Standing by means of functional electrical stimulation (FES) after spinal cord injury is a topic widely reported in the neurorehabilitation literature. This practice commonly uses surface stimulation over the quadriceps muscle to evoke knee extension. To date, most FES neuroprostheses still operate without any artificial feedback, meaning that after a fatigue-driven knee buckle event, the stimulation amplitude or pulse width must be increased manually via button presses to re-establish knee-lock. This is often referred to as ‘hand-controlled (HC) operation’. In an attempt to provide a safer, yet clinically practical approach, this study proposed two novel strategies to automate the control of knee extension based on the kinematic feedback of four miniaturised motion sensors. These strategies were compared to the traditional HC strategy on four individuals with complete paraplegia. The standing times observed over multiple trials were in general longer for the automated strategies when compared to HC (0.5–80%). With the automated strategies, three of the subjects tended to need less upper body support over a frame to maintain balance. A stability analysis based on centre of pressure (CoP) measurements also favoured the automated strategies. This analysis also revealed that although FES standing with the assistance of a frame was likely to be safe for the subjects, their stability was still inferior to that of able-bodied individuals. Overall, the unpredictability of knee buckle events could be more effectively controlled by automated FES strategies to re-establish knee-lock when compared to the traditional user-controlled approach, thus demonstrating the safety and clinical efficacy of an automated approach.

One of the most crucial aspects of closed-loop control of upright mobility using functional electrical stimulation (FES)-evoked muscle contractions is the development of a robust control algorithm. The need for compliance with certain safety requirements when FES-evoked movements are deployed to neurological patients is essential, and the computational tools employed play a fundamental role in this task. This paper reports on some new developments regarding closed-loop control of upright mobility and discusses some important issues concerning FES control design. Our laboratory has focused on the restoration of mobility in spinal cord-injured individuals by means of FES-evoked muscle contractions, and currently we deploy a Pocket PC to control an ambulatory neuromuscular stimulator, while movement feedback data is collected from portable motion sensors. Programmed in Matlab Simulink, the control algorithm is structured with sensor feedback (thighs and shanks kinematic data) acting as the model input and the stimulator channels (gluteus, quadriceps and common peroneal nerve) as the model output. In both input and output components, properties such as sampling rate and stimulation parameters are handled. The functions of standing up, sitting down and stepping are designed as finite-state models and grouped into different subsystems, thereby assuring that conditions of ‘worst-case’ system behavior can be identified and treated independently. The operator initiates the changes between states via button presses. The two major control actions of the model’s algorithm are the modulation of quadriceps stimulation to avert knee buckle during stance and the stimulation of the common peroneal nerve to refine swing leg motion during stepping. This control algorithm has proven to meet the differing requirements of both the clinical and research environments, by achieving flexible programming with a fast algorithm development, real-time operation, reliable processing of heavy mathematical calculations and the ability to counteract external disturbances or unexpected behavior.

In paraplegics, gait can be restored by means of functional electrical stimulation (FES). Because the electrophysiological responses of the lower limbs to the neuromuscular stimulus are not completely deterministic, several stimulation strategies have been reported in an attempt to refine stepping motion. In open-loop (OL) systems, the electrical stimulation sequences applied over the leg muscles are often tuned for each patient in order to improve the quality of gait. Our aim was to contrast this traditional technique against variable stimulation sequences based on motion sensors (MS) data feedback. Both strategies were tested over 240 stepping trials in three complete paraplegics. In comparison to OL, which used a customised stimulation sequence for each subject, the same MS strategy was as functional for all three subjects. Despite MS producing a lower variability on step lengths, the toe clearances had a similar pattern of variability regardless of the strategy applied. Although the novel MS showed promising results, the reliability of OL was also demonstrated. Therefore, we still recommend the use of OL mainly due to its faster donning and doffing, since this is a matter of importance for the user acceptance of any rehabilitation systems.

After 20 years of research into functional electrical stimulation (FES) there is still no widely accepted strategy for closed-loop control of gait. This might explain the low number of stroke and spinal cord-injured (SCI) patients who have benefited from this technique, which is still primarily restricted to the research environment. We developed a novel closed-loop strategy that used a finite-state algorithm and motion sensors upon the thighs and shanks to control knee extension during stance, and manage stimulation delivered to the common peroneal nerve (CPN) during the swing-phase of gait. Neuromuscular stimulation was applied by means of skin surface electrodes over the quadriceps muscles, the gluteal muscles and near the CPN. If knee flexion exceeded 10° during stance, then the stimulation amplitude of the quadriceps was increased by a step of 10mA. If knee-locking was achieved (flexion less than 5°), then stimulation amplitude was ramped down by 5mA over 2s in order to reduce the isometric components of quadriceps contractions. During stepping, stimulation of the quadriceps and glutei muscles of the swing leg was switched off, whilst stimulation of the CPN was activated. During leg swing-phase, when knee flexion and thigh flexion were greater than 30° and 25° respectively, CPN stimulation was terminated and quadriceps stimulation ramped up to the previous levels. When these conditions for swing-phase failed, the next step involved CPN stimulation increased by 5mA. Initial experiments performed on two SCI subjects demonstrated significant increases in standing times and decrease of CPN stimulation when compared to the traditional open-loop hand-controlled operation. During walking our system did not require the user to adjust the stimulation levels, simplifying the operation and improving safety. This strategy seemed to successfully ameliorate three significant problems observed during FES gait: early quadriceps fatigue, CPN accommodation to stimulation and inter-individual differences to FES-evoked walking.

Standing after spinal cord injury by means of FES is a topic of interest in neurorehabilitation research. This study compared a handcontrolled (HC) strategy to avert knee-buckle during stance with two event-driven anti-fatigue strategies automated by the use of motion sensors. The control variable modulated was the quadriceps m. stimulation amplitude. The two automated strategies differed in their rate of increases of stimulation when knee unlock was detected. Two ASIA-A subjects participated in standing trials over a standing frame. Surface stimulation was applied to quadriceps and glutei. For both subjects, the total standing times over three trials per day showed significant advantages in using the automated strategies. From total standing times of approximately 10min using HC, the best automated strategy improved this time by 2min15s and 2min51s for subjects S1 and S2, respectively. An automated approach demonstrated efficient control of knee extension, which minimized muscle work and improved user safety.