A Mobile Game for Domain Adaptation and Deep Learning in Autism Healthcare

点击翻译按钮获取中文摘要

Project Summary Neuropsychiatric disorders are the single greatest cause of disability due to non-communicable disease worldwide, accounting for 14% of the global burden of disease. The current standards of care suffer from subjectivity, inconsistent delivery, and limited access with growing waitlists. New informatics solutions, in particular artificial intelligence (AI) that can port to more ubiquitous mobile health devices and that are not restricted for use in clinical settings, have great potential to complement or even replace aspects of the standards of care. We propose to develop a novel informatics solution for one of the most pressing mental health burdens, autism, which is up in incidence by more than 600% since 1990, among the fastest growing pediatric concerns today, and highly representative of many other neuropsychiatric conditions. We have invented a prototype mobile system called Guess What (guesswhat.stanford.edu) (GW) that turns the focus of the camera on the child through a fluid social engagement with his/her social partner that reinforces prosocial learning while simultaneously measuring the child’s developmental learning progress. At its simplest level, the GW app challenges the child to imitate social and emotion-centric prompts shown on the screen of a smartphone held just above the eyes of the individual with whom the child is playing. But more, as a home-based repeat-use system, GW uses computer vision algorithms and emotion classifiers integrated into gameplay to detect emotion in the child’s face via the phone’s front camera, automatically finding agreement with the displayed prompt, while capturing features such as gaze, eye contact, and joint attention. Preliminary work with more than 20 autistic children resulted in positive user feedback, evidence of high engagement for both the parents and children, and importantly, evidence of clinically meaningful gains in socialization. A single session produces 90 seconds of enriched social video and sensor data, opening up an exciting opportunity for the game play itself to passively generate labeled computer vision libraries that enable the development of better models with higher diagnostic precision going forward. Our proposed project will show that GW can (a.) serve as a mobile therapy that can be used repeatedly by families to target core deficits of autism while inherently tracking progress during use, and, (b.) serve as a distributed system to crowdsource the acquisition of new labeled image libraries for AI models that can automatically classify diagnostic features relevant to autism and extend to other sectors of mental health (and even beyond).