AU2016269469B2 - Non-regular electrical stimulation patterns for treating neurological disorders - Google Patents

Abstract

Systems and methods for stimulation of neurological tissue generate stimulation trains with temporal patterns of stimulation, in which the interval between electrical pulses (the inter-pulse intervals) changes or varies over time. Compared to conventional continuous, high rate pulse trains having regular (ie, constant) inter-pulse intervals, the non-regular (ie, not constant) pulse patterns or trains that embody features of the invention provide a lower average frequency. Non-Regular 33 Pulse Non-Regular Pulse Train (Singlet) Pulse Train is Repeated I n-let (n=2) (Doublet) Kn-let looms Inter-Pulse Inter-Pulse Minimum Singlet Interval Inter-Pulse Interval (Non-Regular) Singlet (Non-Regular) Inter-Pulse Interval Fig. 4 Interval Between n-lets (Non-Regular) Minimum Inter-Purse Non-Regular SinletPulse Tri Interval Non-Regular is Repeae PulsePls TrintralOm n-oulets ~~~~Inter-Pulse(NnRglr SingletBewn Interval -e (Non-Regular) Inter-Pulse Fg Interval (Non-Regular)

Description

NGN - REGULAR. ELECTRICAL STIMULATION PATTERNS FOR TREATING NEUROLOGICAI. DISORDERS Related Application T|iis application claims the benefit of Unite# States Provisional Patent Application Serial No. 61/102,575, filed October 3, 2008,- and entitled "Stimulation Patterns For Treating Neurological Disorders Via Deep Brain Stimulation," which is incorporated herein by reference.

Field of the Invention

This invention relates to systems and methods for stimulating nerves in animals, including humans. Baelsspound of the Invention

Deep Brain Stimulation (DBS) has been found to be successful in treating a variety of brain-controlled disorders, including movement disorders. Generally, such treatment involves placement of a DBS type lead into a targeted region of the brain through a burr hole drilled in the patient's skull, and the application of appropriate stimulation through the lead to the targeted region.

Presently, in DBS, beneficial (symptom”relieving) effects are observed primarily at high stimulation frequencies above 100 Kz that are delivered in stimulation patterns or trains in which the interval between electrical pulses {the inter-pulse intervals) is constant aver time. Ike trace of a conventional stimulation train for DBS is shown in Fig. 2. The beneficial effects of DBS on motor symptoms are only observed at high frequencies, while low frequency stifloatation may exacerbate symptoms See Benabid et al. 1931, Limousin et al. 1995 . Thai ami. c DBS at less "than or equal to 50 Hz increases tremor in patients with essential t: retnor. See Kune el et al. 3006. Similarly, SO Hr DBS produces tremor in pain patients receiving simulation of the ventral posterior medial nucleus of the thalamus (VPM)f but t he tremor disappears when the frequency is increased. See Const ant oyhnnis 2004 . Likewise, DBS of the subthalamic nucleus (STN) at 10 Hz worsens akinesia in patients with Parkinson's disease {PD) , while DBS at 130 Hz leads to significant improvetttSht in motor function See Timmermann et al. 2004, Fogelson et al. 2005. Similarly., stimulation of the globus pallidus (GP) at or above 13 0 Hz significantly improves dystonia, whereas stimulation at either' 5 or 50 Hz leads to significant worsening, See Kupsch et al. 2003 .

Model studies also indicate that the masking of pathological burst activity occurs only with sufficiently: high stimulation frequencies. See Grill et al. 2004, Figure 1. Responsiveness of tremor to changes in DBS amplitude and frequeue/ are strongly correlated, with the ability of applied stimuli to mask neuronal bursting. See Kuncel et al. 2007, Figure 2.

Although effective, conventional high frequency stimulation generates stronger side-effects than low frequency stimulation, and the therapeutic window between the voltage that generates the desired clinical effect(s) and the voltage that generates undesired side effects decreases with increasing frequency. Precise lead placement therefore becomes important. Further, high stimulation: frequencies increase power consumption. The need for higher f requeueies and increased power consumption shortens the useful lifetime and/or increases the physical size of battery-powered implantable pulse generators. The need for higher frequencies and increased power consumption requires a larger battery size, and frequent charging of the: battery, M the battery is rechargeable,

Summary of the Invention -

The invention provides strr.ru 1 a ti on patterns or trains with different temporal patterns of stimulation, than conventional stimulation trains. The invention also provides methodologies to identify and characterize stimulation patterns or trains that produce desired relief of symptoms, while reducing the average stimulation frequency.

According to one aspect of the invention, the intervals between stimulation pulses in a pulse pattern or train (in shorthand celled. ."the inter-pulse intervals") is not constant over time, but changes or varies over time. These patterns or trains are consequently called in shorthand "non-regular According to this aspect of the Invention, the non -regular (i.e., not constant) pulse patterns or trains provide a lower average frequency for a given pul se pattern or train, compared to conventional continuous, high rets pulse trains having regular {i.e,, constant) inter-pulse intervals, Having a lower average frequency, the non-regular stimulus patterns or trains make possible an increase in the efficacy of stimulation by reducing the intensity of side effects; by increasing the: dynamic range between the onset of the desired clinical effect(s) and side effects (and thereby reducing sensitivity to the position of the lead electrode) ; and by decreasing power consumption, thereby providing a longer useful battery life and/or a smaller implantable pulse generator, allowing battery size reduction and/or, for rechargeable batteries, longer intervals between recharging.

The non-regular stimulation patterns or trains can be readily applied to deep brain stimulation, to treat a variety of neurological disorders, such as Parkinson's disease, movement disorders, epilepsy, and psychiatric disorders such, as obsessive-compulsion disorder and depression. The non-regular stimulation patterns or trains can also be readily applied to other classes electrical stimulation of the nervous system including, but not limited to, cortical stimulation, spinal cord stimulation, and peripheral nerve stimulation iincluding sensory and motor), to provide the attendant benefits' described above and to treat diseases such as but not limited to Parkinson's Disease, Essential Tremor, Movement Disorders, Dystonia, Epilepsy, Pain, psychiatric disorders such as Obsessive Compulsive Disorder, Depression, and Tourefcte's Symdrome.

According to another aspect of the invention, systems and methodologies make it possible to determine the effects of the temporal pattern: of DBS on simulated and measured neuronal activity, as 'well as motor symptoms in both animals and humans. The methodologles make possible the qualitative determination of. the temporal features of stimulation trains.

The systems and methodologies described herein employ a genet ic algorithm, coupled to a computational model of DBS of the STS, to develop non-regular patterns of stimulation that produced efficacy {as measured by a low error function, E) at lower stimulation frequencies, F. The error function, E, is a quantitative measure from the model which assesses how faithfully the thalamus transmitted motor commands that are generated by inputs from the cortex. A very high correlation exists between E and symptoms in persons with PD, and therefore E is a valid predictor for the efficacy of a stimulation train in relieving symptoms (see Dorvai et al., 2007).

Previews efforts (see Peng ec al. 2007) sought to design stimulation trains that minimized the total current injection. The systems and methodologies disclosed herein include an objective function that maximizes therapeutic benefit (by minimizing the error function) and improves stimulation efficiency (by reducing the stimulation frequency) ., using a model of the STN that reproduces the frequency tuning of symptom reduction that has been documented clinically. In contrast, Lae Feng et al. model showed, incorrectly, symptom reduction With regular, low frequency stimulation. Ίhe inventors have identified novel non-regular temporal patterns of stimulation, while Feng et al, identified regular low frequency (- 10 Hz) trains that previous clinical work has demonstrated to be ineffective,

Brief Pescripfciqq g£ the Bracings

Fig. 1 is an anatomic view of a system for stimulating tissue of the central nervous system that includes an lead implanted in brain tissue coupled to a pulse generator that is programmed to provide non-regular (i.e., not constant) pulse patterns or trains, in 'which the interval between electrical pulses (the inter-pulse intervals) changes or varies over time.

Fig. 2 is a diagrammatic trace that shows a conventional regular high frequency stimulation train, in which the interval between electrical pulses (the inter* -pulse intervals) is constant.

Fig. 3 is a diagrammatic trace showing a representative example of a repeating non-regular pulse pattern or train in which the inter-pulse intervals are linearly cyclically ramped over tiffig.

Figs> 4 and f> are diagrammatic traces stowdng other representative examples of repeating non-regular pulse patterns or trains comprising within, a single pulse train, a combination of single pulses (singlets) and embedded multiple pulse groups (η··lets) , with non-regular ixrter-pulse intervals between singlets and n-letg as well as non-regular inter-pulse intervals within the multiple pul.o n-lets.

Description of the Preferred Embodiments

Fig. 1 is a system 10 for stimulating tissue of the central nervous system. The system includes a 1 ead 12 placed in a desired position in contact with central nervous system tissue. In the illustrated embodiment, the lead 12 is implanted in a region of the brain, such as the thalamus, subthalamus, or globus pallidus for the purpose of deep bra in. stimulation. However, it should be understood, the lead 12 could be implanted in, on, or near the spinal cord; or in, on, or near a peripheral nerve (sensory or motor} for the purpose of selective stimulation to achieve a therapeutic purpose.

The distal end of the lead 12 carries one or more electrodes 14 to apply electrical pulses to the targeted tissue region. The electrical pulses are supplied by a pulse generator IS coupled to the lead 12.

In the i.i lust rated embodiment, the pulse generator 16 is implanted in & suitable location remote from the lead 12, e. g. , in the shoulder region. It should be appreciated, however, that the pulse generator 16 could be placed in other regions of the body or externally.

When implanted, the case of the pulse generator can serve as a reference or return electrode. Alternatively,-the lead 12 can include a reference or return electrode (comprising a bi-polar arrangement) , or a separate reference or return electrode can be implanted or attached elsewhere on the body (comprising a mono-polar arrangement).

The pulse, generator 16 includes an on-board, programmable microprocessor 18, which carries embedded code. The code expresses pre-programmed rules or algorithms under which a desired electrical stimulation waveform pattern or train is generated and distributed to the electrode (s) 14 on the lead 12. According to these programmed rules, the pulse generator 16 directs the prescribed stimulation waveform patterns or trains through the lead 12 to the electrode(s) 14, which serve to selectively stimulate the targeted tissue region. The cods is preprogrammed by a clinician to achieve the particular physiologic response desired.

In the illustrated embodiment, an on-board battery 20 supplies power to the microprocessor 18. Currently, batteries 2 0 must be replaced every 1 to .9 years, depending on the stimulation parameters needed to treat a disorder» when the. battery life ends, the replacement of batteries requires another invasive surgical procedure to gain access to the implanted pulse generator. As will be described, the system 10 make s possible, among its several benefits, an increase in battery life.

The stimulation waveform pattern or train generated by the pulse generator differs from convention pulse patterns or trains in that the waveform comprises repeating non-regular (i.e., not constant) pulse patterns or trains, in which the interval between electrical pulses (the inter-pulse intervals or IPI) changes or varies over time. Examples of these repeating non-regular pulse patterns or trains are shown in Figs. 3 to 5. Compared to conventional pulse trains having regular (i.e., constant) inter-pulse intervals (as shown in Fig. 2), the non-regular {i.e., not constant) pulse patterns or trains provide a lower average frequency for a given pulse pattern or train, where the average frequency £or a given pulse train (expressed in herbs or Hz) is defined as the sura of the inter-pulse intervals for the pulse train in seconds (ΣΙΡ3) divided by the number of pulses <n) in the given pulse train, or (Σ1ΡΙ)/η. A lower average frequency makes possible a reduction in the intensity of side effects, as well as an increase in the dynamic range between the onset of the desired clinical effect(s) and side effects, thereby increasing the clinical efficacy and reducing sensitivity to the position of the electrode(s). A lower average frequency brought about by a non-regular pulse pattern or train also leads to a decrease in power consumption, thereby prolonging battery life and reducing battery size.

The repeating non-regular (i.e., not consfeaast) pulse patterns or brains. can take a variety of different forms. For example, as will be described in greater detail later, the inter-pulse intervals can be linearly cyclically ramped over time in non-regular temporal patterns (growing larger and/or smaller.or a combination of each over tirae); or be periodically embedded in nonregular temporal patterns comprising clusters or groups of multiple pulses (called n-lets), wherein n is two or more. For example, when n==?2, the n-let can be called a doublet; when n«3, the n-let can be called a triplet ; when n~4, the n··let can be called a quadlet; and so on. The repeating non-regular pulse patterns or trains can comprise combinations of single pulses (called singlets) spaced apart by varying non-regular inter-pulse intervals and n-lets interspersed among the singlets, the n-lets themselves being spaced apart by varying non-regular inter-pulse intervals both between adjacent n-lets and between the n pulses embedded in the n-let. If desired, the non-regularity of the pulse pattern or train can be accompanied by concomitant changes in waveform and/or amplitude, and/or duration in each pulse pattern or train or in successive pulse patterns or trains,

Sash pulse comprising a singlet or imbedded in an n~ let in a given train comprises a waveform that can be isosopfeasic, biphasic, or' multiphssic. East waveform possesses a given amplitude (expressed, e.g., in amperes) that can, fey way of example, range from 10 pa (E"a) to 10 ma (E"·’) . The amplitude of a given phase in a waveform can foe the same or differ among the phases. Each waveform also possesses a duration (expressed, e.g., in seconds) that can, by way of example, range from 10 ps (Ε"®) to 2 ms {£~3)'. The duration of the phases in a given waveform can iikewise foe the same or different. Xt is emphasised that all numerical values expressed herein are given by way of example only. They can be varied, increased or decreased, according to the clinical objectives.

When applied in deep brain stimulation, it is believed that repeating stimulation patterns or trains applied with non-regular inter-pulse intervals can regularize:: the output of disordered neuronal firing, to thereby prevent the generation and propagation of bursting activity with a lower ave rage stimulation frequency than required with convent i onal constant frequency trains, i.e., with a lower average frequency than about 100 Ha. .

Fig. 3 shows a representative example of a repeating non-regular pulse pattern or train in which the inter-pulse Intervals are linearly cyclically ramped over time. As shown in: Fig, 3s, the pulse pattern or train includes singlet pulses (singlets) spaced apart by progressively increasing; inter:-pulse intervals: providing a decrease in frequency over time, e.g., having an initial instantaneous frequency of 140 He, decreasing with doubling inter-pulse intervals::, to a final instantaneous frequency of 40 Hz. The inter pul se intervals can vary within a specified range selected based upon clinical objections, e .g. ,. not to exceed 25 ms, on not to exceed 100 fils, or not to exceed: :2 00 ms, fen take into account burst responses and subsequent disruption of thalaaiie fidelity. ) . The non·regular pulse trains repeat the®seIves for a clinically appropriate period of time. As shown in Pig. 3, the first pulse train comprises progress i ve1y increasing inter-pulse intervals from smallest to largest, followed immediately by another essentially identical second pulse train comprising progressively increasing inter-pulse intervals from smallest to largest, followed immediately by an essentially identical third pulse train, and so on. Therefore, between successive pulse trains, there is an instantaneous change from the largest inter-pulse interval (at the end of one train) to the smallest, inter-pulse interval {at the beginning of the next successive train) . The train shown in Fig. 3 has an average frequency of 85 Hz and is highly non -regular,. with a coefficient of variation (CV) of about 0.5. As. is demonstrated in the following Example (Batch 3} , the increased efficiency of the pulse train shown in Fig. 3 (due to the lower average frequency) also can provide greater efficacy, as compared to a constant 100 Hz pulse pattern.

The train shown in Fig. 3 exploits the dynamics of burst generation in thalamic neurons. The early high frequency phase of the train masks intrinsic activity in subthalamic nucleus (STW) neurons, and the inter-pulse interval increases reduce the average frequency. A family of trains can be provided by varying the initial frequency, final frequency, and rate of change within the train, with the objective to prevent thalamic bursting with a lower average stimulation frequency than required with constant frequency trains.

Figs. 4 and 5 show other representative examples of repeat i ng non - regular pulse patterns or trains. The pulse trains irt Figs. 4 and 5 comprise within, a single pulse train, a combination of single pulses (singlets)· and embedded multiple pulse groups (n-lets), with non-regular inter-pulse intervals between siaglets and n-lets, as well as non-regular inter-pulse intervals within the n-lets themselves. The non-regular pulse trains repeat themselves for a clinically appropriate period of time.

The non-regular pulse train can be character!zed as comprising one or more singlets spaced apart by a minimum inter;-pulse singlet interval and one or more n- let s comprising, for each n-let, two or more pulses spaced apart by an inter-pulse interval {called the "li-let inter-pulse interval"} that is leas than the minimum singlet inter-pulse interval. The n-let intsr-pplse interval can itself '.'ary within the train, as cam the interval .between successive n-lets or a successive n-lets and singlets. The non-regular pulse trains comprising singlets and n-lets repeat themselves for a clinically appropriate period of time.

In Fig. 4, each pulse train comprises four singlets in succession (with non-regular inter-pulse intervals there between) ,- followed by four doublets in succession (with non · regular .inter-douhlet pulse intervals there between and non-regular inter-pulse intervals within each n-let); followed by a singlet, three doublets, and a singlet {with non-·regular inter- pulse intervals there between and non-regular inter-guise intervals within each n-let) . The temporal pattern of this pulse train repeats itself in succession, for a clinically appropriate period of time. The non-regular temporal pulse pattern Shown in Fig, 4 has an average frequency of 67.82 Hz without loss of efficacy, as is demonstrated in the following Example,

Batch 17.

In Fig, 5, each pulse train comprises four singlets in succession (with non-regular inter-pulse intervals there between); followed by three doublets in succession (with non-regular inter-doublet pulse intervals there between and non-regular inter-pulse intervals within each n~let), The temporal pattern of this pulse train repeats itself in succession for a clinically appropriate per'iod of time. The non-regular temporal pulse pattern shown in Fig. 5 has an average frequency of 37.62 Hr without loss of efficacy, as is demonstrated in the following Example, Batch IS .

The following Example illustrates a representative methodology for dewlqpihg and identifying candidate nonregular stimulation trains as shown in Figs, 3 to 5 that achieve comparable or better efficacy at a lower average frequency (i.e. , more efficiency) than constant inter-pulse interval trains,

EX&amp;5SPLE

Computational models of thalamic DBS (McIntyre et al, 2004, Birdno, 2003) and subthalamic DBS (Rubin and

Terraan, 2004} can be used with genetic-a.lgo.rithm-based optimization (Davis, 19.91} (GA) to design non-regular stimulation patterns os' trains that produce desired relief of symptoms with a lower average stimulation frequency than regular, high-rate stimulation. McIntyre et al. 2004, Birdno, 2009; Rubin and Terrnan, 2004; and Davis, 1991 are incorporated herein by reference.

In the GA implementation, the stimulus train (pattern) is the chromosome of the organism, and each gene in the chromosome is the IPI between two successive pulses in the train. The implementation can start, e.g., with trains of 21 pulses (20 genes) yielding a train length of ~400 ms (at average frequency of 50 Hz) , and the 6 s trains required for stimulation are built by serial conoatenation of 1:5 identical pulse plains. The process can start with an initial population of, e.g., 50 organisms, constituted of random IPIfs drawn from a uniform distribution. At each step {generation} of the GA, the fitness of each pulse train is evaluated using either the TC or basal ganglia network model (identified above) and calculating a cost function, C, From each generation, the 10 best stimulus trains (lowest C) are selected, to be carried forward to the next generation, They will also be combined (mated) and random variations (mutations) introduced into the 40 offspring, yielding 50 trains in each generation. This process assures that the best stimulation trains (traits) are carried through to the next generation, while avoiding local minima (i.e., mating and mutations preserve genetic diversity). See Grefenstette 1986. The GA continues through successive generations until the median and minimum values of the cost function reach a plateau, and this will yield candidate trains.

The objective is to find patterns of non-constant inter-pulse interval deep brain stimulation trains that provide advantageous results, as defined by low frequency and low error rate. An error function is desirably created that, assigns the output of each temporal pattern of stimulation a specific error fraction (E) based on how the voltage output of the thalamic cells correspond to the timing of the input stimulus. Using this error fraction, a cost function CO is desirably created to minimise both frequency and error fraction, according to representative equation € * W*E + K*£, where; C is the cost., Έ is the error fraction, f is the; average; frequency of the temporal pattern of stimulation:, w is an appropriate weighting factor for the error: function, and: K is an appropriate weighting factor for the frequency. The weighting factors W and K allow quantitative different! a. t i on between ef:: icacy (E3 and efficiency {£} to generate patterns of nor··· constant inter-pulse internal deep brain stimulation trains that provide advantageous results with lower average frequencies, compared to conventional constant frequency pulse trains.

With this cost, function, the voltage output of several candidate temporal patterns of stimulation can be evaluated and the cost calculated. Temporal patterns of stimulation with a low cost: can then be used to create new1 temporal patterns or similar features in in attempt to achieve even lower costs. In this way, new temporal patterns of stimulation can be -bred" for a set number of generations and the best temporal patterns of stimulation of each batch recorded.

Several batches of the genetic algorithm yields useful results in that they achieve lower costs than the corresponding constant frequency DBS waveforms;. Some batches can be run ip an attempt to find especially low frequency temporal patterns of stimulation, by changing the cost function to weight frequency more heavily, or vice versa (i.e., by changing W and/or K). These batches can also yield lower cost results than the constant -frequency waveforms.

By way of example, a total of 14 batches of the genetic algorithm were run and evaluated with various cost functions and modified initial parameters.

Before the trials were run, a baseline was established by running constant-frequency patterns of stimulation through the model and analyzing the associated error fractions (Example Figure 1) . As can be seen from Example Figure 1, the healthy, condition produced a low error fraction of 0.1 while the Parkinsonian condition without DBS yielded a higher error fraction of 0.5. From these results, constant high-frequency patterns of stimulation ranging from 100-200 Hz gave near perfect results. Novel non-constant temporal patterns of stimulation would then be considered advantageous if they showed error fractions very close to 0.1 with average frequencies less than 100-200 Hz,.

Example Figure 1

The first set of batches was run by minimizing only the error fraction <E). Thus, the associated cost fnnctiph was simply C -·- E, The results are summarized according to average frequency and error fraction (Example Table i). The associated inter-pulse intervals (IPI's) can be seen iri Example Figure 2, Batch 3 outputted an error fraction 0.054. Another interesting feature is that the IPX's in Batch 3 gradually increased until about 40 msec, and then repeated itself.-. This provides support that ramp trains are advantageous,. The trace shown in Fig, 3 generally incorporates the temporal features of Batch 3, .

The remaining batches yielded error fractions higher than 0.1 and were no better than the 150 Hz constant-frequency case.

Example Table 1; Error Fraction Only, C ~ £

Error Fraction Only, C ~ E

Example Figure 2

Because many batches were yielding error fractions above 0.1 {healthy condition}, and only a small window of error fraction less than the 150 Hz DBS case would be useful, a new cost function was constructed to minimize an alternate, feature of the temporal patterns of stimulation; namely, frequency. This new cost function weighted the error fraction and frequency, yielding the equation C * 10Q0*E + F, where C is cost, E is error fraction, and F is the average frequency of the waveform in Hz, W = 1000, and K=l.

In order to establish a new baseline cost, the constant-frequency patterns of stimulation, were evaluated again according to the new cost function (Example Figure 3) . As can be seen from the graph, the healthy condition reported a cost of .90·. 65 and the Parkinson case with no DBS yielded 505.50, The best constant-frequency pattern of stimulation with the new cost function was the 100 Hz case with a cost of 231.11. This new cost function allowed for a wider rang® of solutions, because a temporal pattern of stimulation would be considered ^useful if it had a cost less than 231.11 but presumably higher than 90.65,

Example Figure 3

The results of the sew cost function can be seen in Example Table 2 and the IPX's visualised in Example Figure 4. The best results were seen in batches 15 and 18, which had the lowest costs. Batch 18 is interesting in that it also exhibits a ramp - like pattern of increasing interpulse intervals. It shows a steadily failing IPX, followed by a sudden rise, and then a quick fail, rise, and fall—almost as if it consists of 3 smaller ramps. The trace shown in Fig. 5 generally incorporates the temporal features of Batch 18 Batch 15 also performed very well, but its qualitative features are more difficult to di scern.

ExampleTable 2; Cost Function, C - 1000*E + F

Cost Function, C = 1000*E + F

Example Figure 4

The advantage of low frequency was emphasized with a new cost function, which weighted frequency more heavily, C = m®0*E + 2*F. Because the frequency of DBS does not affect the healthy condition or the PD with no DBS, these baseline costs stayed the same at 90.65 and 505.50, respectively . The 100 He was again the bast . constant-frequency temporal pattern of stimulation, with a cost of 331.11. The following temporal patterns of stimulation, then, were considered useful if they had low frequencies and hosts less than 331.11 arid greater than 90.65.

The results of the revised coat function can be seen in Example Table 3 and the IPX's visualised in EkampleFigure 5. Of the resulting batches, batch 1? proved most interesting because of its very low average frequency of 67.82 Hz. Even with such a low frequency, it managed to prove better than the 1GG Hs condition-with a reduction in cost of about ID. The waveform of batch 17 is interesting in that it consists of a ramp pattern of decreasing IPX in the first. 100 msec, followed by a continual shift between large IPX and small IPX, The qualitative £eatene of quickly changing between large and small IPX's may prove advantageous. The trace shown in Pig. 4 generally incorporates the temporal features of Batch 17.

ExampleTable 3: Revised Cost Function, Cost » 1000*E + 2*F

Revised Cost Function. C = 1000Έ + 2*F

Example Figure 5

The: most interesting temporal patterns of stimulation in this Example are from hatches 15, 17, and IB. Batch 15 rjroduced a temporal pattern of stimulation with an average frequency of 98 Hs with an error fraction as low as 0.098. Thus, it outperformed the 100 Hz constant-frequency case by managing to lower the error even further at roughly the same frequency. Still, the qualitatively useful features of batch 15 are difficult to discern. Batch 17 was also appealing because of its very low frequency of 67.82. This low frequency was gained at the coat of increased error at 0.253, hut it may nonetheless be useful if emphasis is placed on maintaining low frequency DBS. The qualitative features of batch 17 indicated at first a ramp followed by a continual switching between low and high IPI's. Lastly, batch 18 stood somewhere in the middle with1 a fairly low frequency of 87,62 and low error fraction of 0.116, only marginally higher than the healthy condition of 0.1. The dominant qualitative feature of batch 18fs waveform is that it too shows a ramp nature in that the IPX initially steadily falls, then quickly rises, falls,.: and then rises. The rapid changing between high and low 1PT of batch 17 can be envisioned as a sat of steep ramps. A comparison of Batch 17 (Fig. 4) and Batch IS (Fig. 5) demonstratea how the: balance between efficacy <E) and efficiency (f) in non-regular temporal patterns of stimulation can be purposefully taillored to meet clinical objectives. The systems and methodologies discussed allow changing the cost function by weighting efficacy (E) or frequency (f) more heavily (i.e,, by changing W and/or K), while still yielding temporal patterns of stimulation with lower cost results than the constant-frequency waveforms. Comparing Batch 17 with Batch 18, one sees that the error fraction (Ei {i.e., the efficacy of the temporal pattern) of Batch 17 (0.253) is greater than the error fraction (E) (i.e., the efficacy of the temporal pattern) of Batch 18 (0,116). However, one can also see that the efficiency (i.e., the average frequency) of Batch 17 (67.82 Hz} is lower than the efficiency (i.e., the average frequency) of Batch 18 {81.28 Hr). Through different in terms of efficacy and efficiency, both Batch 17 and Batch 18 have costs better than const ant-frequency temporal patterns.

The non-regular temporal patterns of stimulation generated and disclosed above therefore make possible achieving at least the same or equivalent (and expectedly better) clinical efficacy at a lower average frequency compared to conventional constant-frequency temporal patterns. The lower average frequencies of the non-regular temporal stimulation patterns make possible increases in efficiency and expand the therapeutic window of amplitudes that can he applied to achieve the desired result before side effects are encountered, DBS is a well-established therapy for treatment of movement disorders, hut the lack of under standing of mechanisms of action has limdted full development and optimization of this treatment. Previous studies have focused on DBS-induced increases or decreases in neuronal firing rates in the basal ganglia and thalamus. However, recent data suggest that changes in neuronal firing patterns may be at leant as important as changes in firing rates.

The above described systems and methodologies make it possible to determine the effects of the temporal pattern of DBS on simulated and measured neuronal, activity, as well as motor symptoms in both animals and humans. The methodologies make possible the qualitative and quantitative determination of the temporal features of low frequency stimulation trains that preserve efficacy.

The systems and methodologies described herein provide robust insight into the effects of the temporal patterns of DBS, and thereby illuminate the mechanisms of action. Exploiting this understanding, new temporal patterns of stimulation can be developed, using model-based optimisation, and tested, with the objective and expectation to increase DBS efficacy and increase DBS efficiency by reducing DBS side effects.

The invention provides non-regular stimulation patterns or trains that can create a range of motor effects from exacerbation of symptoms to relief of symptoms. The non-regular stimulation patterns or trains described herein and their testing according to the me t ho do1egy described herein #111 facilitate the select: ion of optimal .surgical targets as well as treatments for new disorders. The non - regular stimulation patterns or trains described herein make possible improved outcomes of DBS by reducing side effects and prolonging battery life*

Literature Citations

Benafoid Mj, Pollak P, Gervason c, Hoffmann B, Gas DM, Homme1 M, Ferret 3E, da Rougemont #(:1991) Long-term Suppression of tremor by chron1c stimulation of the ventral intermediate thalamic nucleus. Lancet. 337-.403-6.

Birdrio MJ "Analyzing the mechanisms of thalamic deep brain stimulation: computational and clinical studies". Ph.D. Dissertation. Department of Biomedical Engineering, Duke Ohiverbity, Durham, NC, USA, August 2009.

Const ant oyanni s C, Kuma r A, SLoessi AJ, Honey CR (2001} Tremor induced by thalamic deep brain stimulation in patients with complex regional facial pain. Mov Disord. 19:933-936.

Davis L (1991) Handbook of genetic algorithms. Van Nostrand Reinhold, NY.

Dorval AD, Kuncel AM, Birdno MJ, Turner DA, Grill WM (2007) Deep brain stimulation alleviates Parkinsonian bradykinesia by regularizing thalamic throughput in human subjects. Society for Neuroscience Abstracts 32.

Feng XJ, Shea-Brown E, Greenwald B, Kosut R, Rabitz H (2007) Optimal deep brain stimulation of the subthalamic nucleus-a computational study. J Comput Neurosoi. 23(3):265-282.

Fogelson K, Kuhn AA, Silberstein P, Limousin PD, Hariz M, Trottenberg T, Kupsch A, Brown P (2005) Frequency dependent effects of subthalamic nucleus stimulation in Parkinson's disease.. Neuroscience Letters 382:5-9.

Grefenstette JJ (1986) Optimization of Control Parameters for Genetic Algorithms. IEEE Transactions on Systems, Man and Cybernetics 16:122-128..

Grill WM, Cooper SE, Montgomery EB {2003) Effect of stimulus wan*®form on tremor suppression and paresthesias evoked by thalamic deep brain, stimulation:. Socety for Beuroseience Abstracts 29.

Kuncel AM, Cooper SE, Montgomery EB, Baker KB, Rezax SR, Grill WM (SOS) Clinical response to varying the stimulus parameters in. deep brain stimulation for essential tremor. Movement Disorders 21(11):1920-1928.

Kupsch A, Klaffke S, Kuhn ΆΑ, Meissner W, Arnold G, Schneider GH, Maier-Hauff K, Trottenberg T (2003) The effects of frequency in pallidal deep· brain stimulation for primary dystonia. J Neurol 250:1201-12 64,

Limousin P, Pollack P, Benazsouz A (199S5 Effect on Parkinsonian signs and symptoms of bilateral stimulation. The Lancet 345:91-95 .

McIntyre CC, .Grill Wft, Gherman DL, Thakor NV. {2(504) Cellular effects of deep brain stimulation: model-based analysis of activation and inhibition. J Neurophysiol 91:1457-1459.

Rubin JE, Terman D (2004) High frequency stimulation of the subthalamic nucleus eliminates:: pathological thalamic r'nythmicity in a computational model, J Compuf Neurosci 16:211-235.

Timmarmann L, Gross J, Dirks M, Volkmann J, Freund HJ, Schnifczler A (2003) The cerebral oscillatory network of parkinsonian resting tremor. Brain, 126:199-212.

Various features of the invention ate set forth: in the following claims.

Claims (13)

1. THE CLAIMS DEFINING THE INVENTION ARE AS FOLLOWS:-

   1. A method of generating a pulse train for achieving neurological stimulation comprising: (i) quantitatively assessing for a given temporal pattern of stimulation having an average frequency (f) an error fraction (E) indicating how voltage output of thalamic cells correspond to timing of inputs, (ii) applying a cost function (C) for the temporal pattern based upon E and f, the cost function weighting E and f to minimize E and fat a clinically beneficial cost (C), (iii) applying the cost function to evaluate the cost of candidate temporal patterns of stimulation, based upon a selected computational model, (iv) selecting, based upon a selected computational model, temporal patterns of stimulation with a clinically beneficial cost, (v) subjecting the selected temporal patterns to a genetic algorithm to create new generations of temporal patterns bred from the selected temporal patterns, and (vi) selecting a pulse train from among the new generations of temporal patterns.
2. 2. A method according to claim 1 applying the pulse train to an animal to achieve neurological stimulation.
3. 3. A method according to claim 1 applying the pulse train to achieve deep brain stimulation in an animal at an average frequency of less than 100 Hz.
4. 4. A temporal pattern of stimulation for application to targeted neurological tissue comprising a pulse train selected according to the method defined in claim 1.
5. 5. A temporal pattern according to claim 4 wherein the selected pulse train comprises an average frequency of less than 100 Hz.
6. 6. A method comprising: (i) providing an error function (E) for a given temporal pattern of stimulation that quantifies how voltage output of thalamic cells correspond to timing of inputs, (ii) providing a cost function (C) expressed as C = W*E + K*f wherein C is the cost, E is the error fraction, f is the average frequency of the temporal pattern waveform, W is an appropriate weighting factor assigned for the error function, and K is an appropriate weighting factor assigned for the frequency, the weighting factors W and K being applied to quantitatively minimize efficacy (E) and efficiency (f) at a given cost, (iii) applying the cost function to evaluate the cost of candidate temporal patterns of stimulation, using a selected computational model, (iv) selecting temporal patterns of stimulation with a low cost based upon the computational model, (v) using a genetic algorithm to create new temporal patterns bred from the selected- temporal patterns, (vi) repeating (iii), (iv), and (v) to bred batches of new temporal patterns of stimulation for a determined number of generations, and (vii) selecting from the batches the best temporal patterns of stimulation in terms of low cost (C), efficacy (based upon E), and efficiency (based upon f).
7. 7. A temporal pattern of stimulation for application to targeted neurological tissue comprising a pulse train selected according to the method defined in claim 6.
8. 8. A temporal pattern of stimulation for application to targeted neurological tissue comprising a repeating succession of non-regular pulse trains, the pulse train being selected according to the method defined in claim 1 or claim 6, each pulse train comprising a plurality of singlet pulses spaced apart by progressively increasing inter-pulse intervals, the pulse train repeating in succession such that, between successive pulse trains, there is an instantaneous change from the largest inter-pulse interval at the end of one pulse train to the smallest inter-pulse interval at the beginning of the next successive pulse train.
9. 9. A temporal pattern of stimulation for application to targeted neurological tissue comprising a repeating succession of non-regular pulse trains, the pulse train being selected according to the method defined in claim 1 or claim 6, each pulse train comprising a plurality of single pulses (singlets) and embedded multiple pulse groups (n-lets), with non-regular inter-pulse intervals between singlets and n-lets, as well as non-regular inter-pulse intervals within then-lets themselves, the pulse train repeating in succession.
10. 10. A temporal pattern of stimulation for application to targeted neurological tissue comprising a repeating succession of non-reguiar pulse trains, the pulse train being selected according to the method defined in claim 1 or claim 6, each pulse train comprising one or more singlets spaced apart by a minimum interpulse singlet interval and one or more n-lets comprising, for each n-let, two or more pulses spaced apart by an n- let inter-pulse interval that is less than the minimum singlet inter-pulse interval, the pulse train repeating in succession.
11. 11. A temporal pattern of stimulation according to claim 8 wherein each pulse train comprises an average frequency of less than 100 Hz.
12. 12. A temporal pattern of stimulation according to claim 9 wherein each pulse train comprises an average frequency of less than 100 Hz.
13. 13. A temporal pattern of stimulation according to claim 10 wherein each pulse train comprises an average frequency of less than 100 Hz.