In a recent study published in JAMA Network Open, researchers from the Karolinska Institutet introduced a new machine learning model designed to predict autism in young children. This innovative algorithm has the potential to enhance early detection of autism, which is essential for delivering timely and effective support.
Image Credit: Prostock-studio/Shutterstock.com
The research team utilized a substantial US database (SPARK) containing data on around 30,000 individuals, both with and without autism spectrum disorders. By examining 28 different parameters, the team developed four distinct machine-learning models to identify patterns in the data. These parameters were chosen because they can be gathered without extensive assessments or medical tests before a child reaches 24 months of age. The most effective model was named 'AutMedAI.'
With an accuracy of almost 80 percent for children under the age of two, we hope that this will be a valuable tool for healthcare.
Kristiina Tammimies, Associate Professor at KIND, the Department of Women's and Children's Health, Karolinska Institutet
Among approximately 12,000 individuals, the AutMedAI model successfully identified around 80 % of children with autism. In particular, certain factors combined with other parameters—such as the age of the first smile, the age at which the first short sentence was spoken, and the presence of eating difficulties—proved to be strong predictors of autism.
The results of the study are significant because they show that it is possible to identify individuals who are likely to have autism from relatively limited and readily available information.
Shyam Rajagopalan, Study First Author and Researcher, Department of Women's and Children's Health, Karolinska Institutet
According to the researchers, early diagnosis is crucial for implementing effective interventions that support optimal development in children with autism.
“This can drastically change the conditions for early diagnosis and interventions, and ultimately improve the quality of life for many individuals and their families,” said Shyam Rajagopalan.
The study found that the AI model performed well in identifying children with more significant challenges in social communication and cognitive abilities, as well as those with broader developmental delays.
Not a Substitute for Clinical Assessment
The research team is now organizing additional enhancements and clinical context validation for the model. Additionally, efforts are being made to incorporate genetic data into the algorithm, potentially yielding even more precise forecasts.
To ensure that the model is reliable enough to be implemented in clinical contexts, rigorous work and careful validation are required. I want to emphasize that our goal is for the model to become a valuable tool for health care, and it is not intended to replace a clinical assessment of autism.
Kristiina Tammimies, Associate Professor and Study Last Author, Department of Women's and Children's Health, Karolinska Institutet
Journal Reference:
Rajagopalan, S. S., et al. (2024) Machine Learning Prediction of Autism Spectrum Disorder From a Minimal Set of Medical and Background Information. JAMA Network Open. doi.org/10.1001/jamanetworkopen.2024.29229.