Journal Article in Journal of Biomechanics
Quantifying individualized rehabilitation responses and optimizing therapy for each person is challenging. For interventions like treadmill training, there are multiple parameters, such as speed or incline, that can be adjusted throughout sessions.
Aim: This study evaluates if causal modeling and Bayesian Additive Regression Trees (BART) can be used to accurately track the direct effects of treadmill training on gait.
Methods: We developed a Directed Acyclic Graph (DAG) to specify the assumed relationship between training input parameters and spatiotemporal outcomes during Short Burst Locomotor Treadmill Training (SBLTT), a therapy designed specifically for children with cerebral palsy (CP). We evaluated outcomes after 24 sessions of SBLTT for simulated datasets of 150 virtual participants and experimental data from four children with CP, ages 4–13 years old. Individual BART models were created from treadmill data of each step.
Results: Simulated datasets demonstrated that BART could accurately identify specified responses to training, including strong correlations for step length progression (R2 = 0.73) and plateaus (R2 = 0.87). Model fit was stronger for participants with less step-to-step variability but did not impact model accuracy. For experimental data, participants’ step lengths increased by 26 ± 13 % after 24 sessions. Using BART to control for speed or incline, we found that step length increased for three participants (direct effect: 13.5 ± 4.5 %), while one participant decreased step length (−11.6 %). SBLTT had minimal effects on step length asymmetry and step width.
Interpretation: Tools such as BART can leverage step-by-step data collected during training for researchers and clinicians to monitor progression, optimize rehabilitation protocols, and inform the causal mechanisms driving individual responses.
KA Ingraham, HA Feldner, KM Steele (2024) “An Instrumented ‘Explorer Mini’ for Quantitative Analysis of Toddlers Using Powered Mobility for Exploratory, Mobile, and Digital Play”
Journal Article in the 10th IEEE RAS EMBS Intl. Conference on Biomedical Robotics and Biomechatronics (BioRob).
For toddlers with disabilities, assistive technologies can enable developmentally appropriate exploration, play, and participation, but little is known about how children interact with accessible interfaces, such as joysticks.
Aim: The Permobil Explorer Mini is currently the only commercially available, FDA-cleared pediatric powered mobility device in the United States designed for children ages 12–36 months. In this paper, we present an instrumented Explorer Mini that enables us to quantitatively analyze how young children with disabilities learn to use and interact with joystick-based technology.
Methods: We discuss preliminary results from two studies conducted with two toddlers with motor disabilities using the instrumented Explorer Mini in different contexts: 1) during exploratory mobile play (i.e., driving) and 2) during interactive digital play (i.e., playing a simple computer game).
Results: In the first study, we found that, for a given 15–20 minute play session, participants drove between 11.3 and 65.6 m, and engaged with the joystick between 53 and 165 times. In the second study, we found that children could use the joystick to play a simple cause-and-effect computer game, but that they disproportionately used the ‘forward’ direction of the joystick, regardless of the direction of the displayed target.
Interpretation: The novel experimental platform, research framework, and preliminary data presented in this paper lay the foundation to study how children with disabilities learn to use and interact with joystick-based assistive technologies. This knowledge is critical to inform the design and advancement of developmentally appropriate technologies that equitably support toddlers in exploration, mobility, and play.
NL Zaino, KA Ingraham, ME Hoffman, HA Feldner, KM Steele (2024) “Quantifying toddler exploration in different postures with powered mobility”
Journal Article in Assistive Technology
Access to powered mobility can support play and development for toddlers with disabilities. Using powered mobility in a standing posture has been theorized to support development of muscle coordination, balance, head and trunk stability, and transition to ambulation.
Aim: The purpose of this study was to quantify and characterize joystick control, bodyweight support, and muscle activity while using the Permobil Explorer Mini in seated and supported standing postures.
Methods: Nine children with mobility disabilities participated in four visits where they completed two, 15–20 minute play sessions, one in each posture, with a break between.
Results: We found that all toddlers engaged with the joystick in both postures, with individual differences in favored directions and control patterns. Participants had similar loading through their feet in both postures, but had slightly higher muscle activity in standing, especially while driving.
Interpretation: These results demonstrate that young children with disabilities quickly engage with joystick-based powered mobility in seated and standing postures, with important individual differences that can inform future design of devices and interventions to support play and development.
MR Ebers, JP Williams, KM Steele, JN Kutz (2024) “Leveraging arbitrary mobile sensor trajectories with shallow recurrent decoder networks for full-state reconstruction,”
Journal Article in IEEE Access
Sensing is a fundamental task for the monitoring, forecasting, and control of complex systems. In many applications, a limited number of sensors are available and must move with the dynamics. Currently, optimal path planning, like Kalman filter estimation, is required to enable sparse mobile sensing for state estimation. However, we show that arbitrary mobile sensor trajectories can be used. By adapting the Shallow REcurrent Decoder (SHRED) network with mobile sensors, their time-history can be used to encode global information of the measured high-dimensional state space.
Aim: We leverage sparse mobile sensor trajectories for full-state estimation, agnostic to sensor path.
Methods: Using modern deep learning architectures, we show that a sequence-to-vector model, such as an LSTM (long, short-term memory) network, with a decoder network, dynamic trajectory information can be mapped to full state-space estimates.
Results: We demonstrate that by leveraging mobile sensor trajectories with shallow recurrent decoder networks, we can train the network (i) to accurately reconstruct the full state space using arbitrary dynamical trajectories of the sensors, (ii) the architecture reduces the variance of the mean-squared error of the reconstruction error in comparison with immobile sensors, and (iii) the architecture also allows for rapid generalization (parameterization of dynamics) for data outside the training set. Moreover, the path of the sensor can be chosen arbitrarily, provided training data for the spatial trajectory of the sensor is available.
Interpretation: The time-history of mobile sensors can be used to encode global information of the measured high-dimensional state space.
AA Portnova-Fahreeva, M Yamagami, A Robert-Gonzalez, J Mankoff, H Feldner, KM Steele (2024) “Accuracy of Video-Based Hand Tracking for People With Upper-Body Disabilities”
Journal Article in IEEE Transactions on Neural Systems and Rehabilitation Engineering
Utilization of hand-tracking cameras, such as Leap, for hand rehabilitation and functional assessments is an innovative approach to providing affordable alternatives for people with disabilities. However, prior to deploying these commercially-available tools, a thorough evaluation of their performance for disabled populations is necessary.
Aim: In this study, we provide an in-depth analysis of the accuracy of Leap’s hand-tracking feature for both individuals with and without upper-body disabilities for common dynamic tasks used in rehabilitation.
Methods: Leap is compared against motion capture with conventional techniques such as signal correlations, mean absolute errors, and digit segment length estimation. We also propose the use of dimensionality reduction techniques, such as Principal Component Analysis (PCA), to capture the complex, high-dimensional signal spaces of the hand.
Results: We found that Leap’s hand-tracking performance did not differ between individuals with and without disabilities, yielding average signal correlations between 0.7-0.9. Both low and high mean absolute errors (between 10-80mm) were observed across participants. Overall, Leap did well with general hand posture tracking, with the largest errors associated with the tracking of the index finger. Leap’s hand model was found to be most inaccurate in the proximal digit segment, underestimating digit lengths with errors as high as 18mm. Using PCA to quantify differences between the high-dimensional spaces of Leap and motion capture showed that high correlations between latent space projections were associated with high accuracy in the original signal space.
Interpretation: These results point to the potential of low-dimensional representations of complex hand movements to support hand rehabilitation and assessment.