Objective, MRI biomarkers for pre-symptomatic detection of autism spectrum disorder at 6 months old: commercial software development and optimization

Project Summary Autism Spectrum Disorder (ASD) is a broad diagnosis for a disorder characterized by symptoms affecting repetitive behavior, social communication, and cognitive ability. 1 in 68 children in the US is affected with ASD (Centers for Disease Control and Prevention, 2017) and the likelihood that a child will be affected with ASD is 10 times higher if they have a sibling with ASD. Traditionally, diagnosis occurs most frequently between age 3 to 6 and severe ASD can be diagnosed as early as 18 months. The infant brain is most plastic and susceptible to intervention in the first year of life. As such, there exists an infant population that is either diagnosed too late to receive a more-effective intervention or with a false negative diagnosis, or both. Thus, earlier prediction of the development of ASD can deliver improved likelihood of favorable outcomes. This Fast-Track proposal seeks to develop, standardize, and commercialize software methods and algorithms for neurobehavioral disorder prediction, creating objective biomarkers in a space dominated by subjective neurobehavioral testing and providing a software as a medical device that is a single, integrated, easy-to-use solution for MRI processing, analysis, and syndrome prediction. To create a Minimum Viable Product (MVP), we will: pull existing, academically vetted methods and software into a commercial design control process (Phase I, SA1); verify the integrated, full-stack solution on a set of IBIS Network data sets stored in the National Database for Autism Research (NDAR) and evaluate potential sources of classification error (Phase I, SA2); and conduct an Alpha release to IBIS Network sites for user interface and human factors feedback (Phase I, SA3).We will build the MVP into a commercially viable product as we: expand the utility of the pipeline to accommodate data sets not acquired at IBIS Network sites, optimize manual Quality Control (QC) workflows through semi-automated user interface design, integrate cortical surface area, cortical volume, functional connectivity, and other measurements into a common user interface and workflow, migrate image processing, machine learning, and database infrastructure to cloud-based tools that can scale on-demand, and iterate on the user experience given Alpha feedback (Phase II, SA1); evaluate synthetic data designed to approximate results from multiple MRI scanner types and signal to noise ratio conditions to ensure broad applicability of the software in clinical settings; expand the utility of the machine learning feature analysis and classification to include additional features and evaluate non-binary feature spaces, conduct latent variable analysis to identify one or more scale metrics for ASD prediction; and vet the new metrics through a thorough review of IBIS Network data and comparison with Alpha release results (Phase III, SA2); and develop a data report for clinical use and patient education, conduct a Beta release test across IBIS Network sites and at least one non-IBIS Network site; and perform formal software verification and validation to prepare for broader distribution of the software platform (Phase III, SA3).Narrative MRI technology is rapidly advancing our understanding of how the brain develops in the presence of neurobehavioral disorders like Autism Spectrum Disorder (ASD). This Fast-Track proposal seeks to develop, standardize, and commercialize software methods and algorithms for ASD prediction using structural and functional MRI data acquired on infants, creating objective biomarkers in a space dominated by subjective neurobehavioral testing and providing a software as a medical device that is a single, integrated, easy-to-use solution for MRI processing, analysis, and prediction of syndrome development. This technology will provide first-of-its kind, objective, brain-based biomarkers that will enable development of effective, early interventions and improve the standard of care for infants at high risk of developing ASD.