Cognitive benefits of using non-invasive compared to implantable neural feedback

Abstract

A non-optimal prosthesis integration into an amputee's body schema suggests some important functional and health consequences after lower limb amputation. These include low perception of a prosthesis as a part of the body, experiencing it as heavier than the natural limb, and cognitively exhausting use for users. Invasive approaches, exploiting the surgical implantation of electrodes in residual nerves, improved prosthesis integration by restoring natural and somatotopic sensory feedback in transfemoral amputees. A non-invasive alternative that avoids surgery would reduce costs and shorten certification time, significantly increasing the adoption of such systems. To explore this possibility, we compared results from a non-invasive, electro-cutaneous stimulation system to outcomes observed with the use of implants in above the knee amputees. This non-invasive solution was tested in transfemoral amputees through evaluation of their ability to perceive and recognize touch intensity and locations, or movements of a prosthesis, and its cognitive integration (through dual task performance and perceived prosthesis weight). While this managed to evoke the perception of different locations on the artificial foot, and closures of the leg, it was less performant than invasive solutions. Non-invasive stimulation induced similar improvements in dual motor and cognitive tasks compared to neural feedback. On the other hand, results demonstrate that remapped, evoked sensations are less informative and intuitive than the neural evoked somatotopic sensations. The device therefore fails to improve prosthesis embodiment together with its associated weight perception. This preliminary evaluation meaningfully highlights the drawbacks of non-invasive systems, but also demonstrates benefits when performing multiple tasks at once. Importantly, the improved dual task performance is consistent with invasive devices, taking steps towards the expedited development of a certified device for widespread use.

Figure 1

Non-invasive (NI) and Invasive (I) Sensory Feedback Systems. Real-time electro-cutaneous sensory feedback system composed of: (1) wearable sensors capturing in real-time the force exerted by the subject under the prosthetic foot and the knee flexion/extension of the prosthetic leg. Signals are encoded as stimulation by a microprocessor; (2) NI: surface electrodes placed on the stump and I: electrodes implanted tranversally through the tibial nerve; (3) evoked sensation locations in both the NI and I case.

Figure 2

Electrically evoked sensation characterization. (A) Sensory feedback is mapped to the thigh with (NI1) and the foot sole + gastrocnemius muscle (I1) perceived magnitude. (B) The sensitivity of both NI1 and I1 is shown through their just noticeable difference (JND). (C) In both NI1 and I1, the perceived intensity increases with charge injected. The charge range is on the order of µC and nC for NI and I respectively. (D) Type of sensation and proportion of natural sensations for NI1 and I1 are presented in pie charts. All error bars indicate standard error. *p < 0.05, **p < 0.01, ***p < 0.001. Further results for NI2 and I2 are presented in Supplementary Figure S2A. The right column in panels A, B, C, and D are taken with permission from Preatoni et al..

Figure 3

Passive recognition tasks. (A) Touch task results for three locations (NI/I) and four locations (I) as well as the normalized accuracy of each subject are shown. (B) Proprioceptive task results for two flexion angles (NI) and three flexion angles (I) are shown. (C) Combined touch and proprioceptive task results for a variety of locations and flexion angles are shown. All these results are compared in terms of number of distinguishable levels. A Fisher test was used to compare the results of the individual and combined tasks with their chance levels. ***p < 0.001.

Figure 4

Weight perception and embodiment. subjects performed a weight perception discrimination task before and after a sensory feedback task and no feedback task. Psychometric curves representing their perceived prosthesis weight are presented for NI subjects and I subjects. Statistical significance for specific test weights is marked with *p < 0.05. Self-reported embodiment on a −3 to 3 Likert scale are also shown as graph insets. A comparison between NI and I performance is drawn. The right column is taken with permission from Preatoni et al..

Figure 5

Double task performance. Subjects performed a simultaneous cognitive (spelling words backwards) and motor (walking) task with and without sensory feedback. Their cognitive performance was evaluated through the number of correct letters and their motor performance was evaluated through walking speed. Comparisons between the cognitive and motor improvements of NI and I subjects was also drawn by grouping the paired data together. Statistical significance is marked with *p < 0.05, **p < 0.01, ***p < 0.001. Parts of the right column are taken with permission from Preatoni et al..