FeetBack-Redirecting touch sensation from a prosthetic hand to the human foot

Abstract

Introduction: Adding sensory feedback to myoelectric prosthetic hands was shown to enhance the user experience in terms of controllability and device embodiment. Often this is realized non-invasively by adding devices, such as actuators or electrodes, within the prosthetic shaft to deliver the desired feedback. However, adding a feedback system in the socket adds more weight, steals valuable space, and may interfere with myoelectric signals. To circumvent said drawbacks we tested for the first time if force feedback from a prosthetic hand could be redirected to another similarly sensitive part of the body: the foot.

Methods: We developed a vibrotactile insole that vibrates depending on the sensed force on the prosthetic fingers. This self-controlled clinical pilot trial included four experienced users of myoelectric prostheses. The participants solved two types of tasks with the artificial hands: 1) sorting objects depending on their plasticity with the feedback insole but without audio-visual feedback, and 2) manipulating fragile, heavy, and delicate objects with and without the feedback insole. The sorting task was evaluated with Goodman-Kruskal's gamma for ranked correlation. The manipulation tasks were assessed by the success rate.

Results: The results from the sorting task with vibrotactile feedback showed a substantial positive effect. The success rates for manipulation tasks with fragile and heavy objects were high under both conditions (feedback on or off, respectively). The manipulation task with delicate objects revealed inferior success with feedback in three of four participants.

Conclusion: We introduced a novel approach to touch sensation in myoelectric prostheses. The results for the sorting task and the manipulation tasks diverged. This is likely linked to the availability of various feedback sources. Our results for redirected feedback to the feet fall in line with previous similar studies that applied feedback to the residual arm.

Clinical trial registration: Name: Sensor Glove and Non-Invasive Vibrotactile Feedback Insole to Improve Hand Prostheses Functions and Embodiment (FeetBack). Date of registration: 23 April 2019. Date the first participant was enrolled: 3 September 2021. ClinicalTrials.gov Identifier: NCT03924310.

Keywords: discrete feedback; grip force; sensory feedback; touch sensation; upper limb prosthesis; vibrotactile insole.

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2022 Morand, Brusa, Schnüriger, Catanzaro, Berli and Koch.

Figures

Figure 1

FeetBack glove and insole. The glove senses the pressure on the fingertips of the index finger and the thumb. The higher force at either finger is defined to represent the grip force and is sent wirelessly to the insole. Depending on the force one of the five embedded coin vibration motors starts vibrating. The lower the force, the lower the number of the running motor (or no motor at all when the force is below a minimal threshold). The distribution of the motors was chosen accordingly to the distribution of fast adapting mechanoreceptors on the sole of the foot.

Figure 2

Setup of the experiments. The FeetBack glove was donned on the cover of the prosthetic hand and the FeetBack insole was worn on the right foot. The participants remained seated throughout the tests. (A) In the object sorting task, the participants were blindfolded and were wearing ear muffs. They rested the arm with the prosthesis on the table with the open hand upward. The investigator put the objects between the index finger and thumb. The participants had to answer if the current object was harder or softer than the previous object. (B) In the pick and place tasks, the participants started in a resting position as shown. After an oral start signal by the investigator, the participants moved the hand from the starting position to the clock and started said clock with the prosthesis in step 1. In step 2, the participants moved the prosthesis from the clock to the object and pinched it. In step 3, they lifted the pinched object from the side of the prosthesis to the contra-lateral target and released the object. In step 4, they stopped the clock with the prosthesis.

Figure 3

Confusion matrix of the object sorting task with feedback. In the ideal case, all answers would lie in the diagonal of the confusion matrix. Since the grading is ordinal, errors become more severe the farther they are from the diagonal. (left) Cumulative matrix with observations from all participants; (right) Individual observations per participant.

Figure 4

Success rates of the object manipulation tasks. All participants performed n=20 manipulations per task under each condition FB-on (feedback switched on) and FB-off (feedback switched off).