Bora Banjanin
Michael Rosenberg


Individuals with neuromuscular impairments, particularly cerebral palsy (CP), are often prescribed ankle foot orthoses (AFOs) to improve gait mechanics and assist locomotion. However, predicting how an AFO will impact an individual’s gait remains challenging. Improving predictive models of how AFOs impact gait mechanics and motor control may inform clinical prescription and design of AFOs. While candidate predictive modeling paradigms in biomechanics exist, the complexity of motor control and kinematic variability during gait suggests a novel approach may be useful to generate predictions from data.

To this effect, our project attempts to model the interaction between the subject and the the AFO using data-driven modeling to predict gait mechanics when walking with AFOs of varying mechanical properties. Our project applies methods derived from control theory to a biomechanics framework, with the end goal of guiding clinical device design and rehabilitation. We first tested the model on a small (25-stride) dataset of AFO walking, and have collected preliminary data of 700 strides of walking data in one healthy adult, which has been used to build a pilot model.

Initial results from a single individual with a unilateral AFO indicate that we can predict a significant portion of the device’s influence on gait. We are conducting further experiments on a more representative set of AFO and patient combinations; a richer dataset will aid in the design and implementation of a standardized training regime for our model. A trained model allows practitioners to evaluate an AFO design for an individual before fabricating a custom AFO. This new approach may offer a valuable clinical tool to improve intervention outcomes providing accurate predictions of AFO impacts on gait kinematics and kinetics.


Mentored by:

Sam Burden
Kat Steele