Earlier and More Accurate Autism Diagnosis Through Digital Technology and AI

Diagnosing Autism Spectrum Disorder (ASD) in Germany currently takes, on average, until a child reaches seven years of age. This is far too late, as earlier support significantly improves developmental outcomes. Researchers at the German Center for Mental Health (DZPG) are working to fundamentally improve this process through digital technologies.

Prof. Dr. Christine Falter-Wagner, Heisenberg Professor of Clinical Developmental Psychology at LMU University Hospital Munich and a researcher at DZPG, is pursuing an innovative approach. Rather than relying solely on time consuming interviews and subjective behavioral observations, her team uses advanced sensor technology to capture communication behavior and analyzes the resulting data using machine learning methods.

Specifically, multiple channels of communication are recorded simultaneously, including eye movements, facial expressions, body movements, speech patterns, and physiological signals.

How Well Are Two People Synchronized?

A particular focus of the research is interpersonal synchrony, which refers to the degree to which two people are attuned to one another during a conversation. Neurotypical individuals generally demonstrate more finely coordinated interaction patterns than people on the autism spectrum, for example in the alignment of gaze and gestures.

These differences can be detected by AI algorithms with an accuracy of 80 to 90 percent based on just ten minute video recordings.

Another important advantage of the method is its ability to distinguish autism from other mental health conditions with similar symptoms, such as borderline personality disorder, social anxiety disorder, or attention deficit/hyperactivity disorder (ADHD). This type of differentiation remains a major challenge in current diagnostic practice and frequently contributes to misdiagnosis.

Technology as Support, Not a Replacement

Falter-Wagner emphasizes that technology is not intended to replace clinical assessment: “Technology assisted diagnostics cannot replace clinical diagnostics, as the latter also includes developmental history, among other factors. Ultimately, a comprehensive clinical evaluation remains decisive for diagnosis. Nevertheless, technologies can support clinicians in the future and help reduce waiting times for diagnostic assessments.”

Next Steps: Validation and Acceptance Research

The next phase of the project involves a large scale validation study across multiple clinics, funded by the Innovation Fund of the Federal Joint Committee (G-BA).

In addition, a companion study will investigate how individuals with autism and their family members perceive and accept this new form of digital diagnostics.

Original Publication: Koehler JC, Dong MS, Bierlich AM, et al. Machine Learning Classification of Autism Spectrum Disorder Based on Reciprocity in Naturalistic Social Interactions. Translational Psychiatry. 2024;14(1):76. Published February 3, 2024. doi: 10.1038/s41398-024-02802-5.

SOURCE: DZPG