AI-Powered Motion Tracking Shows Promise for Faster Autism and ADHD Diagnoses
A novel artificial intelligence (AI) tool leveraging motion-tracking data could significantly accelerate and refine the diagnosis of autism and attention-deficit/hyperactivity disorder (ADHD) in children, according to research published in Scientific Reports on July 8, 2025. The technology offers a potential pathway to overcome lengthy diagnostic delays and subjective assessments currently plaguing the identification of these neurodevelopmental conditions.
Bridging the Diagnostic Gap
Children with neurodivergent disorders often face substantial waits – up to 18 months in some regions, including Indiana – before receiving a formal diagnosis. Current diagnostic methods heavily rely on behavioral observations and surveys, processes that can be both time-consuming and open to interpretation. This new study proposes a more objective and scalable alternative: analyzing movement data collected during a simple reaching task to identify potential indicators of neurodivergence.
“By studying the statistics of the motion fluctuations, invisible to the naked eye, we can assess the severity of a disorder in terms of a new set of biometrics,” explained a lead researcher from Indiana University. “No psychiatrist can currently tell you how serious a condition is.”
How the Technology Works
The study involved participants wearing wireless motion sensors while performing tasks on a touchscreen. These sensors meticulously recorded linear acceleration, angular velocity, and roll-pitch-yaw (RPY) orientation at millisecond resolution. This high-resolution kinematic data was then processed using a long short-term memory deep learning model, trained to categorize participants into four groups: neurotypical, autistic, ADHD, or with both conditions (comorbid autism and ADHD).
Promising Accuracy, Nuances Remain
The deep learning models achieved a mean test accuracy of 71.48% when utilizing all three kinematic signal types. However, performance varied depending on the data source. RPY data proved most accurate individually, with a 67.83% success rate, while linear acceleration data yielded the lowest at 44.44%, and angular velocity at 32.17%. Combining RPY and linear acceleration data resulted in a 71.79% accuracy, surpassing the performance of all three data types combined.
Notably, the tool demonstrated the highest accuracy in differentiating neurotypical individuals from those exhibiting signs of neurodivergence. Identifying children with both autism and ADHD proved more challenging, mirroring the complexities clinicians often encounter with comorbid diagnoses.
Beyond Diagnosis: Quantifying Severity
The research extended beyond simple identification, exploring novel biomarkers – specifically the Fano Factor and Shannon Entropy – derived from the statistical patterns in participants’ subtle movements. These metrics quantified the randomness of movement, which researchers correlated with symptom severity. Children with more pronounced autism or ADHD symptoms exhibited higher entropy and distinct fluctuation patterns in their acceleration data. For instance, participants with more limited functioning autism displayed significantly greater variability in their hand movements compared to those with milder forms of the condition.
A Future of Accessible Screening
While not intended to replace clinical expertise, the authors envision this technology as a valuable triage or screening tool deployable in primary care offices, schools, and telehealth settings, particularly in areas with limited access to specialized care. A data collection session could be completed in as little as 15 minutes, making it suitable for early intervention efforts.
“Some patients will need a significant number of services and specialized treatments,” a researcher stated in a news release. “If, however, the severity of a patient’s disorder is in the middle of the spectrum, their treatments can be more minutely adjusted, will be less demanding and often can be carried out at home, making their care more affordable and easier to carry out.”
The study team anticipates that, with training on larger and more diverse datasets, this deep learning approach will play an increasingly important role in identifying individuals suspected of having a neurodivergent disorder. The growing affordability and accessibility of micro-electromechanical (MEM) sensors – found in smartphones and smartwatches – further enhance the potential for widespread application of this innovative technology.
References
Doctor KP, McKeever C, Wu D, et al. Deep learning diagnosis plus kinematic severity assessments of neurodivergent disorders. Sci Rep. Published online July 8, 2025. doi:10.1038/s41598-025-04294-9
Artificial intelligence used to improve speed and accuracy of autism and ADHD diagnoses. News release. EurekAlert. July 8, 2025. Accessed July 8, 2025. https://www.eurekalert.org/news-releases/1090448
Wu D, José JV, Nurnberger JI, Torres EB. A biomarker characterizing neurodevelopment with applications in autism. Sci Rep. 2018;8(1):614. doi:10.1038/s41598-017-18902-w
