Page 56 - Read Online

Page 205 Jabbari et al. Art Int Surg. 2025;5:200-9 https://dx.doi.org/10.20517/ais.2024.77

promote greater axonal elongation and myelination [45,46] . In this effort, ML modeling can significantly
accelerate biomaterial experimentation by identifying optimal biochemical and biophysical properties from
[47]
large datasets . For instance, Li et al. developed a library of 2,000 peptide-based self-assembling hydrogels
[48]
to identify optimal motifs for hydrogel self-assembly . In another ML model of biomaterial synthesis,
Kosuri et al. discovered chondroitinase ABC complexes that best retained enzymatic activity for neural
[49]
regeneration applications . Such AI-driven advances in NGC and biomaterial design may be applied to
emerging strategies in lower extremity nerve repair and the parallel application of AI technology to nerve
regenerative strategies has potential for revolutionary biotechnologies.

AI in lower extremity prosthetic use and design
The ability to stimulate and record signals from the peripheral nervous system (PNS) is an important
component of new bioelectronic systems. In neurologically intact individuals, sensory signals from the
[50]
lower limbs, such as tactile sensation in the foot and proprioception, influence motor output . Traditional
prostheses do not restore sensory feedback in amputees, which contributes to asymmetric gait, poor
balance, risk of falls, and perception of the prosthesis as an external object (low embodiment) [50-53] . Several
strategies have been employed to restore somatosensory feedback to lower extremity amputees [54-57] . Notably,
advances in PNS interfacing represent a promising alternative to current neuromodulation modalities .
[58]
Direct interface with remaining nerves in the residual limb may restore the sensations necessary for human
locomotion among patients with LLA [59,60] . Charkhkar et al. mapped elicitation sensations in transtibial
amputees with implanted nerve cuff electrodes . Neural stimulation was perceived by patients as
[61]
originating from the missing limb, with discrete localization to missing toes, foot, and ankle, as well as the
residual limb. These findings reflect the paradigm shift in prostheses development, where high-density cuff
technology can be applied to neuroprosthesis with natural sensory feedback [Figure 2]. To this end, AI-
driven methodology can be applied to the evolution of prosthesis development. Koh et al. used CNN to
correlate signals from naturally evoked compound action potentials (CAPs) and neural pathways of
interest . Using a rat model, nerve cuff electrodes were implanted on the sciatic nerve and afferent activity
[62]
was selectively evoked in different fascicles via mechanical stimuli. Based on the predicted firing patterns
from the CNN, a recurrent neural network was used to predict joint angles. They showed high accuracy in
CAP-based classification, which can track physiological measurements such as joint ankles. These results
demonstrate the role of AI in the development of more effective neuroprosthetic systems.

Although promising, the above reports lacked prosthesis connection or functional assessment. This was
addressed by Petrini et al., who utilized intraneural electrodes to develop a leg neuroprosthesis with real-
time tactile and proprioception feedback through nerve stimulation . Functional assessment showed
[63]
improved mobility, fall prevention, and increased embodiment of the prosthesis. It has become evident that
induced sensory feedback integration is an important component of care for LLA patients. As such, there is
a need to optimize neural interface design. Zelechowski et al. developed a computational model of sciatic
nerve behavior in response to electrical stimulation . Their model reported optimal interfaces for use in
[64]
humans and their surgical placement. The authors noted, however, that limitations in imaging technique
and computational power precluded their ability to develop patient-specific devices. Instead, their study
suggests indications for the use and design of these devices. This barrier represents yet another potential
application of AI in the natural evolution of lower limb prostheses.

Osseointegration of prosthetic implants has recently emerged as a viable alternative to traditional socket
prostheses, which are not always suitable for LLA patients . Yet, to our knowledge, the application of AI
[65]
technology to osseointegration strategies is not well studied outside the field of implant dentistry [66,67] . Lu