Improving Postural Stability among Amputees by Tactile 1 Sensory Substitution

Background For lower-limb amputees, wearing a prosthetic limb helps restore their motor abilities for 25 daily activities. However, the prosthesis's potential benefits are hindered by limited 26 somatosensory feedback from the affected limb and its prosthesis. Previous studies have examined various sensory substitution systems to alleviate this problem; the prominent 28 approach is to convert foot-ground interaction to tactile stimulations. However, positive 29 outcomes for improving amputees' postural stability are still rare. We hypothesize that the 30 intuitive design of tactile signals based on psychophysics shall enhance the feasibility and 31 utility of real-time sensory substitution for lower-limb amputees. were tested with a classical postural stability task in which visual disturbances perturbed their quiet standing. With a brief familiarization of the system, the participants exhibited better posture stability against visual disturbances when switching on sensory substitution than without. The 45 body sway was substantially reduced, as shown in head movements and excursions of the center of pressure. The improvement was present for both amputees and able-bodied controls and was particularly pronounced in more challenging conditions with larger visual disturbances. postural stability for lower-limb intuitive of the mapping the interaction the tactile is the surrogated tactile signals for postural control, for situations that their postural control is


73
The lower-limb amputee lacks direct foot contact with the ground and the feedback from foot 74 mechanoreceptors, critical for balance control (11). With a broken sensorimotor loop, 75 amputees often show poor balance and gait function with fear of falling and a high prevalence 76 of falls (12,13). When an amputee wears a prosthesis, the residue limb of the amputee 77 physically interacts with the prosthetic sockets and provides limited haptic feedback that 78 indirectly reflects foot-ground interaction. Augmenting this essential feedback for prosthesis 79 wearers has the potential to close the sensorimotor control loop and subsequently improve 80 their gait control and postural stability (14, 15). 81 5 Sensory substitution is to encode the missing sensory information and route it to the nervous 82 system via an alternative, intact sensory channels. For example, auditory and haptic feedback 83 has been used to surrogate visual feedback for the blinded to explore the surroundings (16).

84
For upper-limb amputees, sensory substitution has been shown to provide effective sensory 85 feedback for controlling robotic arms (17). Previous researchers have also explored the 86 coding of movement-related information via visual, auditory, or tactile channels for lower-87 limb amputees. For example, Zambarbieri, Schmid (18) used a pressure-sensing insole to 88 estimate the center of pressure (CoP) underneath the foot and visually present the estimate to 89 the participant. This method is apparently impractical since the processing of the surrogated 90 visual information is cognitively demanding and thus limits the benefit of sensory substitution 91 for gait and postural control, which are typically controlled with minimal cognitive load.

92
Other researchers have also used auditory feedback to deliver gait balance information and 93 demonstrated a positive effect on gait asymmetry (19,20 amputated leg with a force magnitude linearly scaled by the pressure measurements from the 103 insole of the prosthesis. They found that, based on the data from a single transtibial amputee, 104 the intensity and the order of pressing forces applied by the balloon actuators could be 105 estimated with decent accuracy (24, 25). However, they did not assess the efficacy of the 106 system in any motor task with prosthesis use. Furthermore, the large size of the balloon 107 actuators might prevent its wide use in the amputee population. Plauché, Villarreal (29) and

148
In the present study, we designed an intuitive tactile stimulation system to provide real-time 149 feedback on plantar pressure. We tested its efficacy in improving postural stability among 150 amputees and the non-disabled. We measured plantar pressure at four insole locations and 151 mapped it nonlinearly to tactile intensity. Critically, to make the learning of sensory 152 substitution easy and intuitive, our system only encodes CoP excursions in the anteroposterior 153 direction, a more critical direction of instability among amputees than other directions (38).

214
To reduce the ambiguity of vibrotactile signals, we only activated one vibrator at a time: 215 when the BI was larger than the EP, the vibrator placed in the front would vibrate to signal a 216 forward lean, and vice versa. The intensity of vibration for each vibrator was determined by 217 the absolute difference in BI between the current state and the equilibrium state at EP: Where BIEP is the average BI estimated at EP when no visual perturbation was applied, and 220 BImax is the maximum BI in the forward or the backward direction estimated from the trials 221 12 when the participants first encountered visual perturbation on day 1 (sensory substitution was 222 off; see below). The relation between the vibration intensity and the BI followed a logarithmic 223 function ( Figure 1B). When the BI slightly oscillated around the equilibrium point as 224 participants maintained a relatively neutral position, the vibrotactile feedback was weak. As 225 the BI deviated more from EP, the intensity would increase, approaching the maximum

523
The oscillation frequency of visual disturbance showed an inconsistent effect on body sway.

524
For example, control participants tend to increase their power of CoP displacement and head 525 movement with increasing stimulus frequency, but amputee participants showed an opposite 526 tendency ( Figure 5 and 7). When visual stimuli moved with a lower frequency (e.g., 0.1 Hz), 527 the body swayed periodically in synchrony with the driving visual stimuli. When visual 528 stimuli moved with a high frequency (e.g., 0.5 Hz), it became hard for the body sway to keep 529 up with the stimuli, resulting in a smaller power (42,56). This saturation effect appears to be 530 more evident for amputees than for non-disabled participants. 531 29 We also computed the performance difference before and after sensory substitution to 532 compare the effect size of sensory substitution across conditions. Three out of the four 533 measures (i.e., the power of CoP displacement, the range and the power of head movement) 534 showed a larger effect size in conditions with larger visual-stimuli amplitudes. The range of 535 CoP displacement, the last measure, did not increase with visual amplitude, but it did increase 536 with visual frequency. Thus, the sensory substitution system benefited both groups of 537 participants more when they were faced with more challenging visual disturbance.

538
We found that sensory substitution stabilized the head and CoP with similar effect sizes. For 539 the CoP range, the effect size of sensory substitution was 0.60 in partial 2 , which is 540 equivalent to a 35.3% reduction after sensory substitution. In comparison, for the head 541 movement range, the effect size was 0.48 with a 24.8% reduction. The same pattern was

545
Furthermore, if we assume that the standing body resembles an inverted pendulum as in 546 typical postural models (58), the head movement should decrease more when the CoP 547 decreases. Thus, theoretically, we shall expect a more significant stabilizing effect of sensory 548 substitution for the head than for the CoP. The lack of difference between the head and the 549 CoP, or even a slightly more significant effect for the CoP, does not fit the theoretical 550 prediction. We postulate that this might be attributed to the specificity of surrogated sensory 551 information delivered by our sensory substitution system: the vibrotactile feedback reflects 552 30 plantar pressure changes directly related to CoP excursion, not to head movement. Thus, 553 when the nervous system integrates this surrogate sensory information, it readily responds to 554 CoP displacement induced by visual disturbances. Therefore, our findings appear to suggest 555 that sensory substitution exerts its influence on motor control in a stimulus-specific way, at 556 least for the situation investigated here where sensory substitution is adopted for a short 557 period of time. Future studies could test this hypothesis by comparing the responses to 558 substituted stimuli that encoded different body motion signals, e.g., head motion instead of 559 CoP displacement.

560
Interestingly, no group difference of postural stability between amputees and the control 561 reached significance for all the performance measures investigated. We expected that 562 amputees would be perturbed more by the visual disturbances since previous studies have 563 shown that amputees are more dependent on visual inputs (39)(40)(41). However, we recognize 564 that these studies used paradigms that reduced visual sensory feedback for the participants.

565
Understandably, it was harder for amputees than the non-disabled to accommodate visual 566 deprivation due to the loss in somatosensory feedback associated with amputation. In the 567 present study, however, we used a visual perturbation paradigm rather than visual deprivation.

568
According to multisensory integration theory in postural control (45-47), both amputees and 569 the non-disabled could adjust the weights of different sensory channels when sensory inputs 570 (i.e., visual input) became inaccurate. Furthermore, previous studies reported worse standing 571 balance among amputees typically used short trials, e.g., 20 s per trial (59). Our experiment 572 instead used as long as 140 s per trial; thus, both groups had ample time to adjust their 573 31 weights of different sensory channels and adapt to the visual stimuli. The other factor is that 574 most of our participants have worn artificial limbs for more than ten years. After prolonged 575 use of prosthesis, their performance in simple motor tasks such as quiet standing become 576 indistinguishable from that of the non-disabled. In sum, the lack of group difference thus 577 suggests that lower-limb amputees can effectively accommodate continuous visual 578 disturbances.

579
The development of robotic artificial limbs has been made dramatic progress in fusing signals 580 from various sensors for sensing the environment and the internal state of the prosthesis, but 581 the research focus is more on intelligent control of prostheses (60). It is equally essential to 582 route real-time sensory feedback for the agent, i.e., the human controller, to reduce the fear of 583 falling, enhance the sense of embodiment of the prosthesis, and better motor control. This 584 sensory augmentation for the agent can be achieved by invasive methods such as electrical 585 peripheral nerve stimulation of the sciatic nerve (61) or noninvasive methods such as sensory 586 substitution. As we pointed out in the introduction, substituting the missing feedback of foot-587 ground interaction is probably most important for lower-limb amputees. Still, the previous 588 endeavors have been hampered by high demands of cognitive loads, unintuitive design, and 589 inconsistent behavioral benefits. Our study has shown that these shortcomings of noninvasive 590 sensory substitution can be overcome. It paves the way for us to integrate this method with 591 robotic lower limbs. As most actuated lower-limb prostheses still lack afferent feedback to the 592 user, it would be interesting to examine the outcome when our sensory substitution system 593 integrates with these systems to achieve better human-centered close-loop control.