Abstract

In the paper the sensory system for a FES gait re-education system is proposed. The gait training orthosis for incomplete spinal cord injured patients is being developed. Sensory feedback and walking phase detection are the most important parts of the orthosis. Implementation of several types of sensors into the rehabilitative system can make it more reliable. Four accelerometers, gyroscope and two goniometers were used in the measurement system. Data collected by the sensors were input into mathematical algorithms. As a reference measurement device an optical position measurement system was used. The data collected were analyzed offline in order to develop a suitable walking phase detection.

In order to choose the appropriate sensors, the assessments were performed in three healthy subjects. The results have shown a need and advantages of the multisensor use. The appropriate feedback signal will be delivered to the patient who will voluntarily control the functional electrical stimulation.

Introduction

In the recent years our research studies are focused on incomplete spinal cord injured (SCI) patients. In earlier studies we realized the necessity of functional electrical stimulation (FES) training of the paralyzed and atrophied muscles in the early period after spinal cord injury [1]. The candidates are all patients with upper motor neuron lesion, in more clinical terms the patients with thoracic or cervical lesion to the spinal cord. Only a few of incomplete SCI patients are candidates for permanent FES, most of them use FES only in the early period of the rehabilitation process. Peroneal nerve stimulation was found useful to provoke flexion response resulting in the swing phase of walking.

Several existing FES systems employing peroneal nerve stimulation used sensory feedback to improve walking pattern. The sensory feedback was usually provided by use of simple artificial sensors [3,4]. Data collected by a pair of miniature accelerometers were used to distinguish between the stance and swing phase. Automatic detection algorithms were used to identify the appropriate phase of walking and to control the FES. On the basis of the results obtained, development of a small implantable sensor-stimulator device was proposed. No sensor failure or minimal misadjustment was considered.

The aim of a FES rehabilitative system for re-education of walking is not only to deliver electrical stimulation to the paralyzed muscles, but also to assess the sensory information from the paralyzed limb. The sensory information is fed back to the patient and not to the stimulator control unit. The FES rehabilitation systems for re-education of walking are intended to be used in incomplete SCI persons soon after the accident or onset of disease [2]. These systems are to be used within the rehabilitation centers and applied by therapists. Surface electrical stimulation is therefore appropriate. We are developing two separate systems for swing and stance phase detection. The adequate approach must be selected according to the patient’s gait deficits. In this paper we are proposing a swing phase detection.

Gait re-education system for swing phase detection is based on multisensor use, simple feedback signal, which is fed back to the patient and is independent of single sensor failed measurement. The feedback signal can be delivered to the patient with vibrotactile or electrical stimulation or with a small earphone. It represents the successfully or unsuccessfully performed swing phase. The patient has an opportunity to simultaneously control the amplitude of FES by control lever [5] in order to improve swinging of the paralyized lower extremity.

Methods

We have developed a sensory system together with a walking phase detection algorithm. The measuring device consists of several sensors and can be considered as a multisensor device. Obviously, every sensor provided an output that served as input into the phase detection algorithm. The goal was to analyze the data captured and to decide whether all sensors or sensor groups are needed for swing phase detection.

The employed sensors were mounted on an aluminum plate as shown in figure 2. Four uniaxial accelerometers (ACCESS, Switzerland) were attached to measure tangential and radial components of the movement (figure 1). The idea of the accelerometer employment is not new [4]. This placement allows us to compute the acceleration in the ankle joint. A uniaxial gyroscope (Murata ENC 03JA) was used to measure the angular velocity of shank. The gyro’s output signal was low-pass filtered (0.2 Hz) to eliminate the temperature drift, high-pass (1 kHz) filtered to remove noise and amplified. Two goniometers were used to estimate the ankle joint angle and knee joint angle, respectively. There were some problems when attaching the goniometers on human leg. First of all we had to take in consideration that every attachment of the goniometer differs from the previous one. The problem was solved by self-initialization routine performed before each measurement.

In order to test the developed sensory system we employed the contactless optical position measurement system OPTOTRAK (Northern Digital Inc., Ontario, Canada).

Data captured by accelerometer were filtered and converted into an appropriate form for comparison with OPTOTRAK ankle joint acceleration data. With a simple solution of two independent mathematical equations we obtain the ankle joint tangential and radial acceleration:

(1)

Figure 1. Coordinate system of the lower extremity and the sensory system.

The reference measurement results acquired by OPTOTRAK were compared with the data assessed by sensory system. We adapted the reference data to the system data by using simple mathematical tools. The ankle joint angle and the knee joint angle were computed by using cosine relation:

(2)

The ankle joint (marker T₃) tangential and radial acceleration was calculated using numerical derivative of the ankle joint reference position and coordinate system transformation. This operation added derivative noise. Data were filtered before derivation with third order Butterworth low-pass filter with cutoff frequency 5 Hz. After derivation Gauss filter (5 Hz) was used.

Figure 2. Sensory equipment attachment. Goniometers were attached by using velcro straps or adhesive tape. OPTOTRAK active markers were placed in joints and along segments.

The second part of the data analysis was swing phase detection. We were trying to detect the beginning of the phase by every single sensor in order to avoid dependence on sensor failure. While this was not acceptable in a case of accelerometers, the phase detection was focused on sensor groups.

Swing phase was split into three subphases: initial swing (push-off), middle swing and terminal swing (foot down), but we detected only the swing/stance transition. For phase identification artificial neural networks were used. The network with one hidden layer, containing 10 neurons, and one output was trained for each sensor group to detect the stance/swing transition.

Results

Three healthy subjects participated in the experiment. They were asked to walk with normal speed, faster than normal and slower than normal. Their walking characteristics were recorded by sensory system and OPTOTRAK and analyzed on-line and off-line.

The recording track was 7 to 15s long; the subject has made approximately 5 steps, before OPTOTRAK went out of range. Figure 3 presents the comparison of optical measurement and our sensory information.

Figure 3. Data assessed by our sensory system and with optical measurement. Bellow accelerometers output based basogram.

According to the optical measurements we can conclude that our sensory system is sufficiently accurate. The wave in gyro’s data is the easiest way to present the swing phase. Accelerometer data show clearly the transition between stance and swing phase. The detection with ankle and knee joints is less accurate. The last figure presents the results of swing/stance phase detection with accelerometers. The swing phase was successfully detected by every sensor group and the corresponding outputs will be integrated into single stance/swing basogram.

Conclusions

FES gait re-education system is based on artificial sensory feedback, which should replace the natural sensory input in a process of relearning of the gait cycle phases in walking of incomplete SCI patients [2]. The feedback signal is fed directly to the patient and represents a successful accomplishment of selected gait phase.

The signals assessed have helped us to define the appropriate sensors or group of sensors. All time courses (except from the gyro) were compared with optical sensory system with the aim of testing the reliability of the artificial sensory system. According to the results we concluded that the use of the developed sensory system is appropriate for gait cycle phase detection in healthy subjects. In future research the algorithms will be adapted to the gait patterns encountered in incomplete SCI subjects.

The gait phase detection made by each sensor group will be integrated into one single output. This can be achieved by implementing algorithms with ability of calculation with certainty, sensory integration [6]. Therefore we take into the consideration that the sensor failure or failed measurement data can not jeopardize the gait phase detection. Also, the variety of sensors and their specifications make us possible to detect the desired phase in various walking patterns.

References

[1]. T Bajd, A Kralj, M Štefančič, N Lavrač, Use of Functional Electrical Stimulation in the Lower Extremities of Incomplete Spinal Cord Injured Patients, Artif Organs, 23(5):403-409, 1999.

[2]. T Bajd, I Cikajlo, R Šavrin, R Erzin, F Gider, FES Rehabilitative Systems for Re-Education of Walking in Incomplete Spinal Cord Injured Persons, Neuromodulation, 3(3):167-174, 2000.

[3]. A Kostov, BJ Andrews, DB Popović, RB Stein, WW Armstrong, Machine Learning in Control of Functional Electrical Stimulation Systems for Locomotion, IEEE Trans Biomed Eng, 42(6) :541-551, 1995.

[4]. ATHM Willemsen, F Bloemhof, HBK Boom, Automatic Stance-Swing Phase Detection from Accelerometer Data for Peroneal Nerve Stimulation, IEEE Trans Biomed Eng, 37(12) :1201-1208, 1990.

[5]. I Cikajlo, T Bajd, Use of telekinesthetic feedback in walking assisted by functional electrical stimulation. J Med Eng Technol, 24:14-19, 2000.

[6]. HF Durrant-Whyte, Integration, coordination, and control of multi-sensor robot systems. Boston: Kluwer Academic, 1987.

Acknowledgements

The authors wish to acknowledge the financial support of the Republic of Slovenia Ministry of Science and Technology.