AUTOMATED ESTIMATION OF STIMULATION THRESHOLDS
USING SINGLE TWICHES AND
ACCELERATION SENSORS
S. Sauermann, M. Bijak, M. Reichel, D. Rafolt, E. Unger, W. Mayr, H.
Lanmüller
Department of Biomedical Engineering & Physics
SUMMARY
In Functional Electrostimulation (FES) the relation between stimulation input and muscle force output is often called the “recruitment curve”. Especially with surface stimulation systems the properties and position of stimulation electrodes have strong influence on the recruitment curve. In the clinical application it has therefore to be established at every session and even within sessions. This estimation process is of clinical relevance as it takes time and effort which would be better invested into optimisation of the stimulation outcome.
This work was therefore aimed at a method
for automated estimation of the stimulation threshold in the clinical
application of
The algorithm estimated the number of spikes correctly in 15 of 21 trials. It missed a single spikes in two cases and detected one wrongly in one case. In the remaining three cases the number of spikes was wrong by more than one. The automated measurement equipment proved very useful in speeding up the measurement process.
State of the Art
There have been numerous approaches to measure the recruitment characteristic of FES Systems with the help of automated procedures in order to ease and shorten the process of setting up stimulation parameters.
Kilgore and Peckham [Kilgore et.al. 1993] have used the "External Moments Grasp Synthesis Procedure" which is based on measurements of the joint moments under varying joint angles and stimulation parameters, and measurements of the passive joint moments without stimulation [Kilgore et.al. 1993]. This results in a "stimulus map" which describes the stimulation parameters necessary for a certain hand grip movement task.
Other studies used the simpler "rule
based" approach for control of
Muscular dynamics has been described by a so called Hammerstein structure which consists of a static nonlinearity followed by a linear dynamic system [Durfee et.al. 1989]. This concept has been used in many studies of open- and closed-loop control of electrically stimulated muscle. The static nonlinearity is the model for the "recruitment curve" that describes the relation between strength of a stimulus activation and muscle output. This model was also used in this work.
Valencic has described a measurement system for muscular dynamics based on a displacement sensor placed above the belly of the muscle [Valencic et.al. 1997]. This mechanical approach has the advantage that stimulation artefacts are avoided. Nevertheless sensors for force and displacement need an external reference, which complicates the setup. In this work therefore an acceleration sensor was used in order to make the system smaller and easy to use.
The aim was to develop a fast and simple procedure for estimating the stimulation threshold in clinical practice as a first step for assessment of the complete recruitment curve. The following criteria were to be met: (1) The estimation procedure must be based on a single measurement in order to reduce the time needed for measurements, to keep muscle fatigue low and to reduce discomfort for the patient. Single impulses are preferred. (2) The stimulation and measurement procedures have to be performed synchronously and automatically without any need of inputs from the operators. (3) All data must be stored automatically. The output of the estimation is to be fed back to the stimulation control scaled in stimulation parameters. (4) The estimated stimulation thresholds are intended as an initial starting point for optimisation by the operator.
MATERIAL AND METHODS
Experimental Setup
Acceleration signals were measured during
surface electric stimulation of the quadriceps femoris muscle. Measurements
were performed using the FESDaq measurement system [Sauermann
1999]. Stimulation was
delivered by the stimulation system developed in
The acceleration sensor (EGAS-FS-5, Entran
Sensoren GmbH,
Signal Analysis
The analysis algorithm basically identifies and counts spikes within the measured signal. Spikes below a given signal-to-noise level are discarded, as well as spikes out of a given time window around the stimulation period. Together with the known number of impulses and the stimulation amplitude at the start and the end of the stimulation ramp an amplitude interval is then calculated, that contains the stimulation threshold. The number of spikes found by the algorithm was compared to that found in the acceleration signal by an experienced observer in order to estimate the accuracy of the procedure.
RESULTS
Fig. 1 shown an acceleration signal resulting from the measurements. The algorithm estimated the number of spikes correctly in 15 of 21 trials. It missed a single spike in two cases and detected one wrongly in one case. In the remaining three cases the number of spikes was wrong by more than one. There are three causes of error: (1) spikes are lost in noise as marked in Fig. 2. (2) The decay curve of the last spike is mistakenly classified as an additional spike, as marked in Fig. 2. (3) Due to non-linear signal dynamics the shape of the spikes changes, as in Fig. 1. Therefore peaks appear outside the allowed time window.
All testpersons perceived all stimulation impulses as a tingling sensation on the skin beneath the stimulation electrodes, even at the minimum stimulation amplitudes of 5 V. The number of impulses that were observed visually was not assessed properly in all cases because the operators were not able to count as fast as would be necessary. Repeated stimulation or a video camera would have been necessary to get reliable results. The number of visually perceived muscle twitches was therefore not included in the analysis.
Fig. 1: Acceleration of the skin above the bulk of the quadriceps femoris
muscle during
Fig. 2: Acceleration measured as in Fig. 1. Erroneous detections of spikes: one spike is lost in signal noise, one peak of the decay curve of the last spike is mistakenly classified as a separate spike.
DISCUSSION
During the experiments the automated measurement equipment proved very useful in speeding up the measurement process. The acceleration sensors provided signals of sufficient quality at a reasonable effort for setting them up. The graphical display of the acceleration signal was useful for detecting the stimulation threshold, and more reliable than visual perception: The signal may be reviewed in detail on the screen, with much more time and ease than during the very brief period of the stimulation.
Further work will now be directed at more sophisticated signal analysis based on the dataset of signals harvested in this trial. Especially the non-linear behaviour of the signal dynamics of the acceleration signal has to be dealt with. As regards sensing methods, the necessary amount of cabling has to be reduced significantly and the process of setting up the sensor system has to be eased. It will also be of primary concern to establish the accuracy of the recruitment curve which is necessary to achieve satisfying stimulation outcome with the intended applications.
REFERENCES
/1/ Kilgore, K. L. and Peckham, P. H.: Grasp synthesis for upper-extremity FNS. Part 1. Automated method for synthesising the stimulus map. Med.Biol.Eng Comput., 31, (6), 1993, 607-614.
/2/ Kilgore, K. L., Peckham, P. H., Thrope, G. B., Keith, M. W., and Gallaher-Stone, K. A.: Synthesis of hand grasp using functional neuromuscular stimulation. IEEE Trans.Biomed.Eng, 36, (7), 1989, 761-770.
/3/ Durfee, W. K. and MacLean, K. E.: Methods for estimating isometric recruitment curves of electrically stimulated muscle. IEEE Trans.Biomed.Eng, 36, (7), 1989, 654-667.
/4/ Valencic, V. and Knez, N.: Measuring of skeletal muscles' dynamic properties. Artif.Organs, 21, (3), 1997, 240-242 .
/5/ Sauermann, S.: Computer Based Data Acquisition and Online
Analysis in Functional Electrostimulation. Pages 902-903, in Vol.37, Suppl.2,
Proceedings of the European Medical & Biological Engineering Conference,
Part II, held on
/6/ Bijak, M., Hofer, C., Lanmuller, H., Mayr, W., Sauermann, S., Unger, E., and Kern, H.: Personal computer supported eight channel surface stimulator for paraplegic walking: first results. Artif.Organs, 23, (5), 1999, 424-427.
AUTHOR’S ADDRESS