron on a conductance level, connect thousands of such independent cells together, each with electrotonic dendritic processing, and keep track of each synaptic event and time step on a 0.25 ms timescale.

We created a basal ganglia model to study DBS effects on the subthalamic nucleus (STN)73 and have now used this to study the effects of cortical stimulation on movement disorders. In the STN DBS model, we had a very simplified cortical representation. In studies on the mechanisms of MCS for pain syndromes, we created a more sophisticated cortical model that includes intrinsic six-layer cortical anatomic connectivity, and used this in a separate model to examine cortico–cortical and cortico–thalamic processing, to explore how pain sensation within S1 may be affected by M1 stimulation.

The circuitry representing cortex is shown by the schematic in Figure 3. Anatomical findings to support these connections in appropriate ratios and synaptic configurations are found in the work of several groups.74, 75, 76, 77, 78, 79 Although either more or less complexity within cortical regions can be either supported or criticized, we tried to account for a representation of the majority if not all of the major known connections and ratios. In addition, it was necessary to use subthreshold levels of stimulation in cortex. This is determined clinically by trying to see driving of motor units on EMG with 10 Hz stimuli.

View Large Image Download to PowerPoint

FIG. 3. Schematic showing all of the interconnections of two regions of cortex simulated using the UNCuS software program71, 72 and regions of sensory and motor thalamus used to model pain processing and the effect of motor cortex stimulation (MCS) on S1. Arrows in red indicate inhibitory synapses on their targets; those in green, excitatory. All six layers are represented in M1 and S1, but have slightly different ratios of connectivities, based on supporting literature. The rendering at bottom shows the regions of cortex intended to be represented: essentially a core of the surface of the gyrus of M1 and S1. Interconnections between M1 and S1 are intended to be predominantly the U fibers and are also shown in the lower diagram.

We used a similar thresholding to determine the appropriate level of MCS in the model by using the amplitude of stimulation in the electrodes that would change the firing rate in <50% of the cortical cells. Although many of these cells were closer to their baseline thresholds than without stimulation, they were not altered significantly in their baseline firing rates by the stimulation alone. However, and perhaps more importantly, it is likely that surface stimulation fields from a typical paddle-type lead are affecting horizontally oriented fibers,80 rather than affecting vertically oriented fibers exiting the cortex. Many of the horizontal fibers are from the stellate category of interneurons within cortical layers 1, 2, and 3, and these are predominantly inhibitory in nature.81 Moreover, the Toronto group70 has found that typical renderings of MCS stimulate both types of fibers, rather than directly hyperpolarizing them, as has been at times speculated. Combining these observations led us to apply MCS in the model—at subthreshold levels for motor unit driving—only to the stellate interneurons of M1 and not to the pyramidal cells (FIG. 3), which are excitatory and are more likely to be output cells from the cortex overall.

Clinical 

Overall, we have operated on 15 patients using MCS as a treatment (Table 2). Specifically in the PD group of four patients, there was one infection 3 months after implantation and one intraoperative seizure that was generated during motor mapping. There were no other complications from the device, the implant procedure, or postoperative programming of the device. One patient developed a superficial infection of the extension wire site and required removal of the device. He had improvement of ∼50% over baseline in dyskinesias, but regressed significantly when the device was removed (this was the unilaterally implanted patient). After reimplantation of the device several months later, he again improved over preoperative baseline by nearly 50% again.

TABLE 2. Our Series of MCS Patients

	Patient ID	Diagnosis	Time Since Surgery,⁎ mo	Age, years	Outcome
	P1	PSM	39	64	Fair
	M1	ET	36	56	Poor
	P2	PSM	36	61	Good
	P3	NPSP	35	41	Fair
	P4	PSM	34	64	Fair
	P5	PSP	30	58	Good
	M2	PD	27	73	Fair to Poor
	P6	PSM	23	49	Poor
	M3	PD	21	67	Fair to Poor
	M4	ET	21	89	Poor
	M5	PD	20	45	Good to Fair
	P7	PSP	17	64	Poor
	M6	PD	16	66	Fair
	M7	CBD	17	62	Poor
	P8	CRP	1	45	—
	Mean		25	60
	Std Dev		10	12
	Max		39	89
	Min		1	41

CBD = cortico–basal degeneration; CRP = complex regional pain; ET = essential tremor; NPSP = non-poststroke pain; PD = Parkinson’s disease; PSM = poststroke mixed pain–movement disorders; PSP = poststroke pain.

⁎ As of the time of this report.

Two patients had a 20–30% (M2 and M3) reduction in medication requirements at 3 months, but patient M3 started to need more medication at the 6- and 12-month time points. One patient (M5), however, had a 37% increased medication requirement. One patient had a return of dyskinesias after 3 months (M5); this patient had initially improved, but then the amplitude of stimulation was reduced and the benefit was lost. Once the stimulation amplitude was increased again, the patient’s dyskinesias were again better controlled.

We examined scores for UPDRS Part III (motor disability) over the initial 12 months of the study (FIG. 4; Table 3). Although significant benefit in these motor scores can be discerned within the initial 6 months, the trend was returning to baseline by 12 months, suggesting that there is only a transient benefit with MCS, at least using parameters described herein. TABLE 3, TABLE 4 show how the UPDRS total scores changed over time in the ON medication–ON stimulation state for the UPDRS part III scores and the total UPDRS. Table 5 shows the changes in medication over the study period.

View Large Image Download to PowerPoint

FIG. 4. UPDRS part III (on stimulation, off medication) over the initial 12 months in four patients with Parkinson’s disease. Averaging at each time point shows that benefit was seen over the first 6 months, but then returned nearly to baseline by 12 months.

TABLE 3. UPDRS Part III (Motor) Scores During ON Medication and ON Stimulation

	Patient	UPDRS III Score
		Pre ON	1 mo ON	3 mo ON	6 mo ON	1 yr ON	New 6 mo ON
	MCS1	19	26	25	(Inf.)	(Inf.)	23
	MCS2	32	47	13	18	35	—
	MCS3	24.5	12	18	27	5	—
	MCS4	11	—	OFF	OFF	—	—

Inf. = infection.

TABLE 4. UPDRS Total Scores During ON Medication and ON Stimulation

	Patient	UPDRS Total Score
		Pre OFF	Pre ON	1 mo ON	3 mo ON	6 mo ON	1 yr ON	New 6 mo ON
	MCS1	54	46	52	47	(Inf.)	(Inf.)	51
	MCS2	71	57	79	42	94	62	—
	MCS3	70.5	53.5	23	32	54	37	—
	MCS4	29	22		OFF	OFF	—	—

Inf. = infection.

TABLE 5. Levodopa Daily Milliequivalents

	Patient	Daily Levodopa Milliequivalents
		Pre	1 mo	3 mo	6 mo	1 yr	New 6 mo
	MCS1	2140	2140	1874	(Inf.)	pre 1044; post 1320	1520
	MCS2	1204	754	654	1054	1054	—
	MCS3	820	850	1000	910	950	—
	MCS4	1090	1320	1090	1090	—	—

Inf. = infection; pre = pre new implant; post = post new implant.

In our series of pain patients, results are even less consistent. With the eight patients for whom pain played a prominent role, five cases were poststroke pain, one was atypical head pain due to a traumatic fall (this patient failed both spinal and peripheral nerve stimulation), one was atypical head pain, and one was due to complex regional pain syndrome secondary to a brachial plexus avulsion most likely (this patient had failed both a spinal cord and peripheral nerve stimulator).

Of the five poststroke pain patients, three cases were mixed pain and movement disorders. One of those three (P7) received no benefit from the stimulation for either the pain or the movement disorder. The other two (P4 and P5) received varying degrees of benefit. Patient P5 has good control of pain in her upper extremities and face, with less pain control in her leg region and minimal control of a third-limb sensation. Patient P4 has fair control of her pain and fair control of an internal tremor. The patient with atypical head pain due to a fall has had no benefit from the MCS for pain in his contralateral scalp, but is getting benefit in his ipsilateral scalp, curiously.

Programming of our pain patients includes a wide range of stimulation parameters with voltage varying from 2.2 V up to 4.2 V, pulse width varying from 120 μs up to 330 μs, and frequency varying from 50 Hz up to 130 Hz. All patients start with monopolar settings, but often bipolar settings are used with more anodal contacts.

Benefits in our other MCS patients remain mixed. Anecdotally, however, there have clearly been modest successes in our hands. Recently, for example, we have had three patients who each called in to say they had noticed worsening of their symptoms over the previous month or so and that maybe they needed reprogramming: one patient with PD (M5) and two with poststroke upper extremity pain (P5) or mixed (P1). At the programming visit, it was found that each of them had completely depleted the implantable pulse generator batteries; all three regained the previous benefit level after the batteries were replaced. None had a hand controller device to check battery levels, nor had they been seen in the clinic recently, where their batteries might have been checked and noted to be waning.

Modeling 

Significant findings in our original modeling of STN DBS included appreciating that DBS had effects on globus pallidus pars interna (GPi) and thalamic cells only at frequencies of >130 Hz (the level at which therapeutic effects are seen), that subsequent variance of firing in GPi and thalamus was more notable than average firing rate changes, and that power spectral changes in the thalamic tremor frequency range were significantly diminished with STN DBS.73 Changes in the GPi with MCS in the model, in terms of firing rate and variance of the interspike interval, are shown in Figure 5. Note that both STN DBS and MCS significantly alter the variance in GPi cells. Such changes were unexpected, but were found in our STN DBS modeling73 and are also found within the same basal ganglia model using MCS.

View Large Image Download to PowerPoint

FIG. 5. Variance (Var) and coefficient of variation (CV) of the interspike interval (ISI) within the globus pallidus pars interna (GPi) region of our basal ganglia model with MCS at increasing amplitudes. The variance decreases significantly. This is similar to what we had found previously with stimulation of the subthalamic nucleus in the same model.

A more sophisticated model of the cortical anatomy and physiology was used in a thalamo–cortical circuit, to explore how pain signaling might be affected by MCS subthreshold stimulation. Because only the interneurons, the majority of which are inhibitory, had MCS applied, it is likely that this region of M1 would exhibit a general decrease in firing rate and, by definition, would not reveal motor unit driving by EMG. Figure 6 shows how topographically arrayed synaptic input from sensory thalamus of an 80-Hz regional stimulation (graphically shown by cumulative activity; FIG. 6a) to the receiving layer 4 pyramidal cells of S1 would also lead to decreased activity in these layer 4 pyramidal cells overall (FIG. 7). Likely this is related to the U fiber (subcortical arcuate fasiculi) excitation from M1 being inhibited locally by stimulation of the horizontally oriented axons of the associative inhibitory cells. The lack of even baseline excitation from M1 layer 5 pyramidal cells, which send output to layer 4 pyramidal cells in S1, would create a lower firing rate in those targets. Figure 6b shows the cumulative change in firing rate graphed to reveal location within this cortical region. Note that there is a convergence and focus of activity going from sensory thalamus (FIG. 6a) to S1 pyramidal cells (FIG. 6b). The significant change in activity is centered only within the topographically oriented region which was coding for the painful stimulus coming from sensory thalamus.

View Large Image Download to PowerPoint

FIG. 6. (A) Cumulative activity over 20 s of simulated time with a pulsatile 80-Hz stimulus in a discrete region of the topographically oriented sensory thalamus. The peak activity over time (in red) represents the activity coming from this area and coding for a more specific region of the body. The cells synapse directly on layer 4 pyramidal cells in the S1 representation of the model (FIG. 3). (B) Cumulative activity difference before and after MCS in the model. Note that MCS inhibits the activity level only in the discrete region of S1 that would represent coding for the same specific body region, topographically driving it from the sensory thalamus.

View Large Image Download to PowerPoint

FIG. 7. Mean firing rates in the S1 cell population in the model, over 20 s of simulated time, for topographically active (A) and nonactive (B) pain area in S1. MCS is activated only between 12 and 16 s and is off the rest of the time; the decrease in overall firing rate during this time is significant (panel a). In the absence of stimulation (panel b), there is no change in S1 activity throughout the entire simulation.