Quantifying Selective Motor Control in Preterm Infants: A Computer Vision Approach

PROJECT SUMMARY Selective motor control (SMC), or isolation of one joint at a time, is difficult for individuals with cerebral palsy (CP). Impaired SMC is the largest contributor to gross and fine motor abilities in people with CP, caused by early damage to the corticospinal tract (CST). The development of SMC cooccurs with organization of the CST, therefore, SMC may be produced more often as the CST myelinates. Given that a reduction in SMC is a key clinical feature of spastic CP in childhood, reduced SMC may also be detectable in young infants with CP. However, little is known about the development of SMC in infants (with and without CP) because SMC has traditionally been quantified by asking a subject for a motor response to a verbal command, excluding infants from traditional measurement methods. Our team recently addressed this limitation by creating an observational measure of SMC, called BabyOSCAR (Observational Selective Control AppRaisal). BabyOSCAR measured at 3 months of age, is valid, reliable, and predictive of future CP diagnosis, body topography of CP distribution and motor function in children with CP. However, it requires observer training, relies on observer interpretation, and provides a binary score for presence or absence of SMC by joint. As a result of these limitations, there is a fundamental need for technology which can quantify SMC and infant joint kinematics, allowing for more precise, accurate, and efficient measurement. Recent advances in computer vision techniques make it possible to quantify motion using video recordings, although these methods are less mature for infants. In this proposal, we will develop a computer vision algorithm for reliably tracking limb kinematics in infants and use this to automate scoring of SMC. This approach is safe and imperceptible to the infant, as it does not require any touch or handling; therefore, we are able to longitudinally measure the evolving movement behavior and joint motions that occur in preterm infants hospitalized in a Neonatal Intensive Care Unit and after they transition home. The ability to study this early and evolving behavior with the precision and accuracy that computer vision allows, will help us to understand fundamental questions about the ontogeny of human movement behavior and allow a non-invasive measure of the health of the CST from which these behaviors are derived. In this proposal we will explore the evolution of these early movement behaviors and compare them with two-year neurodevelopmental outcomes, including the presence of CP. Using this approach, we will be able to determine how the course of SMC and infant kinematics may differ in infants with CP compared to infants without CP. Furthermore, we propose to validate a computer vision algorithm capable of obtaining continuous measures of SMC and accurate joint kinematics using a single camera set up. This algorithm could allow greater access to SMC and kinematic measurement in infancy as it would allow data collection to occur using commonly available smartphones. By harnessing the power of technology to measure infant movement behavior more accurately, we will be able to create more targeted and effective interventions for infants with CP, optimizing their functional abilities.