Moving towards Rapid and Mobile Detection of Autism

Background: The incidence of autism has increased dramatically over recent years, making this mental disorder one of the greatest public health challenges of our time.  The standard practice of diagnosis is based on behavioral characteristics, as the genome has largely proved intractable for diagnostic purposes. Yet, the most commonly used behavioral instruments take up to 3 hours to administer by a trained specialist, contributing to the substantial delays in diagnosis experienced by many children, who may go undiagnosed and untreated until ages beyond when behavioral therapy would have had more fundamental positive impacts. In an effort to mitigate these challenges, we have developed a machine-learning based system for classification of autism that requires minutes to administer and that can be delivered via mobile technologies.

Objectives:  Our objectives in this study were threefold: (1) To prospectively validate the sensitivity and specificity of a rapid and mobilized method for detection of the core features of autism that combines home video with a parent-assessment report. (2) To assess the feasibility of obtaining parent recorded home videos of quality and content sufficient to detect behaviors consistent with an ASD diagnosis. And, (3) to detect the value of rapid, pre-clinical assessment of ASD for improving patient management at clinical sites.

Methods:  We use machine learning techniques to analyze a large collection of archived score sheets from two of the most commonly used behavioral instruments, the Autism Diagnostic Interview-Revised (ADI-R) and the Autism Diagnostic Observation Schedule (ADOS), in an effort to identify a small subset of behavioral classes that have most value in detection of children with autism. We then applied the resulting behavioral classifiers to over 5000 independent score sheets and several hundred home videos from children both with and without clinical diagnoses of autism to measure the sensitivity and specificity of the classification system overall. Next we administered and tested this system prospectively in a sample of over 100 children to clinically validate the utility of the classification tool and its potential value for patient triage.

Results:  Our classification approach matched the outcomes of the standard instruments in 97% of all autism and 92% of all non-autism cases, including a set of cases with learning delay and clinically challenging symptom presentation. Our results confirm that rapid analysis of home videos strengthens the confidence in classification, and that the method of video scoring can scale to match the size of the risk population. Finally our results demonstrate that pre-clinical screening through a mobilized system could have significant positive impact on the practice of screening and prioritization of the full risk population.

Conclusions: Approaches that enable families to bridge the gap between initial warning signs of developmental delay and clinical diagnosis of autism quickly and effectively are critically needed for the field. Our tool demonstrates the feasibility of pre-clinical assessments and highlights the possibility of using mobile techniques to reduce bottlenecks and reach a larger percentage of the population in need.