Closed-loop control of FES standing: How can strategies to reduce knee-buckle be improved ?

Braz GP ¹, Nightingale EJ ¹, Deckers DGM ¹, Armstrong NM ², Lee MJ ¹, Middleton JW ³,
Smith RM ², Davis GM ^1,2

¹ Rehabilitation Research Centre, ² School of Exercise and Sport Science, Faculty of Health Sciences, University of Sydney ³ Royal Rehabilitation Centre, Sydney, Australia

E-mail: g.braz@student.usyd.edu.au

Abstract

The present study analyzed two strategies for controlling knee-buckle during FES-evoked standing – one automated and the other hand-controlled – with the increment and decrement of stimulation amplitude in ramps and steps, respectively. As a control condition, tests with continuous maximal amplitude stimulation were performed. Two ASIA-A SCI subjects participated in multiple trials of timed FES-evoked standing until muscle fatigue. Although the automated strategy provided more stability for knee displacement after a buckle, the hand-controlled paradigm elicited knee lock more effectively. Making a comparison with a previously published study, our results demonstrated that knee lock is achievable without the need of stepping FES amplitude stimulation up to maximum intensities. The analysis of our anti-knee buckle strategies suggested that increments of stimulation amplitude in steps and decrements in steps combined with ramps, may be a promising approach. If automated, the adjustments of the stimulation parameters in real-time would reduce mean values of stimulation amplitude, thereby prolonging standing times with a balanced posture.

1. INTRODUCTION

The success in the design of a control strategy for FES standing requires attention to different details – the reliability of the system, prolonged standing time [1], maintenance of balance and good posture [2], clinical applicability [1,3] and users’ acceptance in the home-care market.

Previous studies have focused on automated and hand-controlled strategies to prevent knee-buckle during standing, and irregardless of how sophisticated they have been, the basic principle was simple – to achieve effective muscle recruitment with the minimal neuromuscular stimulation [4].

The present study analyzed two strategies for averting knee-buckle during FES-evoked standing – one hand-controlled, executing stimulation increment/decrement in discreet steps, and one automated ramping algorithm.

2. METHODS

Figure 1 portrays a schematic description of the two anti-knee buckle strategies.

Figure 1: Block diagrams of the knee-buckle strategies: a) hand-controlled and b) automated.

Quadriceps and glutei were stimulated with surface electrodes, with the latter at 80% of the stimulation amplitude applied to the former. All experiments started with subjects in an upright posture supported by a harness system, and they used parallel bars for balance once standing.

The hand-controlled (HC) strategy incremented and decremented stimulation amplitude in fixed steps of 10mA. Using button presses, a physiotherapist determined when the neuromuscular stimulator should make necessary adjustments to avert knee-buckle. Based on the real-time kinematic data of 4 microprocessor-controlled sensors [5] strapped over shanks and thighs, the automated (AU) strategy used the following control criteria: (i) flexion less than 5º (“knee lock”) – the stimulation ramped down at 5 mA•s^-1; (ii) flexion between 5º and 10º (“extension”) – no changes in stimulation amplitude; (iii) flexion greater than 10º (“knee unlock”) – stimulation ramped up at 10mA•s^-1. For both strategies, trials commenced with amplitude stimulation ramping up to the maximum of the neuromuscular stimulator (180mA) within 2.5s. After 5s, the sensors collected the reference knee angle (180º) and then stimulation ramped down to 90mA during the next 10s. A continuous maximal amplitude (MA) stimulation test was also performed (at 180mA), as a ‘control’ condition. Knee angles were sensor-monitored under all conditions and the trial was terminated when either knee buckled more than 33º. HC and MA had the same initial ramp up as AU, to allow sensors to read the initial knee lock position (180º).

The Neopraxis Exostim^TM FES system [6], comprised an 8-channel neuromuscular stimulator (constant current, with balanced biphasic square wave pulses at 33 Hz and 150μs), a controller (Pocket PC Cassiopeia EG-800) and sensors packs, which sampled knee angles at 10Hz, connected in series to the controller. The anti-knee buckle strategies were programmed in Simulink (Mathworks Inc.) using a specialised toolbox. Codes were built in C++ and transferred to the controller. After each strategy was completed, sensor and stimulation parameter data were downloaded to a PC.

The subjects were two male paraplegics (details in Table 1), with greater than 3 years experience in FES cycling. Each performed 10 trials: 2 MA, 4 AU and 4 HC. Each subject completed 2 trials of the same strategy per day with a standardized recovery of 30 minutes. The order of the tests was randomized (S1: AU, MA, HC, AU, HC; S2: MA, AU, HC, HC, AU).

Table 1: Physical characteristics of the subjects.

Subject	SCI Level	Mass	Height	Post injury	Age
S1	T4 (ASIA-A)	59kg	1.64m	5yr6mo	58yr
S2	T8-T9 (ASIA-A)	69kg	1.75m	11yr2mo	51yr

3. RESULTS

Table 2 presents the standing times normalized against the control condition (MA) for first and second trials, independently. Subject S2 showed a good consistency between the AU and the HC strategies, standing two to three times longer than the MA ‘control’ trial. For S1, the first AU test (AU-1) had a low time in relation to the second (AU-2), possibly because AU-1 was his first experience with FES standing.

Table 2: Standing times normalised in relation to MA for 1^st and 2^nd trials independently. * denotes trial stopped due to muscle spasm.

Mode	Standing Time/ MA Standing Time
	S1		S2
	trial		trial
	1^st	2^nd	1^st	2^nd
MA	1	1	1	1
AU-1	1.79	1.51	2.59	2.36
AU-2	2.69	2.71	2.10	2.11
HC-1	2.47	1.3*	3.10	3.01
HC-2	2.07	5.16	2.94	3.31

Figure 2 shows S1 instantaneous knee angles and stimulation amplitude for one trial of each anti-knee buckle strategy (tests MA, AU-2 and HC-1, all 1^st trials, from Table 2). Although MA demonstrated more knee-lock stability, early local muscle fatigue was observed [1, 4]. AU tended to keep knees in the extension region (5-10º), but the 10 mA•s^-1 ramp-up was not as effective in fully locking the knees when compared to the 10mA steps of HC. Table 3 demonstrates this finding, showing the number of increments that resulted in knee lock and extension with AU and HC strategies.

Table 3: Number of increments in stimulation amplitude that elicited knee lock (flexion < 5º) and knee extension (flexion between 5-10º), indicating the better ability of step increments to evoke knee lock. * - not considered (refer to Table 2).

Mode	Ratio of increments that elicited          knee lock : knee extension
	S1
	1^st trial		2^ndtrial
	Right	Left	Right	Left
1^st AU	0:9	0:11	0:8	0:13
2^nd AU	12:6	4:10	1:15	4:18
1^st HC	5:0	5:0	*	*
2^nd HC	3:0	3:0	4:0	4:1
Mode	S2
	1^st trial		2^nd trial
	Right	Left	Right	Left
1^st AU	8:8	0:20	1:15	0:15
2^nd AU	0:7	0:3	0:19	0:4
1^st HC	4:0	1:3	5:0	3:2
2^nd HC	5:0	3:2	5:0	5:0

4. DISCUSSION AND CONCLUSIONS

The stimulation applied on glutei seemed to benefit balance and posture. However, we noted that the 80% ratio in relation to the quadriceps stimulation amplitude was not high enough for subject S1 (injury level: T4 ASIA-A) who often did not achieve full hip extension. A future embodiment of the AU strategy will make this proportion adjustable, under separate trunk sensor-feedback control.

Figure 2: Knee angles and stimulation amplitude for 3 trials of subject S1: a) MA, b) AU-2, and c) HC-1 from Table 2.

Mulder and colleagues [1], developed a rule-based anti-knee buckling strategy, whereby unlocking triggered an step increment to their stimulator’s maximal amplitude for 200-400ms (evoking knee lock), followed by a step-down and a ramp-down until knee unlocking was detected. Their approach prolonged standing time by 3-5 times the MA duration. Similar results were observed in this study, even with the stimulation amplitudes maintained after increment changes. Further investigations will quantify eventual vertical forces applied over the parallel bars.

The aim of the described strategies was to trial different methods of incrementing stimulation amplitude that would elicit bilateral knee lock. Although ramp increments created smaller knee displacement, step increments were more effective to evoke knee lock, also requiring a fewer number of adjustments (refer to Table 3). The HC strategy showed that step increments after knee buckle did not require a neuromuscular stimulator’s maximal stimulation amplitude, since the increment values of the steps were adequate. Although AU maintained good balance by controlling knee angular displacements, it is appropriate that an automated strategy with step increments use knee velocities as the control variable to provide better knee stabilization after lock (refer to the increased knee velocities in Figure 2c).

The combination of increment in steps followed by decrement in steps and ramps (to decrease mean values of stimulation amplitude) seems to be a promising strategy for the future, especially if automated, where these variables could be adjusted in real-time.

References

[1] Mulder AJ, Boom HB, Hermens HJ, et al., Artificial-reflex stimulation for FES-induced standing with minimum quadriceps force. Medical & Biological Engineering & Computing. 28(5):483-8, 1990.

[2] Crago PE, Lan N, Veltink PH, et al., New control strategies for neuroprosthetic systems. Journal of Rehabilitation Research & Development. 33(2):158-72, 1996.

[3] Bijak M, Rakos M, Hofer C, et al. Stimulation parameter optimisation for FES supported standing up and walking in SCI patients. In: 8th Vienna International Workshop on Functional Electrical Stimulation. Vienna, 2004.

[4] Mulder AJ, Veltink PH, Scheerder COS, et al. Impact of recruitment level on local muscle fatigue: a clinical evaluation. in Advances in External Control of Human Extremities. Dubrovnik, 1990.

[5] Simcox S, Parker S, Davis GM, et al., Performance of orientation sensors for use with a FES mobility system. Journal of Biomechanics. (In press), 2005.

[6] Simcox S, Davis G, Barriskill A, et al., A portable, 8-channel transcutaneous stimulator for paraplegic muscle training and mobility - A technical note. Journal of Rehabilitation Research and Development, vol:(1): p. 41-51, 2004.

Acknowledgements

This research was supported by National Health and Medical Research Council project grant 302013. This work comprises a component of PhD studies of the presenting author, sponsored by the Brazilian Government – CAPES – Brazil.