Cognoa Demonstrates Advances of its AI-Based Technology for Identifying Autism in Children

Two clinical studies published in peer-reviewed journals validate continued accuracy improvements of the company’s artificial intelligence (AI) platform

PALO ALTO, Calif., Aug. 27, 2018 (GLOBE NEWSWIRE) -- Cognoa, a digital behavioral health company, today announces results from a clinical study of its second-generation AI based algorithms which demonstrates accuracy improvements over its first-generation screening algorithm in supporting the diagnosis of autism spectrum disorder (ASD) in young children. Published in Journal of the American Medical Informatics Association, the data provides the foundation for continued clinical validation of Cognoa’s technology on its path toward becoming a commercially available diagnostic device to be used by physicians.

The multi-center clinical study found Cognoa’s improved second-generation machine learning algorithm was reliable, efficient within clinical workflows and achieved significantly higher accuracy than current standard practice screeners on the same age span. Completed at a sensitivity and specificity of 90 percent and 60 percent, the results demonstrate the potential of AI-based technology to contribute to and improve the process of ASD detection of young children.

“Too many children are still not getting diagnosed early enough to get maximum benefit from therapy,” said John N. Constantino, MD, of Washington University and Cognoa Advisory Board Member. “Cognoa is demonstrating that AI represents an important area of innovation and opportunity. As the incidence of autism has continued to increase and yet the average age of diagnosis has not improved, technologies like Cognoa’s are needed to empower physicians to diagnose ASD sooner and provide every child with the opportunity to benefit from earlier treatment and improved lifetime outcomes.”

Recent studies published by the Centers for Disease Control and Prevention (CDC) indicate that the prevalence of autism has increased to one in 59. However, the median age of diagnosis has not improved in over 15 years. In 2000, the median age of earliest diagnosis was 4.4 to 4.6 years. Currently, on average, children are still being diagnosed at age four or older, beyond the primary window of brain development when interventions have the greatest impact. Cognoa’s algorithm has been clinically validated to screen for autism as early as 18 months of age.

“Families should not have to endure a lengthy and difficult diagnostic process for any aspect of their children’s health,” said Tom Megerian, MD, PhD, Clinical Director of the CHOC Children’s Thompson Autism Center. “Long waiting times place an enormous amount of stress on families and can seriously diminish a child’s ability to reach their full developmental potential. Cognoa’s approach is promising for its potential to enable clinicians and families to take appropriate actions sooner by reducing diagnosis waiting times.”

Cognoa’s first-generation algorithm, developed in 2015 for screening purposes, showed better accuracy across a broader age range than other screening tools in a clinical study published in Autism Research. The platform’s continued improvements, demonstrated by its second-generation AI algorithm developed in 2016 and used in the study published in JAMIA, indicate that Cognoa is advancing its machine learning platform as the basis for a reliable diagnostic tool to improve the timeliness of autism diagnosis within the critical early childhood years.

Cognoa’s platform uses machine learning technology to determine the most predictive data to identify risks for behavioral delays. Cognoa enables the collection and analysis of key data inputs that otherwise would be difficult or impossible for physicians to collect. The machine learning algorithm analyzes this data quickly and accurately to make determinations of a child’s current and future state of behavioral health.

“One of the strengths of machine learning is the ability to evaluate performance and adapt to ensure high accuracy and clinical value on an ongoing basis,” said Dr. Sharief Taraman, who is a clinical informaticist and pediatric neurologist in addition to being the Chief Medical Officer of Cognoa. “We are committed to further advancing our technology, conducting rigorous scientific studies to support the safety and effectiveness of our devices, and working with parents and their physicians with the ultimate goal of increasing access to behavioral health diagnostics for all children during the critical early childhood years.”

Summary of Methodologies and Findings
Published in Autism Research in May 2018, Screening in Toddlers and Preschoolers at Risk for Autism Spectrum Disorder: Evaluating a Novel Mobile-Health Screening Tool, the study compares the performance of four standard autism screening tools with Cognoa’s first generation algorithm in their ability to accurately screen for ASD in children 18-72 months old. 230 children, who were referred to one of three tertiary care centers specializing in diagnosis of ASD, enrolled in this study. Prior to their ASD diagnostic appointment, parents completed the Cognoa assessment. The performance of the Cognoa algorithm was compared against best estimate clinical diagnosis. At 75 percent sensitivity, Cognoa’s tool was 62 percent specific in correctly identifying children with ASD. The Cognoa algorithm significantly outperformed other ASD screeners including the Modified Checklist for Autism in Toddlers Revised with Follow-up, the Social Responsiveness Scale, Second Edition, and the Social Communication Questionnaire, and the Child Behavior Checklist. The study also demonstrates the advantages of Cognoa as a single platform that covers all age ranges, compared to other screening tools that each have their own limitations in either implementation, sensitivity, or specificity.

Published in JAMIA in May 2018, Machine learning approach for early detection of autism by combining questionnaire and home video screening, the combined application of two of Cognoa’s second-generation machine learning algorithms was demonstrated to be reliable, efficient within clinical workflows, and achieved higher accuracy than current standard practice screeners on the same age span. The algorithms utilized information provided by the parent including video examples of the child’s natural behavior at home. In a multi-center clinical study of 162 children aged between 18 and 72 months of age, every child was evaluated with standard screeners including Modified Checklist for Autism in Toddlers (M-CHAT) and standardized research instruments including Autism Diagnostic Observation Schedule (ADOS) with diagnosis ascertained by a licensed healthcare provider using Diagnostic and Statistical Manual of Mental Disorders, 5th Edition, (DSM-5) criteria. Cognoa’s improved algorithms demonstrated statistically significant gains compared to existing screeners and previous algorithms in accuracy and AUC, with sensitivity, and specificity of 90 percent and 60 percent respectively.

About Cognoa
Cognoa is developing AI-based digital diagnostics and personalized therapeutics that are designed to provide accurate, earlier diagnoses and more effective treatments to improve outcomes and lower behavioral healthcare costs. When physicians are empowered to identify behavioral conditions and developmental delays early, children have the opportunity to benefit from interventions within the critical window when they have the greatest potential for improved lifelong outcomes. Cognoa also provides a child development app that is available through employers and empowers parents and caregivers to better support their children’s unique behavioral health and growth. For more information, visit


Contact Data