CAREER: Learning Visual Representations of Motor Function in Infants as Prodromal Signs for Autism

点击翻译按钮获取中文摘要

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that affects social communication and flexible behavior in up to 1.5% of the population. Movement disorders are considered one of the first signs of ASD, and probably precede social or linguistic abnormalities. But differences in motor movement quality are also associated with other conditions, such as Cerebral Palsy and Developmental Coordination Disorder, so while early motor deficits are not in themselves diagnostic of ASD they are risk indicators. This research will directly aid in early identification of motor deficits in infants, thereby enabling early treatment resulting in better quality of life as well as reduced healthcare and education costs. Additional broad impact will derive from inclusion of an underserved, low-income population with health disparities from Puerto Rico, which will help ameliorate a critical public health research gap (since prior studies have focused overwhelmingly on White, high socio-economic status samples, with only limited relevance to historically marginalized and at-risk communities). In addition, educational activities will engage high school through graduate students to create a pipeline of future data scientists and engineers that democratizes access to broader communities. To maximize impact, project outcomes will be disseminated in peer-reviewed articles, outreach programs, and open code/data repositories.The goal of this work is to establish a computer vision-based, artificial intelligence-guided infant motor function monitoring and assessment system to enable unobtrusive tracking of measures of motor impairment while the infant is in their natural environment. To this end, the research will learn and quantify visual representations of motor function in infants and develop novel data/label-efficient AI techniques, including biomechanically constrained synthetic data augmentation, semantic-aware domain adaptation, and human-AI co-labeling algorithms. The collaboration with the Puerto Rico Testsite for Exploring Contamination Threats (PROTECT) cohort will enable large-scale clinical validation of the extracted measures of early motor function in infants between the ages of 5-10 months and their relationship with the standardized risk screening tests performed at 18 and 24 months of age. A public outreach activity presenting live demonstrations of the developed AI tools will help raise awareness of the importance of AI-guided automatic motor function monitoring early in life.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.