AN EXPERIMENTAL EVALUATION OF THE GYROSCOPE AS A SENSOR IN FES FOOT-DROP CORRECTION SYSTEMS

SN Ghoussayni, JR Henty, DE Wood*, DJ Ewins

Biomedical Engineering Group, University of Surrey, Guildford, UK

*Dept. of Medical Physics and Biomedical Eng., Salisbury District Hospital, Salisbury, UK

SUMMARY

This paper presents the work done on the use of a Gyroscope (Gyro) as a sensor for foot-drop correction systems using Functional Electrical Stimulation (FES). These systems usually employ foot switches to control the timing of stimulation. It is believed that the replacement of the heel switch with the gyroscopic sensor would offer several advantages. The performance of the new sensor and associated software in gait event detection have been tested and compared to foot switch and kinematic data. The tests were carried out with both able-bodied and hemiplegic patients walking over different terrains. The results indicate that the Gyro can be used to detect the necessary gait events for controlling the stimulator timing.

STATE OF THE ART

Since the work of Liberson in 1961 /1/, many researchers have been interested in the application of electrical stimulation for the restoration of lost or impaired function. The correction of foot-drop is one common application with increasing clinical use. FES foot-drop stimulators typically employ a physical sensor (usually a foot switch) as a feedback source controlling the timing of stimulation. These sensors are reported – by more than one study /2/, /3/, /4/ – to suffer from a number of limitations, including detection errors, in particular with pathological gait, and a relatively short mean time to failure with some patients, for example children with cerebral palsy. Previous work at the University of Surrey /5/, /6/, has investigated the use of a gyroscopic sensor as an alternative sensor to be used with FES foot-drop correction systems. The use of a vibratory gyroscope in combination with other sensors has also been investigated by other researchers /7/, /8/. A wide variety of other sensors, both artificial and natural, has been investigated in detail to be used with FES systems. Accelerometers, goniometers, and inclinometers are some of the artificial sensors that have been researched for use with FES systems /2/, /4/. Recordings of both electroneurogram (ENG) and electromyogram (EMG) have also been studied as possible natural sensors for FES systems /9/, /10/.

The replacement of the foot switch by the Gyro offers several advantages, which could improve system reliability and function. The Gyro is a small and lightweight sensor (with potential for further miniaturisation and implantation), which can be easily donned and doffed (positioning is not very critical) with minimal encumbrance to the patient. The nature of the Gyro contributes to its high reliability and long lifetime during which there is little or no deterioration in its performance. When compared to the information provided by foot switches, which is essentially of a binary nature, the Gyro based sensor system has the advantage that it could predict a gait event - and therefore begin, end, ramp up or ramp down stimulation - before its occurrence.

As a replacement for foot switches in FES foot-drop correction systems, the sensor output (voltage proportional to angular rate of moving limb segment) has to be adapted to be suitable for timing the stimulation. In our work this has been done by developing detection software for the Gyro signal to give four gait events (Heel Contact HC, Foot Flat FF, Heel Rise HR, and Toe Off TO). A portable microcontroller based unit with data logging capability was developed. This unit runs the gait event detection software and provides real time detection. This system was used for the evaluation of the sensor

and associated software by comparing the detection times to those obtained from simultaneously recorded foot switch and kinematic data (kinematic data was not available for evaluating the system when used by the patients, as these tests were not performed in a lab setting). The evaluation of the system involved its use by both able-bodied subjects and patients with foot-drop, and the system performance was tested over different terrains including leveled and inclined walking, and staircase climbing.

MATERIAL AND METHODS

A Murataa ENC-05E piezoelectric vibrating gyroscope was used to capture the angular velocity of the foot in one plane (worn on the anterior side of foot). The output of the Gyro and two foot switches (located under the heel and first metatarsal head) was sampled by the data logger at 164 Hz. The tests were split into two sets.

In the first set of tests, the system was used by five able-bodied subjects (average age 38 years), and simultaneous kinematic data was captured using a Qualysis ProReflex motion capture system sampling at 240 Hz. Three retroreflective markers were attached to the foot at the lateral malleolus, the lateral side of calcaneus, and above the first metatarsal head. The positions of these makers were visually inspected afterwards to determine the times of the gait events. Each subject performed six trials walking on leveled floor, up and down a 7º ramp, and a 7 step staircase, at two speeds (self-selected normal and slow). The performance of the sensor was analysed by comparing its predicted times to those given by the foot switches, and then by comparing both methods against the kinematically determined gait events.

In the second set of tests, the system was used by four patients with foot-drop condition (average age 53 years, 3 females and 1 male, 2 MS and 2 CVA patients, average time since diagnosis 4 years, average time since stimulator was first used 5 months). The patients were asked to perform two trials: with and without using their Odstock Dropped Foot Stimulator (ODFS, Salisbury District Hospital, Salisbury, UK). The data from the Gyro and two foot switches was collected using the data logger while the patients walked around the rooms of the hospital over both inclined and leveled floor. The data was then analysed by comparing the Gyro and foot switches detection times for each gait event (foot switch detection times were determined by thresholding). Two of the patients also ascended a stairway, however this data is still being processed.

RESULTS

For the first part of the study, the differences in time between the three methods are presented in Tables 1-3. These show the average difference between each of the methods used for different events against the three terrains.b The absolute difference was also calculated and averaged to avoid any misleading conclusions from considering the average difference alone.

Table 1: The results of the comparison of detection times (in ms) between the Gyro-Kinematic method. (D = Average Difference; |D| = Average of Absolute Difference)

Event\Terrain	Level Floor		Ramp Up/Down		Stairs Up/Down
Event\Terrain	D	\|D\|	D	\|D\|	D	\|D\|
HC	-7	28	-11	25	-47	95
FF	22	25	28	30	37	47
HR	4	29	25	68	110	136
TO	56	58	69	69	80	98

Table 2: The results of the comparison of detection times (in ms) between the Foot switch-Kinematic method.

Event\Terrain	Level Floor		Ramp Up/Down		Stairs Up/Down
Event\Terrain	D	\|D\|	D	\|D\|	D	\|D\|
HC	31	37	42	63	55	111
FF	43	79	15	64	11	58
HR	-44	81	-65	112	-73	108
TO	-29	90	23	73	-47	78

Table 3: The results of the comparison of detection times (in ms) between the Gyro-Foot switch.

Event\Terrain	Level Floor		Ramp Up/Down		Stairs Up/Down
	D	\|D\|	D	\|D\|	D	\|D\|
HC	-38	38	-53	57	-102	160
FF	-21	78	14	66	-4	64
HR	48	99	90	112	179	179
TO	85	118	45	101	128	138

The results from the second part of the study are presented in Table 4. As the trials with the 4 hemiplegic subjects were performed in a non-lab setting, the data collected is from the Gyro and 2 foot switches only.

Table 4: The results of the comparison of detection times (in ms) between the Gyro-Foot switch when used by 4 hemiplegic patients walking on leveled and inclined floor.

Event\Trial	No Stimulation		With Stimulation
Event\Trial	D	\|D\|	D	\|D\|
HC	-7	48	-3	32
FF	-3	16	42	51
HR	118	135	86	156
TO	98	98	-3	144

DISCUSSION

The results from the first part of the study show that when compared to the foot switches detection times the Gyro had a similar performance and was generally closer to the kinematically determined times. As the trials with the 4 hemiplegic subjects were performed in a non-lab setting, the data collected was from the Gyro and 2 foot switches only. This meant that the Gyro and foot switch times could not be compared against a third method. However, one observation is that the differences between the Gyro and foot switch detection times (Table 4) are comparable to those from the first set of results (Table 3). It is suggested that this is because the sensor system performed equally well as in the first part of the study. It can be concluded from the presented results that the Gyro sensor system is capable of detecting four gait events in both able-bodied and hemiplegic patients. The sensor also appears to be appropriate for staircase and slope walking.

After some necessary modifications to the software, the sensor was also tested when worn on the anterior aspect of the shank. The preliminary results for this are promising. The use of the sensor when worn on the shank may offer a more convenient use with potential for developing a miniature unit encompassing the sensor, stimulator, electrodes and microcontroller. Future work includes completion of analysis of patient data, and further testing of the system when used on the shank. This will be followed by real time use with the stimulator, evaluation of patient feedback and the development of a take home system.

REFERENCES

/1/ Liberson WT, Holmquest HJ, Scot D, Margot D. (1961). Functional Electrotherapy: Stimulation of the Peroneal Nerve Synchronised with the Swing Phase of the Gait of Hemiplegic Patients. Archives of Physical Medicine and Rehabilitation, 42, pp 101-105

/2/ Dai R, Stein RB, Andrews B. (1996). Application of Tilt sensors in Functional Electrical Stimulation. IEEE Transactions on Rehabilitation Engineering, 4, pp 63-71

/3/ Taylor PN, Burridge J, Ewins DJ, Swain ID. (1995). A Two Channel Stimulator for Gait Assistance. Proceedings of the BES Symposium on Electrical Stimulation – Clinical Systems, University of Strathclyde, pp 41-42

/4/ Willemsen ATM, Bloemhof F, Boom HBK. (1990). Automatic Stance-Swing Phase Detection from Accelerometer Data for Peroneal Nerve Stimulation. IEEE Transactions on Biomedical Engineering, 37, pp 1201-1208

/5/ Henty JR, Ewins DJ. (1998). Applications of Gyroscopic Angular Velocity Sensors in FES Systems. Proceedings of the 6^th Vienna International Workshop on Functional Electrical Stimulation, Vienna, pp 157-160

/6/ Henty JR, Wood DE, Ewins DJ. (1999). Detection of Gait Events using a Vibratory Gyroscope. Proceedings of the 4^th Annual conference of the international FES Society, Japan, pp 73-76

/7/ Popovic MR, Keller T, Ibrahim S, Bueren GV, Morari M. (1998). Gait Identification and Recognition Sensor. Proceedings of the 6^th Vienna International Workshop on Functional Electrical Stimulation, Vienna, pp 153-156

/8/ Williamson R, Andrews B. (1997). Sensors For FES Control. Proceedings of the Second Annual IFESS Conference and Neural Prosthesis: Motor Systems V, Canada, pp 213-215

/9/ Strange KD, Hoffer JA. (1999). Gait Phase Information Provided by Sensory Nerve Activity DuringWalking: Applicability as State Controller Feedback for FES. IEEE Transactions on Biomedical Engineering, 46, pp 797-809

/10/ Kershaw RA, Jones R, Bateman A. (1993). The Use of EMG for Real Time Closed Loop Control of Functional Electrical Muscle Stimulation. Proceedings of The Ljubljana FES Conference, Ljubljana, pp 123-125

ACKNOWLEDGEMENTS

The authors would like to thank the University of Surrey, the Engineering and Physical Sciences Research Council, and the Lebanese National Council for Scientific Research, and the staff and patients at Salisbury District Hospital for supporting this project.

AUTHOR’S ADDRESS

a Murata Electronics (United Kingdom) Ltd., Hampshire, GU13 8UN, UK.

b Negative values indicate that the predicted time was earlier than the time given by the second method. For Table 1 this would be the kinematic method.