Control of sway using vibrotactile feedback of body

313

Journal of Vestibular Research 15 (2005) 313–325 IOS Press

Control of sway using vibrotactile feedback of body tilt in patients with moderate and severe postural control deficits C. Wall IIIa,∗ and E. Kentalaa,b a

Department of Otology and Laryngology, Harvard Medical School, and Jenks Vestibular Diagnostic Laboratory Massachusetts Eye & Ear Infirmary Boston, USA b Helsinki University Hospital, Department of Otolaryngology, Helsinki, Finland

Received 17 December 2003 Accepted 1 November 2005

Abstract. We evaluated the effect of the vibrotactile display of body tilt upon the postural stability of vestibulopathic subjects during standing. Two groups were studied: those with moderate and with severe deficits as defined by postural stability test scores. They were studied under conditions of distorted sensory input, and during anterior-posterior perturbations. Seventeen subjects, with uni- or bilateral vestibular deficits, as determined by electronystagmography and vertical axis rotation, were tested using Equitest computerized dynamic posturography (CDP). Based on their performance on the CDP they were divided into two groups having either moderate (nine subjects) or severe (eight subjects) postural control deficits. Their anterior-posterior (A/P) body motion at the waist was measured with a micromechanical rate gyroscope and a linear accelerometer. The resulting tilt estimate was displayed by a vibrotactile array attached to the torso. The vibration served as a tilt feedback to the subject. The subject’s performance was evaluated using the root-mean-square (RMS) of both the A/P body motion and center-of-pressure (CoP) estimates. Sensory distortions were introduced using the Equitest Sensory Organization Tests (SOT). These tests are designed to distort A/P sensory inputs while standing. The SOT 5 distorts proprioceptive information about ankle joint movement, while the subject stands eyes-closed on a moving support platform that measures foot pressure. The SOT 6 adds distorted visual information about body movement instead of testing with eyes closed. Perturbations were introduced using the Equitest Motor Control Tests (MCT). These move the support platform forward or backward with small, medium and large displacements in the horizontal plane while measuring subjects’ foot pressure responses. We used the medium and large backward tests. Vibrotactile display of body tilt reduced the subjects’ A/P sway and improved their balance. The finding was more evident for those subjects with severe deficits than those moderate ones. This trend was found for both SOT 5 and 6, as well as the medium and large MCT. Additionally, during the MCT, the peak deflection and mean recovery time also decreased significantly. Keywords: Balance rehabilitation, postural control, balance prosthesis

1. Introduction There are several groups of subjects who might benefit from some form of vestibular or balance prosthesis [1]. These include subjects with bilateral vestibu∗ Corresponding author: Conrad Wall, III, Jenks Vestibular Diagnostic Lab, 243 Charles Street, Boston, MA 02114, USA. Tel.: +1 617 573 4153; Fax: +1 617 573 4154; E-mail: [email protected].

lar hypofunction who cannot adequately rely on motion cues from proprioception and vision during their activities of daily living, those with unilateral vestibular hypofunction whose central nervous systems have a tonic imbalance to which they cannot adjust, those with fluctuating vestibular function who cannot adapt to the fluctuations, and those in the elderly population who are prone to fall. Direct electric stimulation of the vestibular portion of the eighth nerve, and sensory

ISSN 0957-4271/05/$17.00 © 2005 – IOS Press and the authors. All rights reserved

314

C. Wall III and E. Kentala / Control of sway using vibrotactile feedback of body tilt in patients

substitution comprise two means for providing missing motion cues to the central nervous system. Vibrotactile display of body tilt is one way of implementing sensory substitution. This method, while lacking the potential of stimulating the vestibulo-ocular and vestibulo-spinal reflex directly, may have potential advantages over direct electric stimulation in some cases. These advantages include use as a temporary balance aid for use after ablative inner ear surgery, and for fall prevention in the elderly. Also, non-invasive devices carry no risk of surgery, and may be significantly less expensive. Ideally, balance prostheses should be evaluated during activities of daily living (ADL) or in situations designed to mimic these activities. Standing while other sensory information, such as vision or proprioception, is distorted or denied, and recovering from perturbations while standing are two components of ADL which may require accurate motion information to maintain stability. Thus, they provide fairly realistic, but relatively simple paradigms for the initial evaluation of vibrotactile display of body tilt for use as a possible balance prosthesis. The experimental sensory substitution apparatus used in this study, which we will refer to as a “VibroTactile Tilt Feedback (VTTF) device” from now on has been described in detail elsewhere [2,3]. Briefly, it consists of a motion-sensing device on the subject’s back that sends body tilt information to a digital processor, which in turn delivers electrical signals to an array of small vibrotactile stimulators called tactors. The idea of the prosthesis is based on that of sensory substitution, where information normally coming in one “sensory channel” is coded for delivery to a different channel [4]. In our scheme, information normally delivered to the vestibular system is provided to the somatosensory system. We selected the somatosensory system as the substitution channel because it had already proven useful by Angus Rupert and his colleagues in transmitting orientation information to pilots and scuba divers [5,6], and because we did not want to increase the information load on senses like vision and hearing that are needed for ADLs [7]. The number of tactors, their positioning, and coding scheme used in this study were chosen based on earlier results [3,8]. The device developed by Rupert for use by US. Navy pilots differs somewhat from the device we use in that we use the physical position of the tactors to display the magnitude of body movement, while the Navy device reserves the physical location of the tactors on the body to display the direction of the movement, or other orientation information.

One aim of this study was to compare the performance of two different groups of balance impaired subjects having either mild or severe postural control deficits as objectively determined by computerized dynamic posturography scores. A second aim was to characterize the ability of the VTTF device to stabilize the subjects during abrupt disturbances. Would the effect of VTTF be similar for both groups, but just be matter of degree, or would these two groups display two distinct kinds of responses? The answer to this question helps to define the class of subjects for which VTTF has potential utility as a balance aid. In a preliminary report, we have previously shown that a subset of the severely impaired subjects could reduce their tilt using vibrotactile feedback during postural control conditions that deny or distort other sensory inputs [9].

2. Material and methods 2.1. Subjects Recruitment and testing. Seventeen subjects with varying amounts of balance dysfunction were recruited to the study either from a tertiary referral center (Massachusetts Eye & Ear Infirmary) or from the Acoustic Neuroma Association. All subjects gave informed consent before entering the study, which was approved by the Mass. Eye & Ear Infirmary Human Studies Committee. Subjects had a vestibular test battery to evaluate their vestibular and balance function. This battery included: an ANSI Standard 3.45 electronystagmography test battery (Dix-Hallpike tests, binaural, bithermal closed loop water irrigations recorded eyes open in the dark), sinusoidal vertical axis rotation (SVAR), (0.01 Hz to 1.0 Hz 50 deg/sec recorded eyes open in the dark using electrooculography) visual vestibular interactions (0.05 Hz, 50 deg/sec sinusoids), and computerized dynamic posturography (CDP). SVAR responses were characterized by fitting the combined gain and phase results using a frequency domain linear systems model. This yields an overall estimate for the gain of the aVOR over the 0.2 Hz to 1 Hz range (Midrange gain), where it is relatively “flat”, and yields an estimate of the long VOR time constant. This process is described detail for our lab in [10]. Table 1 gives test data, demographics and other clinical information. Subjects were divided into two groups based on the score on the CDP sensory organization test (SOT) 5 and 6. All trials of the SOT 5 and 6 runs were averaged together to obtain the score given in the column labeled “SOT 5 & 6

C. Wall III and E. Kentala / Control of sway using vibrotactile feedback of body tilt in patients

score” in Table 1. Group 1, consisting of nine subjects, 3 female (average group age 56.8 years, range 27-68, standard deviation ±14.6) had an average score above 45 (mean 57, range 48–71, standard deviation ±7.4). This group consisted mainly of vestibular schwannoma patients that had compensated for their unilateral loss. Group 2 consisted of eight subjects, 5 female (average age 48.9, range 23–61, standard deviation ±12.2). The scores for Group 2’s SOT 5 and 6 performance was less than 45 (mean 18, range 0–37, standard deviation ±14.7). These patients were referred to the study by clinicians due to acute severe balance problems. Group 2 also had two patients that hadn’t recovered from their vestibular schwannoma surgery and were still experiencing balance problems. The mean SOT score from all SOT 1 – SOT 6 trials was 72 for Group 1 subjects and 54 for Group 2 (The average score for normal is above 72). The motor control test (MCT) score was 147 for Group 1 and 152 for Group 2 (The average score for normal is below 158). For MTC ranges and standard deviations, please see Table 1. Limits for these test scores are given in the legend of Table 1. Self reports. All subjects were interviewed for their medical history. The impact of their balance impairment on their daily life was evaluated by the American Academy of Otology – Head and Neck Surgery (AAOHNS) function level scale designed for Meniere’s disease [10]. The scale was used because it is well described in the clinical literature, and because it is gives a good estimate of function level. Scores by subject are shown in Table 1. Most subjects scored themselves in a range of 1 to 3 out of 6 levels, where level 6 represents complete disability. Four subjects scored themselves as level 1: activities not effected by dizziness. Four scored themselves as level 2: cessation of some activities during spells of dizziness, but otherwise no change in any plans or activities to accommodate dizziness. Seven subjects scored themselves as level 3: having to change some plans or activities to allow for dizziness. The only striking difference in scores between groups is the presence of two scores of 5 in Group 2 subjects. With this exception it would appear that both groups of subjects were had compensted for any dificits moderately well. 2.2. Vibrotactile Tilt Feedback (VTTF) device The VTTF device contains three major elements: an inertial instrumentation package, a Macintosh PowerBook, with analog and digital I/O, and a vibrotactile display. This “proof-of-concept” device is limited to

315

only one direction of motion. The inertial instrumentation consists of a micromechanical linear accelerometer and rate gyroscope (gyro) package that is secured on the subject’s lower back (Fig. 1A). The accelerometer and gyro provide linear acceleration and angular rate information, respectively, about the subject’s anterioposterior (A/P) body tilt. A tilt estimate is calculated by adding the integrated, high pass filtered gyro signal to the low pass filtered accelerometer signal, as shown in Fig. 1(B). If the gyro signal alone were integrated to estimate body tilt, the output would have a error, called “drift” due to the integration of an unwanted noise or “bias” term. If the accelerometer signal alone were used as a tilt estimate, then it would also have an error because it would measure both gravitational force and linear acceleration of the body at right angles to the rostal-caudal axis. The drift from the integrated gyro signal can be filtered off using a very low frequency high pass filter, but this also removes the static and low frequency body tilt angle information. This latter information can be re-introduced by lowpass filtering the accelerometer signal to remove most of the linear acceleration component, while retaining the low frequency and static gravity components. These two signals, when thus combined, have been shown to give a very accurate and nearly drift free estimate of orientation of the instrument package with respect to gravity vertical over the frequency range of normal body tilts [2]. This process is shown diagrammatically in Fig. 1(B). We will refer to this processed signal from the instrument package when attached to the body as “Tilt”. The vibrotactile display consists of small tactile vibrators called tactors, which are placed in an elastic belt that ring the subject’s torso just above the waist. There are two parallel columns of three tactors each that are located on the lower portion of the subject’s stomach, one on each side of the navel, and two more identical columns on the subject’s back, one on each side of the vertebrae. Distance between each of the two parallel columns is 3 inches. The spacing (gap) between each row of tactors is 1.5 inches. Each tactor is 0.5 inches high. Thus, columns of tactors are used to display direction, while rows of tactors are used to display magnitude. The lowest row of tactors was positioned approximately at the level on the torso at which a subjects normally wore their belts. We did not position the level of the tactors to correspond to a particular vertebra, since different subjects have different “belt lines” relative to the vertebrae, and our primary objective was to have a comfortable, functional,

Tactors

C. Wall III and E. Kentala / Control of sway using vibrotactile feedback of body tilt in patients

3 2 1

3 2 1

Gyro

s s+0.18 Highpass filter 0.03 Hz

1 s Integrator

Sensors

Accelerometer

Sum

0.18 s+0.18 Lowpass filter 0 - 0.03 Hz

K 1/g

62.8 s+62.8 Lowpass filter 0 - 10 Hz

Activated Tactor Level

316

3 2 1 None Forward

Backward

0 1 3 5 Tilt Signal (PD)

A

B

C

Fig. 1. A) Diagram of the VTTF device. The instrumentation package, consisting of one rate gyroscope (gyro) and one linear accelerometer, is mounted on the small of the subject’s back. The gyro measures angular motion in the anterioposterior forward-backward) direction, while the accelerometer measures linear acceleration at right angles to the subjects’ rostral-caudal axis, and also a component of the gravity vector. Tactile viberators (tactors) are mounted in dual columns on the front and on the rear of the subject’s torso. The lowest of three rows of tactors, labeled “1” are mounted at the level of the subject’s beltline. B) Tilt estimation scheme. The angular rate output from the Gyro is first high pass filtered at 0.03 Hz to eliminate an undesirable slowly varying noise or “bias” term, which, if put through the integrator, would cause the angular position output to drift. The output of the integrator thus does not contain a “DC” term because it has been filtered out. This “DC” term can be replaced by using the output of the linear accelerometer, whose signal is first low pass filtered so that it mainly measures the input due to gravity and not much of the more rapidly changing tangential acceleration of the body. Finally, the two signals are added, then filtered again over a bandwidth in which most body motion occurs. C) Schematic presentation of the firing pattern of the tactors. Only one row of tactors is activated any time that the subjects is outside of a pre-determined “dead zone”. A small positive tilt signal activates the lowest row of tactors in the front of the subject, labeled “1” on the subject’s front torso in Fig. 1(A). Progressively larger tilt signals activate row 2, and so on. Typically, rows 1, 2, and 3 are activated at inputs of 1 deg, 3 deg, and 5 deg, respectively. Similarly, a small negative signal outside the dead zone activates the lowest row of tactors on the rear of the subject.

mounting site. Stomach-mounted tactors are used to display anterior tilt direction, while back-mounted tactors are used to display posterior direction of tilt. Each row of tactors represents a tilt angle range that is determined by the signal processor using the step-like coding scheme shown in Fig. 1(C). The exact switching point from one step or level to another is determined for each subject based upon their maximum range of static body tilt while they are asked to control posture as if they were an inverted pendulum that pivots about the ankle (see Section 2.4). The control or input signal to the step-wise coding scheme is the sum of the tilt estimate (Tilt) and its first derivative. Thus, this signal is that of a proportional plus derivative (PD) controller. This processing step was done because it is well known from control theory that a single inverted pendulum cannot be stabilized using just proportional control. In addition, our preliminary experiments (manuscript in preparation) showed about a 20% reduction in subject body motion in vestibulopathic subjects when we provided vibrotactile display of body motion that was based upon PD control as compared to proportional control alone. Thus, the input signal to the vibrotactile display represents both the subject’s estimated tilt angle and tilt

velocity. The firing range of tactors is set individually for each subject, based on their maximum forward and backward tilt range. To account for normal body motion, no tactors fire in a defined dead zone that is one degree in both the anterior and posterior directions. With increasing body tilt the active tactor moves from bottom to top on the tilting side in a stepwise manner. Only the location of the tactor was used to display subject tilt. Stimulus intensity, as defined by amplitude and frequency of the input signal to the tactors, was kept constant over subjects and was set at the highest current rating (200 mA) of the tactile vibrators, as recommended by the manufacturer (Tactaid, Cambridge, MA). Frequency was always 250 Hz. With these parameters, all subjects were able to perceive the tactor vibrations, and no subjects reported aversive effects. 2.3. Balance training and measurement equipment A modified NeuroCom Equitest device (NeuroCom, International) was used for this study. The device has a platform designed to pitch the feet toes-up and toesdown about the ankle joints, and to provide linear anterioposterior (A/P) perturbations. There is also a movable visual surround that pivots on a line co-linear with

1 1 1 1 1 1 1 1 1

2 2 2 2 2 2 2 2

1 2 3 4 6 8 11 13 14 Mean S.D.

5 7 9 10 12 15 16 17 Mean S.D.

61 51 56 56 57 43 44 23 48.9 12.2

27 64 56 58 56 58 40 68 31 50.9 14.6

Age

female female female male male female male female

female male male female male female male male male

Gender

bilateral no tumor right no tumor no tumor no tumor no tumor no tumor

right left right right left right left left left

Side of tumor

264 N/A 30 N/A N/A N/A N/A N/A

49 71 55 52 48 59 55 46

0 37 34 19 5 25 23 0 17.9 14.7

161 136 145 178 135 156 136 131

100 2 100 32 45 8 28 25

0.86 0.61 0.93 0.60 0.89 0.64 0.36 0.32

N/A 9.9 8.2 8.6 1.5 3.9 0.2 1.3

N/A 5.21E-05 0.0402 0.0054 8.12E-13 1.53E-07 1.92E-28 6.07E-16

Time since SOT SOT 5 & MCT RVR VOR VOR Normal surgery score 6 score score midrange time VOR (mo) gain constant probability 24 72 67 139 100 0.68 9.3 0.0115 87 79 71 146 100 0.54 9.5 0.00216 107 73 55 148 100 0.79 7.8 0.00928 68 72 55 150 100 0.96 9.3 0.345 24 70 53 155 100 0.86 6.5 0.0057 98 69 52 155 100 0.79 3.3 1.26E-10 13 70 57 129 100 0.71 8.0 5.73E-10 16 68 48 146 100 N/A N/A N/A 5 71 53 154 100 0.76 6.3 4.16E-08 56.8 7.4 N/A UVH UVH UVH (