ARGUS-MDS: automated, quantitative and scalable system for social processes in behavioral health - Argus Cognitive STTR Grant Application
Abstract
Standardized behavioral observation methods are integral to developmental, educational, and behavioral
science research. However, existing observational strategies are too laborious to use in large-scale, intervention
and dissemination trials needed in autism spectrum disorder (ASD). In addition, current observational
strategies do not yield sufficiently quantitative, comparable and granular assessment that could drive the
comparison of therapies in clinical trials or the optimization and personalization of intervention. We are
developing a minimally intrusive medical device technology (“ARGUS-MDS”) to simultaneously
monitor multiple key social and problem behaviors in individuals with ASD and related
neurodevelopmental disorders (NDDs). Our team represents an essential collaboration between
computer and clinical scientists with expertise in artificial intelligence (AI), NDDs, diagnostics, multi-modal
interventions, and psychometrics. We seek support in the form of a Fast Track STTR grant to validate the
psychometric properties of ARGUS-MDS and its ability to provide data on change in target behaviors in early
childhood and school-aged children. This would then support the development of a scalable, digital treatment
progress indicator for behaviors reflecting social, repetitive behavior, and associated symptom profiles in ASD.
In Phase I, video and audio data will be collected during gold-standard diagnostic evaluations individuals with
ASD (n=15). Aim 1.1 will establish quality and clinical validity of ARGUS-MDS algorithms for key social
communication behaviors, while Aim 1.2 will evaluate test-retest reliability of biometric output. Phase I will
show that ARGUS-MDS meets quality metrics for biometric output, validates the clinician-
technician feedback system, and establishes intraclass correlation coefficients for automated
social communication (AutoSC) output. In Phase II our focus shifts to establishing psychometric
properties of derived scores for AutoSC analysis, evaluating convergence with established clinical and
functional measures, and preparing for regulatory filing in Phase III. Aim 2.1. will develop scores from
biometric data through exploratory and confirmatory factor analyses of social communication behaviors. Aim
2.2 evaluates correspondence of AutoSC scores to scores on standardized clinical assessments. Aim 2.3
develops a comprehensive Validation Strategy and executes Analytical Validation, per medical device design
control regulation and FDA guidance. Phase II will develop scores from AutoSC output, evaluate
measurement characteristics of AutoSC scores, reliability & validity of Autos SC scores, and
executes all Analytical Validations per the strategy document and FDA guidance. Phase I and II
milestones will set us up for commercialization in Phase III, including filing for regulatory approval and
product launch. Successful completion of this project will provide a novel, scalable medical device technology
to support objective, automated clinical evaluations of social impairments in ASD and other NDDs.
