2025 Excellence in Neural Engineering Winner Presentations
Hybrid Format: Sears 439 and zoom
9:00 Gabrielle Labrozzi (Winner of the Oustanding BME student Neural Engineering)
Title: Can feedback control of neuromuscular stimulation based on the progression of the CoM improve gait efficiency and reduce gait variability and asymmetry in individuals paralyzed by SCI or other movement disorders?
Advisor: Ron Triolo
9:30 Sedona Cady (Winner of the Kouros Azar Award in Neural Engineering)
Title: Perceptual, Functional, and Psychosocial Impacts of a Wirelessly Connected
Implanted Sensorimotor Interface for Bidirectional Upper Limb Neuroprostheses
Advisor: Dustin Tyler
Gabrielle Labrozi's Abstract: Walking mobility has consistently been reported as a top priority for improvement by people with a spinal cord injury (SCI). It is an important activity for navigating daily life, enhancing metabolic processes and overall health, and participating in society. Stepping post-SCI is routinely improved with neuromuscular stimulation (NS) applied in a feedforward manner to generate contractions of the otherwise paralyzed or paretic muscles. Despite the ability to initiate and facilitate stepping, feedforward NS paradigms often result in variable, discontinuous, and asymmetrical movements that fail to approach neurotypical (NT) gait. Extensive reliance on the upper extremities is required to remain stable and stimulation can result in rapid muscle fatigue, which both contribute to high metabolic energy expenditure. Overcoming these limitations requires a feedback controller that modulates muscle activations appropriately based on a global control parameter and results in a smooth, continuous, and efficient walking pattern. The center of mass (CoM) reflects whole-body movement in space, where the kinematics in the anteroposterior (AP), mediolateral (ML), and inferiosuperior (IS) directions follow well-defined sinusoidal trajectories during neurotypical walking, and significantly deviate from these patterns with pathologies like those resulting from SCI. As such, it indicates the overall quality of the motor control of walking. Thus, my work is based on the assumption that driving deviations of the CoM observed during pathological gait toward a more NT trajectory should result in more natural movements than feedforward control alone. Thus, my overall research question is: Can closed loop (feedback) control of NS based on the progression of the CoM reduce variability, asymmetry, and inefficiency for individuals paralyzed by SCI or other movement disorders? To answer this question, my first aim focused on characterizing the CoM in all three directions during post-SCI gait (Aim I). I hypothesized that the CoM kinematics could discriminate between at least three ambulation categories and have a 0.70 or greater correlation with established, validated, and routinely applied clinical gait assessments. I collected CoM and six established clinical gait assessments from five individuals with incomplete SCI without stimulation over four experimental sessions each. I then applied the model prescribing CoM kinematics developed by Minetti et al. consisting of a truncated Fourier Series, optimized the single-sine coefficients, and computed their proposed symmetry indices in all three directions. CoM symmetry indices strongly or moderately correlated with clinical measures of gait speed, muscle strength, and gait quality depending on the CoM direction. Two clinical measures had weak correlations across all directions: utilization of assistive devices and balance. These findings suggest that CoM analysis captures the salient features of locomotion, is clinically relevant, has promise for assessing rehabilitation interventions post-SCI, and may be suitable for application in a clinical setting including as a control parameter for stimulated assistance. I recently submitted a first-authored manuscript on these results to the Journal of Spinal Cord Medicine which is still in review. For Aim II, I focused on developing a method to reliably derive CoM kinematics from signals obtained from body-mounted inertial measurement units (IMUs). I hypothesized that data from wearable IMUs can be combined via a bilateral Long Short-Term Memory (biLSTM) neural network to robustly predict the CoM trajectories derived from motion capture data during walking. I trained networks with data from four of five NT participants in a leave-one-subject-out cross validation study. The network estimated the CoM with average root mean square errors (RMSEs) of 1.44cm, 1.15cm, and 0.40cm in the ML, AP, and IS directions respectively. Then, I explored the impact of number and location of IMUs on network prediction accuracy via a principal component analysis. Three to five IMUs located on the legs and trunk were the most promising reduced sensor sets to capture movement of the CoM. The networks accurately estimated the CoM on data from an individual with hemiparesis with the greatest error in the ML direction, which could stem from their gait asymmetry. I published these findings in a first-authored paper in the IEEE Transaction on Neural Systems and Rehabilitation Engineering, which established the feasibility of a practical system for estimating CoM in real time for the purposes of control or clinically assessing rehabilitation outcomes. I am now focusing on automating NS assisted stepping based on real time feedback control of CoM. First, I addressed detecting the phases of the gait cycle and hypothesized that features of the CoM could discriminate between six distinct phases with < 50ms latency, and goodness index < 0.70. I developed a fuzzy logic controller (FLC) to regulate the transition between sub-patterns of stimulation based on data from Aim I. After optimizing the FLC to account for the cyclic nature of walking, in silico testing suggested that the controller had acceptable goodness indices for left (0.174, 0.736, 0.289) and right (0.233, 0.742, and 0.232) early swing, terminal swing, and double support phases, respectively. The higher goodness indices for the left/right terminal swings resulted from a delayed transition response into that phase of 0.12 seconds on average. I am currently implementing the FLC in real-time to produce efficient dynamic stepping in post-SCI gait by automatically triggering steps based on the CoM as a control parameter. I hypothesize that producing consecutive steps based on the CoM features will increase walking speed, and reduce gait variability, asymmetry, and upper body effort by more than 25% over current feedforward systems. I anticipate completing this Aim and submitting the resulting manuscript for publication before the end of 2025. I have participated in team science in Neural Engineering and am a contributing author on another manuscript regarding objectively and automatically adjusting stimulation parameters via in silico optimization techniques that was recently submitted to Medical & Biological Engineering & Computing, and presented posters and abstracts related to my work at BMES, the North American Congress of Biomechanics, and the upcoming Scientific Meeting of the American Spinal Injury Association (ASIA). I have mentored upperclassmen from the Mastery School at Hawken and worked with undergraduate BME students in the biomechanics track. I am currently introducing two first-year doctoral trainees to aspects of human biomechanics and the control of stimulated assistance based on the CoM and have assumed responsibility for orienting them to laboratory instrumentation, analytical methods, and human subject data collection. Through these interactions, I am helping to guide their critical thinking regarding their own research questions. I am also volunteering my time and expertise as a former member of the Women’s Varsity Crew Team at the University of North Carolina to the laboratory-wide effort to establish a Pararowing program for individuals with SCI in Cleveland, and I am a founding member of the RePlay for Kids at Case student organization that adapts and repairs toys for children with physical, developmental, or cognitive disabilities.
Sedona Cady's Abstract:
Individuals with upper limb loss are often dissatisfied with commercial prostheses due to factors
such as unreliable control and lack of sensory feedback, contributing to reduced function and
worsened psychosocial experiences. Implanted nerve stimulation and myoelectric recording
systems offer promising solutions by providing sensation and more advanced prosthetic control.
However, several challenges hinder translation of these technologies, including percutaneous
lead maintenance, bulky external processors, a limited number of implanted electrode channels,
and a narrow focus on either sensory feedback or myoelectric sensing, rather than both features.
Additionally, the benefits of implanted bidirectional neuroprostheses over commercially
available state-of-the-art (SOA) prostheses remain unevaluated.
To create a more complete sensorimotor restoration system, our team developed a nonpercutaneous,
implanted somatosensory electrical neurostimulation and sensing (iSENS) system,
featuring bidirectional stimulation and sensing, wireless communication, and an increased
electrode channel count. The central hypothesis was that iSENS would enhance sensory
perception, myoelectric control, psychosocial experiences, and function for upper limb prosthesis
users.
Initial feasibility testing in two individuals with upper limb loss confirmed iSENS long-term
stability and validated bidirectional 3 degree-of-freedom (DOF) prosthetic control with sensory
feedback. The iSENS system’s increased implanted channel count improved 4 DOF myoelectric
controller performance, expanded sensory location coverage across the hand, and increased the
number of distinct percept locations, enabling diverse sensory feedback. Beyond the laboratory
environment, long-term home use of iSENS with a bidirectional prosthesis led to significant
improvements in psychosocial outcomes, including social interaction and embodiment,
compared to a clinically prescribed SOA prosthesis. Although the tendency to favor the sound
arm over the prosthesis did not change, functional dexterity and activities of daily living
performance improved, likely facilitated by the high DOF prosthetic control enabled by iSENS.
This work validates iSENS as a fully implanted bidirectional neuroprosthetic system that
enhances sensory perception and myoelectric control, thus providing meaningful psychosocial
and functional benefits to prosthesis users. These findings represent significant progress toward
clinical translation of implanted sensorimotor interfaces to overcome limitations of conventional
prostheses and ultimately improve the lives of individuals with limb loss.