P-81 Decoding phantom arm movement using superficial electromyography

Background: After limb amputation, many amputees report being
able to perform voluntary movements with their phantom hand.
Commonly reported voluntary movements of the phantom arm/
hand include reaching for an object, forming a fist and moving one
or more individual fingers. This phenomenon raises the question of
how phantom movements are planned and controlled (Scaliti
et al., 2020). Here we used surface electromyography (sEMG) and
high-density sEMG (HDsEMG) in combination with a decoding
approach to examine the distinctiveness of activity patterns associated
with large set of arm and finger phantom movements.
Objective: (1) Decoding phantom hand and finger movements from
stump sEMG activity and (2) revealing muscular activation patterns
associated with specific phantom movements with high spatial
resolution.
Methods: Unilateral transradial amputees (N=3; mean age±standard
deviation of the mean=43±13) performed three sets of phantom
movements in a fixed order: Hand/wrist phantom movements (including
wrist flexion, wrist extension, close hand, open hand and
pinch with three fingers), flexion of individual phantom fingers,
and extension of individual phantom fingers. After performing the
instructed movement, amputees were asked to hold the position
for 3seconds. We recorded sEMG activity from six superficial electrodes
embedded in a thermoplastic socket customized for each
amputee. To reveal phantom movement sEMG patterns with high
spatial resolution, we also recorded high density sEMG (HDsEMG)
signals from one amputee. sEMG and HDsEMG data were fed into
a Linear Discriminant Anlysis (LDA) to classify phantom movements.
Results: For each amputee, the decoding of phantom movements
from sEMG was well above chance for each set of movements (mean
accuracy±SD=87%±9%, chance level: 20%). HDsEMG revealed movement-
specific spatial activity maps over the extensor and flexor
muscles of the stump, yielding to a classification accuracy of 85%
for 18 movements (chance level: 5.6%).
Conclusions: Our results highlight the potential of sEMG and
HDsEMG for studying phantom movements. Elucidating the relationship
between phantom movements and sEMG not only enhances
our basic understanding of phantom movements but also provides
potential avenues for design of novel, more intuitive prosthetic control
strategies.