Sensory Neuroprosthesis Improves Gait Speed under Cognitive Load in Individuals with Lower Limb Loss

Introduction: Lower limb prostheses are essential for walking after amputation. However, a significant limitation of these devices is the high level of attention or cognitive load (CL) required for their operation due largely to the lack of detailed somatosensory feedback. On the intact limb, various receptors in the skin, joints, and muscles provide information about the limb’s state and its interactions with the floor. In contrast, prostheses only communicate interactions with the environment through the loads applied at the residuum-socket interface. With limited sensation, users must rely primarily on vision to monitor gait progression and stability, resulting in increased mental effort during ambulation.

Neural stimulation can provide context appropriate sensory feedback to individuals with limb loss. Electrical stimulation via peripheral nerve electrodes can elicit sensations perceived as originating from the missing limb. Such approaches have demonstrated benefits in terms of balance, gait symmetry, and metabolic cost. However, the interactions between plantar sensation and the CL of walking after limb loss is not well understood. In this study we attempted to examine the effects of elicited plantar sensation on the CL of walking with a lower limb prosthesis in a dual-task experimental paradigm.

Materials and

Methods: Three participants with unilateral limb loss (LL01, LL02: transtibial; LL03: transfemoral) received high density nerve cuff electrodes (C-FINEs) around the sciatic and/or tibial nerves in the posterior thigh of their residual limb. Neural stimulation delivered via various contacts within the C-FINE produced plantar sensations perceived as originating from distinct areas of their missing limb that could be modulated in magnitude with applied charge. This neural interface was incorporated into a Sensory Neurprosthesis (SNP) that provided context dependent sensations corresponding to the location and magnitude of real-time foot-floor interactions.

The dual-task experiment consisted of self-paced walking on an instrumented split-belt treadmill (VGait, Motek Inc.) and a Color-Word Matching Stroop test (CWM) with the SNP active or inactive. When visually presented with two words, participants responded “yes” or “no” to the question: “Does the color of the upper word match the meaning of the lower word?” After receiving a response, the next stimulus was manually triggered. One CWM trial consisted of neutral (NC), congruent (CC), and incongruent conditions (IC). To highlight the effects of higher CL, in this abstract we present NC and IC results.

CWM responses were recorded for post processing. Reaction times were computed as the difference between a response and the time that stimulus was presented. Gait speeds were collected from the self-paced treadmill. A 2-way ANOVA (CWM task type, SNP condition) followed by Tukey post-hoc tests was conducted, with significance set at 0.05. Despite showing only NC and IC results, the analysis included all conditions for completeness.

Results, Conclusions, and Discussions: All participants maintained high accuracy (> 95%) regardless of task difficulty and SNP condition. Overall, the IC showed increased reaction times compared to the NC for all participants (p < 0.01) (Table 1). These results indicate a higher CL in the IC with longer processing times needed to maintain the same level of accuracy.

The effect of CL on gait speed manifested in two ways (Figure 1). LL02 demonstrated a decrease in speed from NC to IC (p = 0.04) as expected. Counterintuitively, there was an increase in gait speed from the NC to IC for LL01 (p < 0.01) and LL03 (p = 0.01). Notably, all participants achieved faster gait speeds with SNP active: LL01, LL03 (p < 0.05), and LL02 (p < 0.01).

Increasing gait speed with task difficulted was unexpected since most dual-task studies report decreasing gait speeds with heightened CL, as seen with LL02. A possible mechanism behind the higher gait speeds exhibited by LL01 and LL03 may be a shift to a more automatic control of gait to relieve the increased CL. Walking utilizes a combination of lower level spinal circuits and higher cortical processes. Previous dual-task work suggests that increased gait speeds offload the processing burden of walking into the automatic spinal circuitry to free up capacity for other cognitive demands.

This automaticity strategy seems to be employed primarily by LL01 and LL03 whereas LL02 compensates for higher CL by lowering gait speed. Interestingly, the SNP appears to enhance the automaticity benefits regardless of gait speed strategy (p < 0.05).

The CWM task suggests that different strategies could be utilized to maintain accuracy as CL increases during walking. One participant slowed while two increased gait speed, potentially to enhance automaticity in the presence of higher cognitive demand. Interestingly, all participants exhibited higher gait speeds with SNP active regardless of condition. Most importantly, these findings suggest that electrically elicited plantar sensation can modulate resource allocation for ambulation and engage neural circuits responsible for automaticity, improving walking performance under high CL.

Acknowledgements (Optional): This work was supported in part by the Department of Defense under Awards No. W81XWH-18-1-0321 and W81XWH-20-1-0802; and the Merit Review Award # 5 I01 RX003566-02 from the US Department of Veterans Affairs, Rehabilitation Research and Development Service.