Tremor suppression using functional electrical stimulation - IEEE Xplore

7 downloads 0 Views 97KB Size Report
Deborah M. Gillard, Tracy Cameron, Arthur. Prochazka, and Michel J. A. Gauthier. Abstract— In this study, we compared digital and analog versions.
IEEE TRANSACTIONS ON REHABILITATION ENGINEERING, VOL. 7, NO. 3, SEPTEMBER 1999

385

Communications Tremor Suppression Using Functional Electrical Stimulation: A Comparison Between Digital and Analog Controllers

Index Terms— Electrical simulation, functional electrical simulation (FES), tremor suppression.

authors designed a controller and implemented it using analog circuitry. The controller comprised a closed- loop system that filtered the sensed tremor signal and used it to stimulate the tremorogenic muscles out of phase, thereby attenuating the tremor without significantly affecting concomitant voluntary movement. Patients diagnosed with three different types of tremor were tested, with varying degrees of attenuation achieved. With the feasibility of the device established, a digital version of the analog controller was implemented [17]. More recent improvements include the use of an accelerometer as the motion sensor and changes to the feedback filter that increase the system’s ability to attenuate tremor. The present study is a comparison between the analog and digital controller’s effectiveness at attenuating tremor. Tests on three subjects with no neurological impairment performing fast tremorlike movements showed an average attenuation of 83% for the digital system and 69% for the analog system with neither system appreciably attenuating slower voluntary movements. Further tests of the digital controller on three parkinsonian patients provided an average tremor attenuation of 85% as compared to 61% with the analog controller of the original study.

I. INTRODUCTION

II. METHODS

Tremor is the most common of all motor disorders, disabling over three million people in North America alone [1]. Until recently, only three methods of treatment existed: medication, inertial loading, and viscous loading. Of the three main types of tremor, approximately 50% do not respond to medication [2]. Viscous and inertial loads provide resistance to motion which increases with frequency and the square of frequency, respectively. Inertial loads are often used to selectively reduce tremor [3] but they also attenuate functional movements to some extent and weigh the limb down, causing muscle fatigue. A combination of inertial and viscous loading has been investigated using feedback-controlled actuators [4], [5] with one prototype showing promise for table-top activities [6]. More recently, work has been done on a tremor suppression system for use in either a teleoperated device or a joystick-based system [7], [8]. An adaptive noise cancelling system is under development for filtering out tremor in microsurgery and in computer input applications [9], [10]. In the last two years, deep brain stimulators have been used in people with Parkinsonian and essential tremors, in some cases with spectacular success [11], [12] but the costs are high, the technique involves invasive brain surgery and the pros and cons in the long term have yet to be evaluated. Tremorous movement differs from most intended movement in its frequency content. Typically, cerebellar tremor is in the 3–5 Hz range, essential tremor is in the 6–11 Hz range, and Parkinsonian tremor is in the 3–7 Hz range [13], while most of the energy of intended movement is in the 0.5–2.5 Hz range [14]. Prochazka et al. [15], [16] explored feedback-controlled functional electrical stimulation (FES) as a tool for tremor suppression. The

A. Experimental Approach The responses to amplitude-modulated trains of electrical stimuli of the wrist flexor and extensor muscles using either the analog filter or the digital filter were recorded for three subjects with no neurological impairment (ages 35–51) and three Parkinson’s patients (ages 63–71). All participants gave their written consent to the experiments, in accordance with the requirements of the University of Alberta Human Ethics Committee and the Declaration of Helsinki. Each subject was seated in a chair with the elbow joint at 90 flexion. Hand and forearm motion was independently monitored using a three-dimensional motion analysis system (Skill Technologies, Inc., Phoenix, AZ 85014 USA). Two 2 2 2.5 cm sensors were attached to the subject’s skin, one on the posterior surface of the lower forearm just proximal to the wrist crease and the other on the midpoint of the third metacarpal segment. The indifferent (anodic) electrode (self-adhesive surface electrodes, ConMed Versa-stim: 45 2 90 mm) was placed on the skin proximal to the wrist crease on the anterior forearm. The stimulating (cathodic) electrodes (45 2 45 mm) were placed on the skin over the motor points of the wrist or finger flexor and extensor muscles, depending on which tremor was targeted. Subjects with no neurological impairment were trained to mimic a 4 Hz tremor by concentrating on an audible 4 Hz signal.

Deborah M. Gillard, Tracy Cameron, Arthur Prochazka, and Michel J. A. Gauthier

Abstract— In this study, we compared digital and analog versions of a functional electrical stimulator designed to suppress tremor. The device was based on a closed-loop control system designed to attenuate movements in the tremor frequency range, without significantly affecting slower, voluntary movements. Testing of the digital filter was done on three patients with Parkinsonian tremor and the results compared to those of a functional electrical stimulation device based on an analog filter evaluated in a previous study. Additional testing of both the analog and digital filters was done on three subjects with no neurological impairment performing tremor-like movements and slow voluntary movements. We found that the digital controller provided a mean attenuation of 84%, compared to 65% for the analog controller.

Manuscript received August 20, 1998; revised December 21, 1998 and June 3, 1999. This work was supported by the Canadian MRC, the Neuroscience Partners Fund, and the Alberta Heritage Foundation for Medical Research. The authors are with the Division of Neuroscience, University of Alberta, Edmonton, Alta. T6G 2S2, Canada. Publisher Item Identifier S 1063-6528(99)07100-1.

B. Filter Design The overall open-loop frequency response was designed for zero phase at the tremor frequency and a maximal gain over the tremor frequency range to attenuate the tremor. The gain was minimized elsewhere to allow voluntary movement at lower frequencies to remain unrestricted and to prevent painful stimulation in the event of instability at higher frequencies. The digital filter was designed using Matlab (The Mathworks, Inc., Natick, MA 01760 USA) software to analyze the frequency response characteristics of the closed-loop, allowing an optimization of the

1063–6528/99$10.00  1999 IEEE

386

IEEE TRANSACTIONS ON REHABILITATION ENGINEERING, VOL. 7, NO. 3, SEPTEMBER 1999

Fig. 1. Block diagram of the setup for the closed-loop experiment used for subjects with no neurological impairment. The switch allowed the selection of either controller without the subject knowing which was in use.

design criteria. Two digital filters were designed: one was tuned to 4 Hz and the other to 6 Hz, while maintaining the same overall closed-loop response.

TABLE I SUMMARY OF TRIAL RESULTS FOR THE ANALOG AND DIGITAL FILTER SYSTEMS ON SUBJECTS WITH NO NEUROLOGICAL IMPAIRMENT

C. Displacement Sensor To allow unrestricted movement of the arm, an unobtrusive but reliable joint displacement sensor was needed. The original analog filter used a small cantilever strain gauge with silastic tubing spanning the joint to provide the displacement signal. The relationship between joint angle and output signal was linear for only a small range of motion. In addition, the silastic tubing tended to “ring” if it caught on clothing or external objects and from turning the system on in mid-cycle, and the resulting signals were propagated through the high-pass digital filter, thus introducing a high-frequency oscillation through the feedback system. The result was large, uncomfortable fluctuations in the amplitude of stimulation. For these reasons, a 10g accelerometer (Silicon Designs, Inc., Issaquah, WA 98027-5344 USA, 9 2 9 2 3 mm) was chosen to be the sensor for the digital system although a digital filter was also designed for use with the strain gauge to verify that any differences in performance between analog and digital systems were not due to the sensors but rather the filters themselves.

TABLE II SUMMARY OF TRIAL RESULTS FOR THE DIGITAL FILTER SYSTEM USED ON PARKINSONIAN PATIENTS

D. Closed-Loop Experiments 1) Subjects with No Neurological Impairment: To obtain a direct comparison between the analog and digital systems, both the strain gauge and the accelerometer were mounted on the subject’s hand. This allowed us to switch between the corresponding analog or digital filters without the subject knowing which system was in use. The stimulator output stage was the same in each case (Fig. 1). Ten trials were run on each subject, five with the analog filter and five with the digital filter. Each trial lasted 30 s with stimulation being applied for 10 s in the middle. Threshold stimulation amplitude was set for each active electrode by increasing the stimulation strength until movement due to muscle contraction was first detected. In the analog system, open-loop gain was set by increasing the stimulation gain until the closed-loop system became unstable and then decreasing the gain by about 50% (6 dB). The digital system remained stable even for large gains, so we simply increased the gain until the maximum amount of tremor attenuation was achieved without discomfort. Audible clicks at a repetition rate of 4 Hz were used to give subjects with no neurological impairment a target frequency for their tremor movements, and a 0.75 Hz audible signal was used for the timing of the slow tracking movements. 2) Parkinsonian Patients: The accelerometer was mounted on each subject’s finger or hand depending on whether the targeted tremor was in the fingers or about the wrist. The 4 Hz digital filter was used for the two patients with tremors at 3.8 and 4.2 Hz, and the

Fig. 2. Sequential fast Fourier spectra of a Parkinsonian patient’s tremor movements before, during (stim on), and after stimulation with the digital FES controller.

6 Hz digital filter was used for the patient with a 6 Hz tremor. Ten trials were run on each patient, each trial lasting 30 s. Trials where a sustained tremor was absent before the application of stimulation were rejected from the data analysis. Thresholds and open-loop gains

IEEE TRANSACTIONS ON REHABILITATION ENGINEERING, VOL. 7, NO. 3, SEPTEMBER 1999

387

(a)

(b)

Fig. 3. (a) Fast and (b) slow wrist movements monitored in a subject with no neurological impairment with and without stimulation using the digital controller. Note that the slow movements were unaffected by stimulation.

were set as described above. Wrist or finger displacement data were analyzed with a fast Fourier transform (FFT) using Matlab software. III. RESULTS A. Displacement Sensor The transfer function characteristics of the accelerometer and its circuitry were determined empirically by displacing the accelerometer at a range of frequencies with an electromagnetic length servo. A least-squares routine was used to fit an equation to the curves. The transfer function includes an ac-coupling term and a second order low-pass filter with a cutoff frequency of 6 Hz and is given by 3

s

: (s + 6:45)(s2 + 77:7s + 1508)

(1)

B. Filter Design for Closed-Loop Stimulation The digital filters were designed in the s-domain with the following transfer functions for the 4 and 6 Hz filters, respectively 2 2 2 2 2 (s + 11s + 40)(s + 23s + 158) (s + 40s + 3198) 2 2 2 2 (s + 34s + 355)(s + 45s + 632) (s + 57s + 987)(s + 15)2

system. The transfer function is given by (4) below and has a gain margin of 19.0 dB at 9.1 Hz which gives a theoretical attenuation of 90% for a 4 Hz tremor. This filter was realized digitally and tested with the length gauge sensor. It performed considerably better than the original analog filter, and was comparable to the accelerometer system, although the comfort level was slightly compromised due to the noise on the length gauge. The filters were transformed into the discrete time-domain using the bilinear transform with a step size of 10 ms. Instability occurs when the phase of a closed-loop, negative feedback system is ±180 and its magnitude is greater than one. The gain margin is a measure of the amount of gain that can be added to the closed-loop system before it becomes unstable. A closedloop system with transfer function G=(1 + GH ) has an open-loop transfer function GH, where G and H represent transfer functions in the forward and feedback paths, respectively. A negative feedback loop attenuates a disturbance signal by a factor of (1 + jGH j)01 . The gain margin of the analog system was 8.1 dB at 7.9 Hz, allowing a theoretical maximal attenuation of 72%. The gain margin of the digital system was 18.7 dB at 10.0 Hz, allowing a maximal attenuation of 90%.

(2)

2 3 2 2 (s + 6s + 10) (s + 53s + 5685) : 2 3 2 2 (s + 68s + 1421) (s + 79s + 1934)(s + 163s + 6672)

(3) It could be argued that any differences between the digital and analog systems is due to the use of different sensors. For this purpose only, a third digital filter was designed for use with the length gauge, but with the same open-loop characteristics as the accelerometer

C. Closed-Loop Trials The results of the trials on the subjects with no neurological impairment and the Parkinsonian patients are summarized in Tables I and II, respectively. The average attenuation for the subjects with no neurological impairment was 68.9% for the analog system, and 83.1% for the digital system. A representative spectrogram illustrating tremor attenuation in a Parkinsonian subject is shown in Fig. 2. The average attenuation

2 3 2 2 (s + s + 0:4) (s + 37s + 2852) : 2 2 2 2 (s + 34s + 355)(s + 45s + 632)(s + 57s + 987)(s + 85s + 2221)(s + 50)2

(4)

388

IEEE TRANSACTIONS ON REHABILITATION ENGINEERING, VOL. 7, NO. 3, SEPTEMBER 1999

electrodes and a thin battery-driven controller-stimulator. The small and portable design now allows the device to be worn in the form of a cuff about the forearm which is slim enough to be worn discretely under a shirt sleeve (Fig. 4). New filter components can be downloaded to the microprocessor quickly and easily. In principle this will allow one of several filters optimized for specific frequencies in the tremor range to be selected for an individual patient. Finally, two limitations with the strain gauge length sensor: a small linear range and susceptibility to mechanical transients, have been minimized with the use of an accelerometer as a sensor. ACKNOWLEDGMENT The authors would like to thank A. Denington for his technical assistance and R. Sustrik for his work on graphical figures. REFERENCES Fig. 4. Tremor suppression system currently being used in an outpatient trial in five Parkinsonian patients with simple flexion-extension hand or finger tremors.

of tremor in the Parkinson’s patients was 84.5% using the digital controller with a standard error or ±2.2. The original analog study had an average attenuation of 60.8 ± 0.93 for the Parkinson’s patients (P < 0:0001, Mann–Whitney rank sum test). Time-domain plots of a subject with no neurological impairment performing a tremor-like movement at 4 Hz tremor and a slow tracking movement at 0.75 Hz with and without stimulation using the digital system are shown in Fig. 3. IV. DISCUSSION In this study we compared a digital version of an FES tremorsuppression stimulator with the original analog prototype. Each system was designed to counteract the tremor by stimulating the wrist or finger flexor and extensor muscles to contract out of phase with their tremorogenic activation pattern. Using a mathematical description of the load-moving muscles derived in a previous study, we designed and implemented two digital filters to attenuate 4–6 Hz tremor without affecting slower, voluntary movements. Each system suppressed tremor within 10% of its maximal theoretical attenuation. A number of factors may have contributed to the variation between predicted and actual tremor attenuation. The use of a relatively simple linear model of the load-moving properties of muscles cannot provide a precise prediction of such a complex system. More realistic nonlinear models, such as that explored by Prochazka et al. [18] may offer improved performance, and are currently under development. It might be expected that stimulation would alter the amplitude of the internal signal used to produce a tremor-like movement by the subjects with no neurological impairment. However, the fact that the tremor-like movement returned to its prestimulated amplitude within a few cycles after stimulation was turned off implies that the internal signal remained constant. In fact, for the digital control system, the difference in average tremor attenuation between the Parkinson’s patients and the subjects with no neurological impairment was less than 2%, giving further credibility to the use of their data. Thus, both the digital and analog tremor suppression systems performed within an acceptable range relative to the theoretical design, and a considerable improvement in suppression was noted with the new digital filter. A. Practical Application As with the previous study, we have shown the feasibility of a new digital tremor suppression system using functional electrical stimulation. The device consists of a forearm cuff with integral

[1] W. J. Weiner and A. E. Lang, Movement Disorders: A Comprehensive Survey. New York: Futura, 1989, pp. 221–256. [2] L. J. Findley and R. Capildeo, Movement Disorders: Tremor. London, U.K.: Macmillan, 1984. [3] R. Hewer, R. Cooper, and M. H. Morgan, “An investigation into the value of treating intention tremor by weighting the affected limb,” Brain, vol. 95, pp. 579–590, 1972. [4] B. L. Morrice, W. J. Becker, J. A. Hoffer, and R. G. Lee, “Manual tracking performance in patients with cerebellar incoordination: Effects of mechanical loading,” Can. J. Neurol. Sci., vol. 17, pp. 275–285, 1990. [5] B. L. Morrice, R. G. Lee, W. J. Becker, D. G. White and J. A. Hoffer, “Viscous loading improves manual tracking accuracy in humans with cerebellar incoordination,” Soc. Neurosci. Abstracts, vol. 13, p. 1700, 1987. [6] M. J. Rosen, A. S. Arnold, I. J. Baiges, M. L. Aisen, and S. R. Eglowstein, “Design of a controlled-energy-dissipation orthosis (CEDO) for functional suppression of intention tremors,” J. Rehab. Res. Dev., vol. 32, no. 1, pp. 1–16, Feb. 1995. [7] J. G. Gonzalez, E. A. Heredia, T. Rahman, K. Barner, and G. R. Arce, “Optimal digital filtering for tremor suppression,” review in progress, IEEE Trans. Biomed. Eng., submitted for publication. [8] J. G. Gonzalez, E. Heredia, T. Rahman, K. Barner, and G. Arce, “A customized optimal filter for eliminating operator’s tremor,” in Proc. SPIE’s Int. Symp. Intel. Syst. Adv. Manuf.—Telemanipulator and Telepresence Tech. II, Philadelphia, PA, Oct. 1995. [9] C. N. Riviere, R. S. Rader, and N. V. Thakor, “Adaptive canceling of physiological tremor for improved precision in microsurgery,” IEEE Trans. Biomed. Eng., vol. 45, pp. 839–846, July 1998. [10] C. N. Riviere and N. V. Thakor, “Modeling and canceling tremor in human-machine interfaces,” IEEE Eng. Med. Biol., vol. 15, pp. 29–36, May/June 1996. [11] A. L. Benabid, P. Pollak, D. Gao, D. Hoffmann, P. Limousin, E. Gay, I. Payen, and A. Benazzouz, “Chronic electrical stimulation of the ventralis intermedius nucleus of the thalamus as a treatment of movement disorders,” J. Neurosurg., vol. 84, pp. 203–214, 1996. [12] P. Limousin, P. Pollak, A. Benazzouz, D. Hoffmann, J. Le Bas, E. Broussolle, J. E. Perret, and A. Benabid, “Effect on Parkinsonian signs and symptoms of bilateral subthalamic nucleus stimulation,” The Lancet, vol. 345, pp. 91–94, 1995. [13] J. Jankovic and S. Fahn, “Physiologic and pathologic tremors—Diagnosis, mechanism, and management,” Ann. Internal Med., vol. 93, no. 3, pp. 460–465, 1980. [14] K. A. Mann, F. L. Werner, and A. K. Palmer, “Frequency spectrum analysis of wrist motion for activities of daily living,” J. Orthopaedic Res., vol. 7, pp. 304–306, 1989. [15] M. Javidan, J. Elek, and A. Prochazka, “Attenuation of pathological tremors by functional electrical stimulation—II: Clinical evaluation,” Ann. Biomed. Eng., vol. 20, pp. 225–236, 1992. [16] A. Prochazka, J. Elek, and M. Javidan, “Attenuation of pathological tremors by functional electrical stimulation—I: Method,” Ann. Biomed. Eng., vol. 20, pp. 225–236, 1992. [17] J. J. Law, “Reduction of pathological tremor by functional electrical stimulation using digital feedback control,” M.Sc. thesis, Dep. Elec. Eng., Univ. Alberta, Edmonton, Alta., Canada, 1995. [18] A. Prochazka, D. Gillard, and D. J. Bennett, “Implications of positive force feedback in the control of movement,” J. Neurophysiol., vol. 77, pp. 3237–3251, 1997.