Summary
Autism spectrum disorder (ASD) is a heterogenous disease, making it difficult to develop diagnostic measures and effective treatments. Oregon Health and Science University researchers have developed a novel AI method for identifying ASD subgroups, which could help refine ASD diagnostic criteria and further the study of precision medicine for individuals with ASD.
Technology Overview
The prevalence of ASD is increasing globally, but the variability in symptoms, severity and adaptive behavioral impairments makes the disease difficult to diagnose and treat. The laboratory of Dr. Damien Fair has designed a machine learning model that utilizes the benefits of network analysis to characterize heterogeneity as it pertains to ASD patients. Unlike other predictive modeling approaches, this method allows for the integration of multiple data types, such as behavioral measurements and MRI brain scans, to capture a more complete patient profile for the identification of ASD subpopulations, and is robust to missing data. Testing of this model found it was able to identify ASD patients with 72% accuracy, 82% specificity and 63% sensitivity. The method also measures the proximity of each subject to every other subject, generating a distance matrix between participants, which led to the identification of 3 ASD subgroups in the initial cohort of 47 patients. These subgroup identifications strongly predicted differences in brain activity patterns as measured by MRI, suggesting this method could potentially parse the variation in brain mechanisms affected by ASD.
Publication
Feczko et al. “Subtyping cognitive profiles in Autism Spectrum Disorder using a Functional Random Forest algorithm.” NeuroImage 172(2018): 674-688. Link
Licensing Opportunity
This technology is available for licensing.