Can MRI Provide Reliable Biomarkers for Autism Spectrum Disorder?

Madeleine Haase

April 07, 2022

The study covered in this summary was published on as a preprint and has not yet been peer reviewed.

Key Takeaways

  • MRI provides an important source of information for the study of autism spectrum disorder (ASD).

  • Predictive MRI biomarkers enable longitudinal follow-ups and prospective epidemiology. 

  • Infants at risk for ASD could be scanned longitudinally, which could allow us to develop early biomarkers useful when behavior is not a sufficient basis for diagnosis.

Why This Matters

  • Autism spectrum disorder is a lifelong neurodevelopmental disorder that affects more than 1% of the population.

  • MRI is an important tool to explore the brain of individuals with ASD, as it is a widely available, fast, and non-invasive method to measure brain anatomy and function.

  • MRI has been extensively used to identify anatomical and functional differences in ASD. However, many of these findings have proven difficult to replicate because studies rely on small cohorts and are built on many complex, undisclosed, analytic choices.

Study Design

  • Researchers launched a data-science prediction challenge, inviting data scientists to submit algorithms to predict ASD diagnostic from provided MRI data. The 10 best submissions were analyzed.

  • The algorithms were first applied to different imaging modalities: only functional MRI, or only anatomical MRI. 

  • The number of available subjects was varied to measure the importance of the sample size. 

  • Lastly, researchers investigated the importance of different brain regions by removing those that appeared as most discriminative and attempting to extract biomarkers from the rest of the data. 

Key Results

  • The combination of the 10 best models provided a good predictor of ASD diagnosis.

  • Used as a screening test, the predictor would correctly detect 88% of the individuals with ASD at the cost of misclassifying 50% of controls as patients.

  • Functional MRI contributed more to prediction than anatomical MRI (area under the curve [AUC] 0.79 using only functional MRI vs AUC 0.66 using only anatomical MRI).

  • Prediction accuracy remained high even after removing up to 50% of the most important brain regions, suggesting that the biomarkers captured information distributed over the entire brain.


  • There were no limitations disclosed in this study.


  • The authors have declared no competing interests.

  • This study did not receive any funding.

This is a summary of a preprint research study, "Insights from an autism imaging biomarker challenge: promises and threats to biomarker discovery," written by Nicolas Traut, University of Paris Department of Neuroscience, and colleagues on, provided to you by Medscape. This study has not yet been peer reviewed. The full text of the study can be found on

For more Medscape Neurology news, join us on Facebook and Twitter


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.