Reducing the Soleus Stretch Reflex With Conditioning: Exploring Game- and Impedance-Based Biofeedback

Abstract

People with spasticity, i.e., stretch hyperreflexia, have a limited functional independence and mobility. While a broad range of spasticity treatments is available, many treatments are invasive, non-specific, or temporary and might have negative side effects. Operant conditioning of the stretch reflex is a promising non-invasive paradigm with potential long-term sustained effects. Within this conditioning paradigm, seated participants have to reduce the mechanically elicited reflex response using biofeedback of reflex magnitude quantified using electromyography (EMG). Before clinical application of the conditioning paradigm, improvements are needed regarding the time-intensiveness and slow learning curve. Previous studies have shown that gamification of biofeedback can improve participant motivation and long-term engagement. Moreover, quantification of reflex magnitude for biofeedback using reflexive joint impedance may obtain similar effectiveness within fewer sessions. Nine healthy volunteers participated in the study, split in three groups. First, as a reference the "Conventional" group received EMG- and bar-based biofeedback similar to previous research. Second, we explored feasibility of game-based biofeedback with the "Gaming" group receiving EMG- and game-based biofeedback. Third, we explored feasibility of game- and impedance-based biofeedback with the "Impedance" group receiving impedance and game-based biofeedback. Participants completed five baseline sessions (without reflex biofeedback) and six conditioning sessions (with reflex biofeedback). Participants were instructed to reduce reflex magnitude without modulating background activity. The Conventional and Gaming groups showed feasibility of the protocol in 2 and 3 out of 3 participants, respectively. These participants achieved a significant Soleus short-latency (M1) within-session reduction in at least -15% in the 4th-6th conditioning session. None of the Impedance group participants showed any within-session decrease in Soleus reflex magnitude. The feasibility in the EMG- and game-based biofeedback calls for further research on gamification of the conditioning paradigm to obtain improved participant motivation and engagement, while achieving long-term conditioning effects. Before clinical application, the time-intensiveness and slow learning curve of the conditioning paradigm remain an open challenge.

Keywords: electromyography; gamification; operant conditioning; plasticity; system identification.

Figure 1

Overview experimental methodology. (A) Nine participants were split in three groups, all following the same 13 session study schedule (3 times per week). Per session, M_max was obtained using electrical stimulation, followed by 4 blocks with stretch reflexes containing either 25 or 75 feedback instances (7). (B) Stretch reflexes were elicited around the right ankle joint using a robotic manipulator. Participants were seated on an adjustable chair to support a static posture. (C) Dorsiflexion perturbations around the ankle joint elicited a stretch reflex response as visualized in the SOL muscle and torque. For the EMG-based groups, a discrete ramp-and-hold stretch profile was used (7), whereas a continuous pulse-step perturbation profile was applied for the Impedance group (14). (D) EMG-based groups received biofeedback on the SOL EMG, specifically background EMG activity and the short-latency (M1) reflex response (shaded area) (7). (E) The Impedance group received biofeedback on background torque and the estimated reflexive joint impedance gain (G). A mechanical-based methodology using recorded torques and kinematics was used to disentangle this reflexive contribution from the intrinsic contribution with parameters: inertia I, damping B and stiffness K (14).

Figure 2

Biofeedback visualization and timing. For the (blue) Conventional group, a background (all trials) and a reflex (conditioning trials only) bar-graph directly represented current magnitudes. Moreover, a (gray) target area was displayed with the bar color visualizing whether this target was met (green) or not (red) (7). The reflex graph also showed a blue reference line based on average baseline (B1-5) reflex magnitude. The reflex biofeedback (gray-dashed vertical) was coupled to a stretch perturbation, displayed after a short data processing delay. Additionally, the completed number of trials and success rate were displayed. The game-based Gaming (red) and Impedance (green) groups had truck left-right position represent current background magnitude with the (gray) road as target area. Reflex activity controlled the number of bananas in the trunk after each feedback instance, visualized as wobble of the truck. After the wobble, all bananas were retained when the (non-visual) reflex target was met and two bananas would fall out on failure. As a result, the continuous perturbations of the Impedance group were decoupled from the feedback instances.

Figure 3

Steadiness M_max and background activity. Individual participant traces of SOL M_max, and SOL, TA, and torque background activity for acclimatization (A1), baseline (B1-5), and conditioning (C1-6) sessions. All variables were required to remain steady throughout data collection. Each data point reflects the average of all blocks (Block 0–3) within a single session. Conventional and Gaming groups received biofeedback on SOL activity, whereas the Impedance group received biofeedback on torque activity. For all groups, TA activity was required to remain at a resting level (<7.5 μV). Each icon (circle, square, and diamond) per group is linked to an individual participant and consistently used across figures.

Figure 4

SOL M1 reflex results and within-session effect. Individual participant traces of the average conditioned reflex (mean Blocks 1-3) and control reflex (Block 0) per session for acclimatization (A1), baseline (B1-5), and conditioning (C1-6) sessions. The within-session effect is derived from the difference between the conditioned and control reflex within a session. Conventional and Gaming groups received biofeedback on SOL M1 activity, whereas the Impedance group received biofeedback on reflexive impedance gain G. A –15% within-session effect in session C4-6 was defined as success criteria to determine feasibility of the biofeedback method for each participant, see (gray) shaded target area. Each icon (circle, square, and diamond) per group is linked to an individual participant and consistently used across figures.

Figure 5

Within- and across-block correlation of reflexive biofeedback variables. Individual participants are visualized with a different color. Correlation analysis for the Impedance group for Session B1–C6 and Blocks 1–3. The within-block correlations were computed using the averaged measures per feedback instance. The across-block correlations were computed using the averaged measures per blocks. Data was Z-score standardized within-block and within-subject, respectively to allow combination of data over sessions and participants. To improve visualization only 10% of all within-block data points are shown.