Balance prostheses for postural control - IEEE ... - BalanceTek

Postural Control

Balance Prostheses for Postural Control ©DIGITAL STOCK

Preventing Falls in the Balance Impaired By Displaying Body-Tilt Information to the Subject via an Array of Tactile Vibrators

CONRAD WALL III AND MARC S. WEINBERG

here is a clear need for a prosthesis that improves postural stability in the balance impaired. Such a device would be used as a temporary aid during recovery from ablative inner-ear surgery and as a permanent prosthesis for those elderly prone to falls. In the article we discuss a single-axis research device we have developed to provide balance-impaired subjects with a noninvasive, vibrotactile display of body tilt, which helps them reduce their body sway during standardized tests.

T

The Need for a Postural Balance Prosthesis Description of Balance System

For those with normal balance function, the inner ear’s vestibular system provides self-motion cues that help stabilize our vision while moving, that enable us to orient ourselves with respect to our surroundings, and that help us stand and walk. Each inner ear can sense, in three dimensions (3-D), angular motion and linear acceleration summed with gravity [41]. The central nervous system (CNS) can process these cues to estimate self motion in six degrees of freedom (dof), three angular and three linear. When either the inner ear, the 8th nerve that connects it to the CNS, or the CNS portion that processes self-motion information malfunctions due to injury, disease, or to prolonged exposure to altered gravity (such as a deep space voyage), this useful information is lost or distorted. This loss of information can force subjects to rely on other cues from vision or proprioception, which may not always be accurate. This reduction of motion cues produces dizziness, blurred vision, inability to orient correctly, and reduced ability to stand or walk. A serious consequence is an increase in the risk of falling in the elderly. Estimated Incidence of Balance Disorders

Over 90 million Americans (more than 40% of the population) will seek medical attention for dizziness at least once in their lifetime. At least 2 million Americans suffer or experience chronic impairment due to dizziness or balance disorders, resulting in annual medical costs exceeding US$1 billion [26]. Depending on the disease, not all patients respond well to existing treatments. As discussed below, many could benefit from a vestibular prosthesis. 84

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

Types and Uses of Balance Prostheses

Biomedical research is presently being conducted on both implantable and nonimplantable balance prostheses. Currently, no patients have had vestibular implants, and only limited laboratory-based experiments have been done using nonimplantable devices. Building upon Suzuki and Cohen [10], [9], [31], [32], Merfeld recently demonstrated responses in monkeys that are first steps toward an implantable prosthesis [19], [23]. An implantable prosthesis offers attractive features including: portability, intuitive operation, replacement of lost natural function, and the ability to work with existing CNS physiology. A potential disadvantage is the surgical risk associated with an implant. A less-invasive means of providing self-motion cues is a nonimplantable prosthesis. Current devices include stimulating the vestibular nerve via surface electrodes and displaying self-motion cues via vibrotactile stimulation or via electric currents applied to the tongue [4], [5] or mastoid bone [30]. Vibrotactile displays have been successfully used by the U.S. Navy to furnish navigational cues that allow blindfolded pilots to control their aircraft [27], [28]—a task with some similarities to postural control. Its use for reducing sway in vestibulopathic subjects will be discussed later in the article. Basic uses for balance prostheses include: 1) a vestibular “pacemaker” to reduce dizziness and imbalance due to abnormal fluctuations in the peripheral vestibular system; 2) permanent replacement of vestibular function; 3) temporary replacement of motion cues that commonly occur following ablative surgery of the inner ear; and 4) vestibular/balance rehabilitation. In terms of postural control, the primary use of a prosthesis would be to prevent falls. Two potential target populations are those recovering from inner-ear ablative surgery and the elderly with impaired balance. Temporary Balance Aid for Use After Ablative Inner-Ear Surgery

The 8th nerve carries hearing and balance information from the inner ear to the brain. The incidence of tumors in the 8th nerve has been estimated at 0.7% based upon thousands of MRI scans [2]. Surgical removal frequently [13] results in balance problems and dizziness while the CNS adapts to this sensory loss. A follow-up study of 141 patients who had acoustic neuroma surgery found that 45% reported balance problems,

0739-5175/03/$17.00©2003IEEE

MARCH/APRIL 2003

The precursor prosthesis is a wearable, distributed, modular design, which builds upon the single-axis research device and consists of a 3-D motion sensor array (three accelerometers and three gyroscopes), a central processor, and vibrotactile stimulators.

while 19% reported dizziness [3]. Five vestibulopathic individuals who had undergone 8th nerve resection were tested with our prosthesis, which is described below. Several commented that a balance prosthesis would have benefited their recovery following surgery [17]. Temporarily using a balance prosthesis could both prevent falls and promote more patient standing and walking activity, and thus hasten recovery. The Need for Fall Prevention

Stability when standing or walking requires functional sensory, motor, and central integration systems. Investigators have shown that the ability to use visual cues to maintain postural stability is significantly reduced in elderly subjects [20]. Akin to presbycusis in hearing, the number of vestibular sensory hair cells decreases markedly with increasing age in healthy normals [22], [34]. At present, there is no direct evidence that this loss causes imbalance, but it is known that the 9.1% incidence rate for dizziness in the geriatric population is about 2.5 times greater that that of the middle-aged population. With 32 million persons in the U.S. over age 65 years and 3 million over 85, this is a significant number of people. One large study of nursing home residents found that the mortality rate for fallers, taken over a one-year period after falling, was more than twice that of a nonfaller control group [15]. The literature suggests it is possible to screen community-based individuals for falling [37], that fallers can be predicted [33], [40], and that many falls can be prevented [1]. Despite this, a recent meta-analysis of fall prevention studies shows a mere 4% decrease of falls in groups getting fall-prevention treatments [14]. A noninvasive balance aid would be an attractive alternative for the aging population. Gyro Input Axis

top computer with analog and digital interfaces, and a vibrotactile display subsystem comprised of drive electronics and small vibrotactile devices called tactors that contact the subject’s skin. The tilt sensor module contains one micro-electrical-mechanical systems (MEMS) gyroscope (gyro), which senses angular rate, and one MEMS linear accelerometer [Figure 1(a)]. The gyro and accelerometer signals are processed to obtain a tilt angle estimate accurate to within 2 milliradians (mrad) over a 0 to 10 Hz bandwidth. This experimental device measures and displays tilt about only one axis. For controlling front-to-back tilt [Figure 1(b)], the accelerometer input axis is nominally horizontal and backward. The gyro’s sensitive axis is horizontal and parallel with the shoulders. The digital controller commands individual tactor amplifiers that drive tactors mounted in columns on the subject’s front and back at 250 Hz, a sensitive frequency for human skin [8], [16]. Control of Body Tilt During Computerized Dynamic Posturography with Vibrotactile Display of Body Tilt

Six vestibulopathic subjects were trained to use the singleaxis device using a standard, rehabilitation training protocol (Balance Master) while they stood on a computer-controlled

Accelerometer Input Axis

Accelerometer Input Axis

Gyro Input Axis

Vibrotactile Display of Body Tilt Reduces Sway in Vestibulopathic Subjects Vibrotactile Research Device

We have developed a single-axis research device to provide subjects with a noninvasive, vibrotactile display of body tilt [17], [21], [35]. The device consists of inertial sensors and their signal conditioning electronics, a lap-

(a)

(b)

Fig. 1. (a) MEMS instrumentation used to estimate body tilt along a single axis. The single-axis tilt module is mounted on the small of the subject’s back. The gyro input axis is sideways relative to the subject, while the accelerometer axis is forward-backward. (b) Stick figure to show orientation of instrumentation axes relative to the body.


MARCH/APRIL 2003

85

The availability within the past five years of good-performance, low-cost, small-size MEMS inertial sensors and advanced electronics has seeded developing the vestibular prosthesis.

posture platform (Equitest). The appropriate Human Studies Committees approved all research protocols. Performance was evaluated by measuring subjects’ tilt as they performed a standardized clinical test of postural control called the Sensory Organization Test (SOT) [6], [18], [24], [25] with and without the balance aid. The SOT distorts visual and proprioceptive sensory information. In subjects with normal balance function the primary source of undistorted information must come from the motion sensor of the inner ear during this test. This information was either absent or greatly reduced in the subjects we tested. Compared to the no-balance-aid condition, subjects’ sway (rms center of pressure) was significantly decreased from 2.7° to 1.4° (p < 0.05) during the SOT runs when they were provided with the vibrotactile display of their body tilt. The most dramatic finding was that subjects who regularly fell during the SOT without the aid were able to stand without falling when it was turned on. This is shown in Figure 2 where falls correspond to a TPI score of zero [17], [36].

2-axis random-noise platform perturbation. The analysis of the rms body tilt estimates from the inertial instrumentation showed a significant reduction in anteroposterior tilt when the subjects were getting the vibrotactile feedback. Diffusion analysis also showed that the vibrotactile feedback of body tilt allowed the subjects to control posture more quickly than without feedback. During conditions that induced a mild 2-axis random platform motion, there was a significant change (p = 0.04) for all subjects for anteroposterior sway with anteroposterior tilt displayed. The change in mediolateral sway was not significant. This is evidence of direction-specific control. Precursor to a Wearable Balance Prosthesis

Tilt Performance Index

The precursor prosthesis is a wearable, distributed, modular design (Figure 3), which builds upon the single-axis research device and consists of a 3-D motion sensor array (three accelerometers and three gyroscopes), a central processor, and vibrotactile stimulators. Because communications are done through a controller area network (CAN) bus, the instrument assembly and tactor array both require their own microprocesSway on a Randomly Moving Platform sors. For initial research applications, an array of 48 small with Vibrotactile Tilt Feedback tactors displays the estimated body tilt. A laboratory-based Sway responses during perturbed and quiet standing have computer is wirelessly linked to the wearable central processor been characterized in five vestibulopathic subjects by meato receive data from the stand-alone wearable components. The suring the root mean square (rms) center of pressure and present wearable configuration has the sensor package also by the use of stabilogram diffusion analysis [11]. mounted on the small of the back. The tactor array is mounted Vestibulopathic subjects were tested, eyes closed both with in an elastic strip or wide belt worn just above the waistline and and without the vibrotactile feedback, during runs featuring the central processor, and batteries are worn in two pouches on the sides of the 2.5 thighs near the pockets. The wearable prosthesis precursor 2.0 was developed in collaboration with 0.5 Drs. Lars Oddsson and Peter Meyer at Boston University’s Neuromuscular 1.0 Research Center and the Swedish 0.5 Royal Institute of Technology (KTH) 0 Department of Mechanical Design Mechatronics Laboratory’s 2001-2002 −0.5 5 10 15 20 25 30 35 40 Advanced Course [12]. Its primary purpose is to serve as a test bed for the deRun Number velopment of motion sensor arrays for the biomedical research community. Fig. 2. Sequential tilt performance index (TPI) scores from one subject’s sensory organization test (SOT) runs during computerized dynamic posturography testing. SOTs measure sway output during different conditions of sensory input. Sensory input can be either normal, denied, or distorted. The SOT 5 has distorted proprioceptive and no vision inputs. The SOT 6 has distorted proprioceptive and visual inputs. Circles are for SOT 5 runs, while squares are for SOT 6 runs. Open symbols show runs when the prosthesis is turned on, while closed symbols are for the no-aid conditions. TPI = 1/rms center of pressure sway and is scored zero if the subject falls during a run. 86


Motion Sensor Array

The motion sensor array is comprised of three MEMS linear accelerometers and three MEMS rate gyros whose sensitive axes are aligned along three orthogonal directions to provide MARCH/APRIL 2003

Estimated Tilt (Deg)

six-degree-of-freedom (dof) motion information. The associated microprocessor module has anti-aliasing filters for the motion sensor signals, which are then digitized (12 bit) in real time (typically at 100 samples per second) and sent to the central processor via the CAN bus. The availability within the past five years of good-performance, low-cost, small-size MEMS inertial sensors and adVibrator Display vanced electronics has seeded developing the vestibular prosthesis. MEMS extends the photolithography, masking, Tilt Sensors Central Unit etching, and batch processing of solid-state electronics to forming mechanical devices. While the sensors are described in [7], [39], key features are listed here. The sensing mechanism is a glass and silicon chip, roughly 3 mm per side (compared to tens of millimeters for prior, non-MEMS instruments). The vibrating gyro senses Coriolis acceleration (angular) while the accelerometer is an unbalanced see-saw. Both devices rely on electrostatic forcing and sensing so that power dissipation within the mechanism is small compared to Processor older technologies. We have determined that the balance prosthesis should estimate the vertical within 0.1-1° to be useful to a patient [17], [35]. The estimated tilt error depends upon the instrumentation and algorithms. Our algorithm that successfully detects subject tilt with available MEMS instruments deserves elaboration. For the single-axis research device, we use a yoked linear accelerometer and rate gyro to estimate tilt along one axis. The linear accelerometer signal is low-pass filtered while the gyro output is high-pass filtered and integrated (Figure 4). The two signals Fig. 3. Photograph of major components of the wearable deare then combined to yield a wide bandwidth (−10 Hz) estivice to show mounting locations on the body. Also shown is mate of the tilt angle. Based on an inverted pendulum model the laboratory computer that communicates with the wearand demonstrated in-patient testing, 0.03 Hz was selected as able device. the filters’ break point to separate gravity from angular (180° phase shift from tilt) and lateral acceleration. Bg Since the accelerometer input axis is nominally hori3 + s2ω (2ς + 1) − sN N 1 zontal, a steady-state patient Gyro Σ 2 2 k s + ( )( s + ω s + ζωNs + ωN2) g N tilt, θ, results in accelerome+ ter output, y = g sinθ + Bias, Low Pass Indicated Σ Ba where Bias is the output volt10 Hz Tilt + age when no acceleration is − sω2(2ς + 1)+ω3 1 applied. The bias is caliAccel Σ 2) + ka (s + ωN)(s2 + 2ςωNs + ωN brated at instrument turn-on but may change no more than (a) 0.001 rad [~1 milligravity (mg)] over roughly 16 hours (one waking day). One mg 1.5 stability over a day represents the best MEMS sensors, 1 while 5 to 50 mg is typical of Indicated the much less expensive senAccelerometer sors used for automotive ap0.5 plications [29], [38]. Since inertial instruments are often Gyro 0 calibrated and corrected as a function of temperature, these −0.5 performance figures should −5 5 0 10 15 20 25 30 35 40 be considered as remaining Time (s) error after compensation. (b) If the gyro signal were not high-pass filtered, small Fig. 4. (a) Block diagram of body tilt estimation. (b) The estimated tilts for a 1° step change in acerrors in gyro bias would in- tual tilt are shown for the gyro, accelerometer, and combined (indicated) channels. IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

MARCH/APRIL 2003

87

Further biomedical research using a 3-D wearable version is justified and will be done by four investigators who will incorporate the wearable prosthesis into their existing research protocols on human postural stability.

tegrate to large low-frequency tilt errors. With 0.03-Hz break frequency, an 80°/h (degrees per hour) gyro bias shift in less than 10 s contributes a 0.1° tilt error, which lasts for roughly 10 s. This creates a transient tilt error. The maximum allowable transient tilt error for a balance prosthesis is not yet known and must be kept small for initial experiments. Short-term bias stabilities of 100°/h are satisfied by high-performing MEMS gyros while lower cost instruments are less stable. We define the motion sensor array’s (MSA) axes by two accelerometer input axes. The other input axes are nominally parallel to the MSA axes; however, mass-produced MEMS instruments are typically misaligned 1 to 3° with respect to their mounting surfaces so that 2 to 5% cross-axis errors (significant magnitudes) ensue. The motion sensor array’s imperfect alignment with the patient’s vertical results in three misalignment angles. Imperfect alignment of the individual instruments within the motion sensor array results in nine more angles for which one must account.

For our initial 3-D studies, 12 input axis misalignments, six biases, and six scale factors (output signal per unit acceleration or rate input) will be calibrated on a two-axis test station (Figure 5). This station will also be used to verify all six-dof tilt estimation algorithms before they are applied to patients. These parameters will be augmented by a 1-s on-patient calibration that will align the motion sensor array axes to the patient’s preferred vertical and will update the bias estimates. Array calibration requires complex apparatus and procedures. For the initial 3-D testing, we wish to demonstrate prosthesis feasibility; therefore, a negative result due to poor sensor selection or performance must be avoided. Calibration coefficients, as determined on the test station, will be embedded in the motion sensor array processor making them transparent to the user. Based on the 3-D testing and further analyses, the set of calibrated parameters may be reduced for future applications. Central Processor

The central processor is a Real Time Devices USA CMG16686GX300 PC104 embedded system running the Linux operating system. The PC104 was chosen beRefrigerator Table Control and Acquisition cause of its high performance, small 0.86 m form factor (3.6 in by 3.8 in), low processor and bus power requirements, and availability of the necessary modules: 1) a 300-MHz PCI Pentium class low power (5 W) unit, 2) a CAN-interface utility module, and 3) a PCMCIA Outer Gimbal Axis PC card carrier utility module for IEEE 802.11b wireless communication and 64-Mb flash memory cards. Linux (kernel version 2.2.14-5.0) was chosen because it is open-source software, and it can be configured as a real-time system. The central processor (CP) uses the signals from the moSlip Ring tion sensors that are sent to it via the Assembly CAN bus to make estimates of the subInner Table Axis Controlled Oven, ject’s body tilt. The CP then codes Sensors and Heaters these estimates into signals that are deInside livered to the vibrotactile display microcontroller. Fig. 5. Photograph of two-axis, motion sensor test station equipped for thermal sensitivity testing. The horizontal (black) cylinders drive outer gimbal rotation axis. The vertical (black) cylinder below the station rotates inner table axis. The system under test is inside the cubic, electrically heated thermal enclosure mounted on the inner axis. Seen behind the test station and connected by insulated flexible tubes (orange) is the cooler, which enables operation to below –40 °C. 88


Microcontroller Modules for Sensors and Stimulators

The real-time microcontroller was developed in modular form by a previous MARCH/APRIL 2003

KTH Mechatronics student team. The module is based on a 1.25-MHz, 16-bit Ristel M167-2 processor board. The original module design provided for both analog and digital inputs as well as for a CAN interface. It was made application specific by adding only the filters, amplifiers, and digital I/O appropriate for the sensor and stimulator applications. Tactile Vibrator Display

The display is a 3 by 16 array of tactors, which are FDA approved for use in a sensory substitution hearing aid (Tactaid) (Figure 6). Mounted on the subject’s trunk, the tactor array is oriented with 16 circumferential columns (22.5° resolution) to indicate the tilt direction with each column having three rows to distinguish magnitude. Spacing between individual tactors is greater than the two-point discrimination distance for the torso. The CP gates a 250-Hz signal to an individual tactor’s driving amplifier.

Activated Tactor Level

3 2 1

Vibrator Display

Tilt Sensors

CAN Bus

CAN devices exchange information at 1 Mb/s using a standardized (ISO 11898), content-oriented addressing scheme instead of a device-oriented one. Local intelligence in each CAN station is used to determine whether or not that station processes a message, which allows for modular electronic device design. Thus, for example, instrumentation to measure the motion of the feet could be easily added to the system at a later date if this became desirable.

(a)

Tactor Display Coding

The laboratory computer is a standard IBM notebook that has an IEEE 802.11b wireless communication card. The lab computer runs dual operating systems: Linux for software development on the PC-104 and Windows XP for interacting with the prosthesis during experiments. Labview is used to display, control, and store prosthesis data. Windows XP is also used for editing and downloading the program to the motion sensor and vibrotactile display microcontrollers. Wireless communication between computers is done using both file transfer protocol (FTP) and hypertext transmission protocol (HTTP). The FTP server provides a way of transferring the collected data, new algorithms, and new code. The HTTP server provides a fast stream of online information from the PC-104 to the lab computer and provides an interactive Web page, built with common gateway interface (CGI) scripts to enable remote functionality such as reboot, shutdown, and start and stop of programs.

Activated Tactor Level

3

Lab Computer

2

1

None Backward

Forward

0 1 Body Tilt (°)

3

5

(b)

Conclusions

There is a clear need for a prosthesis that improves postural stability in the balance impaired. Such a device would be used as a temporary aid during recovery from ablative inner-ear surgery and as a permanent prosthesis for those elderly prone to falls. Research using a one-axis device that estimates body tilt and displays it to vestibulopathic subjects via an array of tactile vibrators has demonstrated feasibility; thus, further biomedical research using a 3-D wearable version is justified and will be done by four investigators who will incorporate the wearable prosthesis into their existing research protocols on human postural stability. Acknowledgments

We gratefully acknowledge Joseph B. Nadol, M.D., and the Massachusetts Eye and Ear Infirmary for facilities support; the IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

Rear Tilt Activated

Forward Tilt Activated

(c) Fig. 6. (a) Diagram of 3 by 16 tactor array worn on the body. (b) The chart shows the relationship between tactor activation in a specific row to body tilt. This is further explained by (c), where forward and backward tilt activate only the forward-most and backward-most columns of tactors, respectively. MARCH/APRIL 2003

89

W.M. Keck Foundation for its support of this project; the Keck Neural Prosthesis Research Center; Neurocom International for the generous donation of Balance Master software; and the National Institutes of Health for support in developing and validating the motion sensor arrays (NIH NIDCD R01 DC6201). Conrad Wall III received his B.S. and M.S. degrees in physics from Tulane University and his Ph.D. in bioengineering from Carnegie Mellon University. He is an associate professor of otology and laryngology at Harvard Medical School and in the joint Harvard-MIT Whitaker College of Health Sciences Technology. Conrad is the founder and director of the Jenks Vestibular Diagnostic Laboratory at the Massachusetts Eye and Ear Infirmary, where he also participates in sponsored research. Conrad is project lead and associate team lead in the Neurovestibular Adaptation Team of NASA’s National Space Biomedical Research Institute and project lead of the balance project of the Infirmary’s Neural Prosthesis Research Center. He chairs the working group in charge of the ANSI standard on the vestibular function test battery. Marc S. Weinberg received his B.S., M.S., and Ph.D. degrees in mechanical engineering from Massachusetts Institute of Technology. Marc is laboratory technical staff in the Hardware Design and Development Directorate. Marc has been responsible for the design and testing of a wide range of micromechanical gyroscopes, accelerometers, hydrophones, microphones, angular displacement sensors, chemical sensors and biomedical devices. He holds 22 patents with 11 additional in application. He was given Draper’s Best Patent (twice), Best Publication (twice), and Distinguished Performance Award for his work on the Tuning Fork Gyro, the first silicon micromechanical gyroscope to demonstrate resolution better than 100°/hr in 60 Hz, and other MEMS work. Address for Correspondence: Conrad Wall III, Jenks Vestibular Diagnostic Laboratory, Massachusetts Eye and Ear Infirmary, 243 Charles Street. Boston, MA, 02114-3096. Tel: +1 617 573 4153. E-mail: [email protected]. References [1] Don’t Let a Fall Be Your Last Trip, American Academy of Orthopaedic Surgeons [Online]. Available http://orthoinfo.aaos.org [2] T.D. Anderson, et al., “Prevalence of unsuspected acoustic neuroma found by magnetic resonance imaging,” Otolaryngol. Head Neck Surg., vol. 122, no. 5, p 643-6, 2000. [3] G. Andersson, et al., “Evaluation of quality of life and symptoms after translabyrinthine acoustic neuroma surgery,” Am. J. Otol., vol 18, no. 4, pp. 421-6, 1997. [4] P. Bach-y-Rita, “Late post-acute neurologic rehabilitation: neuroscience, engineering and clinical programs,” Arch. Phys. Med. Rehab., to be published. [5] P. Bach-y-Rita and M.E. Tyler, “Tongue man-machine interface,” Stud. Health Technol. Inform., vol. 70, pp. 17-9, 2000. [6] F.O. Black, C. Wall, and L.M. Nashner, “Effects of visual and support surface orientation references upon postural control in vestibular deficient subjects,” Acta. Otolaryngol., vol. 95, pp. 199-201, 1983. [7] B. Boxerhorn, “A vibratory micromechanical gyroscope,” in Proc. AIAA Guidance and Control Conf., 1988. [8] R.W. Cholewiak and A.A. Collins, “Sensory and physiological bases of touch,” in The Psychology of Touch, M.A. Heller and W. Schiff, Eds. Hillsdale, NJ: Laurence Erlbaum Assoc.: 1991, pp. 23-60. 90


[9] B. Cohen, J. Suzuki, and M. Bender, “Nystagmus induced by electrical stimulation of the ampullary nerves,” Acta Oto-Laryngologica, vol. 60, pp. 422-436, 1965. [10] B. Cohen, J.I. Suzuki, and M.B. Bender, “Eye movements from semicircular canal nerve stimulation in the cat,” Ann. Otol. Rhino. Laryngol., vol. 73, pp. 153-169, 1964. [11] J.J. Collins and C.J. De Luca, “Open-loop and closed-loop control of posture: a random-walk analysis of center-of-pressure trajectories,” Exp. Brain Res., vol. 95, no. 2, pp. 308-18, 1993. [12] M. Grimheden and M. Hanson. Providing a framework for prototype design of Mechatronic systems: A field study of an international collaborative educational project using the mechatronic learning concept,” in 3rd European Workshop on Education in Mechatronics, 2002. [13] S.J. Herdman, et al., “Vestibular adaptation exercises and recovery: acute stage after acoustic neuroma resection,” Otolaryngol. Head Neck Surg., vol. 113, no. 1, pp. 77-87, 1995. [14] E.E. Hill-Westmoreland, K. Soeken, and A.M. Spellbring, “A meta-analysis of fall prevention programs for the elderly: how effective are they?,” Nurs. Res., vol. 51, no. 1, pp 1-8, 2002. [15] P.O. Jantti, I. Pyykko, and P. Laippala, “Prognosis of falls among elderly nursing home residents,” Aging (Milano), vol. 7, no. 1, pp. 23-27, 1995. [16] K.A. Kaczmarek, et al., “Electrotactile and vibrotactile displays for sensory substitution systems. IEEE Trans. Biomed. Eng., vol. 38, no. 1, pp. 1-16. [17] E. Kentala, J. Vivas, and C. Wall, Reduced postural sway by use of a vibrotactile balance prosthesis prototype in subjects with vestibular deficits,” Annals ORL, to be published. [18] A.D. Kuo, et al., “Effect of altered sensory conditions on multivariate descriptors of human postural sway,” Exp. Brain Res., vol. 122, no. 1, pp 185-195, 1998. [19] R. Lewis, et al., “Vestibular adaptation studied with a prosthetic semicircular canal,” J. Vestibular Res., 2002. [20] A.J. Matheson, C.L. Darlington, and P.F. Smith, “Dizziness in the elderly and age-related degeneration of the vestibular system,” NZ J. Psychol., vol. 28, no. 1, pp. 10-16. [21] D. McGinn, “The bionic eye and ear,” Newsweek, p. 52-53, 2002.. [22] S.N. Merchant, et al., “Temporal bone studies of the human peripheral vestibular system. Normative vestibular hair cell data,” Ann. Otol. Rhinol. Laryngol. Suppl., fol. 181, pp. 3-13, 2000. [23] D. Merfeld and R. Rabbitt, “Vestibular prosthetics,” in Neuroprosthetics: Theory and Practice, K. Horch and G. Dhillon, Eds. Singapore, World Scientific, 2002. [24] L.M. Nashner, F.O. Black, and C. Wall, “Adaptation to altered support and visual conditions during stance: patients with vestibular deficits,” J. Neurosci., vol. 2, no. 5, pp. 536-544. [25] Management of Balance and Mobility Disorders Bibliograpy. Neurocom International, 2002. [26] National Strategic Research Plan: Language and Language Impairments, Balance and Balance Disorders, Voice and Voice Disorders. National Institutes of Health. [27] A. Rupert, F. Guedry, and M. Reschke, “The use of a tactile interface to convey position and motion perceptions,” Advisory Group for Aerospace Research and Development, Rep. AGARD CP 541, 1993. [28] A.H. Rupert, et al., “Tactile interface to improve situation awareness,” Aerospace Medical Association Washington, DC, 1996. [29] P.G. Savage, “Strapdown Sensors,” in NATO/AGARD Lecture Series 95, Strapdown Inertial Systems-Theory and Application. Neuilly-sur-Seine, France, 1978. [30] A.P. Scinicariello, et al., “Enhancing human balance control with galvanic vestibular stimulation,” Biol. Cybern., vol. 84, no. 6, pp. 475-480, 2001. [31] J. Suzuki, et al., “Implantation of electrodes near individual vestibular nerve branches in mammals,” Annals Otolaryngology Rhinology & Laryngology, vol. 78, pp. 815-826, 1969. [32] J.I. Suzuki and B. Cohen, “Head, eye, body and limb movements from semicircular canal nerves,” Exp. Neurol., vol. 10, pp393-405, 1964. [33] P. Trueblood, et al., “Performance and impairment-based assessments among community-dwelling elderly: sensitivity and specificity,” Issues Ageing, vol. 24, no. 1, pp. 2-6, 2001. [34] L. Velazquez-Villasenor, et al., “Temporal bone studies of the human peripheral vestibular system: Normative Scarpa’s ganglion cell data, Ann. Otol. Rhinol. Laryngol. Suppl., vol. 181, pp 14-19, 2000. [35] C. Wall, et al., “Balance prosthesis based on micromechanical sensors using vibrotactile feedback of tilt,” IEEE Trans. Biomed. Eng., vol. 48, no. 10, pp. 1153-1161, 2001. [36] C. Wall and E. Kentala, “Control of sway in vestibulopathic subjects using vibrotactile display of body tilt,” presented at MidWinter Meeting of the Association for Research in Otolaryngology, St. Petersburg Beach, FL, January 27-31, 2002. [37] H.W. Wallmann, “Comparison of elderly nonfallers and fallers on performance measures of functional reach, sensory organization, and limits of stability,” J. Gerontol. A Biol. Sci. Med. Sci., vol. 56, no. 9, pp M580-M583, 2001. [38] M. Weinberg, “Grades of Motion Sensors and Their Characteristics,” 2002. [39] M. Weinberg, et al., “A micromachined comb drive tuning fork rate gyroscope,” in Proc. ION 49th Annual Meeting, 1994. [40] S.L. Whitney, G.F. Marchetti, and A.I. Schade, “Relationship between the Timed Up & Go and reported falls in persons with vestibular disorders,” J. Vestibular Research, 2002. [41] V.J. Wilson and G. Melvill Jones, Mammalian Vestibular Physiology. New York: Plenum, 1979. MARCH/APRIL 2003