Sustained Attention Tasks During MRI Improve Prediction of Autistic Traits

The Search for Objective Biomarkers in Autism Spectrum Disorder

Autism spectrum disorder remains a clinical diagnosis primarily defined by behavioral observations of social communication deficits and restricted, repetitive interests [1]. While functional magnetic resonance imaging has identified potential neural markers, such as local underconnectivity in the posterior cingulate cortex (a key node in the brain network involved in self-referential thought and social cognition), findings across resting-state studies are often inconsistent due to significant phenotypic heterogeneity [2]. Clinicians currently lack objective physiological biomarkers that can reliably stratify patients or predict the severity of social impairments during the pre-symptomatic period [3]. Emerging technologies, including machine learning and advanced connectivity mapping, are being explored to bridge the gap between neurobiological theory and diagnostic practice [4]. A new study now investigates whether specific cognitive challenges during neuroimaging can elicit more reliable markers of autistic traits than traditional resting-state protocols.

Sustained Attention as a High-Yield Brain State

The inherent heterogeneity of autism spectrum disorder has long complicated the search for consistent biological markers that correspond to specific behavioral presentations. While functional magnetic resonance imaging has advanced the general understanding of the neurobiological correlates of autistic features, limited research has focused on identifying the optimal brain states required to reveal these complex brain-phenotype relationships. To address this gap, researchers interrogated four distinct datasets to determine which scanning conditions most effectively improve the prediction of clinically relevant phenotypes. The study utilized connectome-based predictive modeling (a machine-learning method that uses whole-brain functional connectivity patterns, or the synchronized activity between distant brain regions, to predict individual behavioral traits), allowing for a more personalized assessment of neural circuitry. In the first phase of the study, the researchers analyzed dataset one, which consisted of a sample of youth with autism and neurotypical participants (n = 63). The team compared neural data captured during three distinct states: a sustained attention task, a free-viewing social attention task, and a resting-state condition. The findings indicated that the sustained attention task resulted in high prediction performance of autistic traits, providing a clearer signal of the underlying neurobiology than the other conditions. Specifically, the prediction performance during the sustained attention task was higher than during the free-viewing social attention task and also higher than during the resting-state condition. These results suggest that a sustained attention task enhances prediction accuracy for autistic traits compared to other scanning conditions, potentially offering a more reliable method for delineating robust markers of the condition in clinical research.

Cross-Cohort Validation and Generalizability

Validation across diverse populations is the essential litmus test for any potential clinical tool, ensuring that neural markers identified in a discovery group remain reliable in broader patient populations. In dataset two, which consisted of n = 25 participants, the researchers tested the model's ability to identify traits in a different demographic. The results showed that the predictive network model generated from the sustained attention task in dataset one generalized to predict measures of attention in neurotypical adults in dataset two. This finding suggests that the functional connectivity patterns associated with autistic traits are linked to a fundamental neurobiological framework of attention that exists across a spectrum, rather than being isolated to a specific clinical diagnosis. The study further evaluated the model's robustness by applying it to two large-scale, independent cohorts to predict social behavior. Dataset three was the Autism Brain Imaging Data Exchange, which included n = 229 participants, and dataset four was the Healthy Brain Network, consisting of n = 643 participants. The researchers found that the predictive network model generalized to predict measures of social responsiveness in both dataset three (n = 229) and dataset four (n = 643). This successful application across large datasets indicates that the neural signatures of sustained attention are intrinsically connected to the social communication deficits observed in autism. By using data from the n = 229 individuals in the Autism Brain Imaging Data Exchange and the n = 643 individuals in the Healthy Brain Network, the authors demonstrated that these brain-behavior relationships are consistent across different scanning sites and protocols. For the practicing clinician, this evidence suggests that an in-scanner sustained attention task can reveal stable, reproducible markers of social function that resting-state scans may miss.

Clinical Utility of In-Scanner Challenges

The transition from passive resting-state imaging to active cognitive challenges represents a shift toward more functional assessments of neurobiology in a clinical context. While resting-state scans are often easier to collect in pediatric populations, they may lack the sensitivity required to capture the dynamic neural interactions underlying complex social behavior. The researchers utilized connectome-based predictive modeling (a computational method that uses whole-brain functional connectivity patterns to predict individual behavioral or clinical traits) to demonstrate that an in-scanner sustained attention challenge helps delineate robust markers of autistic traits. By requiring participants to engage in a specific cognitive task, the researchers were able to isolate neural signatures that are more closely aligned with the phenotypic expression of autism than those observed during a state of rest or free-viewing social tasks. The clinical relevance of these findings lies in the potential for more objective, biology-based assessments of social responsiveness. In the primary cohort of n = 63 youth, the sustained attention task provided a clearer window into the brain-behavior relationship than either resting-state or social attention tasks. This suggests that the neural circuits involved in maintaining focus are intrinsically linked to the social communication deficits that define the disorder. Because the model successfully generalized across diverse datasets, including n = 229 individuals from the Autism Brain Imaging Data Exchange and n = 643 participants from the Healthy Brain Network, it offers a level of reliability that is often missing in neuroimaging studies. For the practicing physician, this indicates that task-based functional MRI could eventually serve as a tool to quantify the severity of social impairment or to track changes in neural function following therapeutic interventions, providing a more granular characterization of the patient's neurodevelopmental profile.

References

1. Pina-Camacho L, Villero S, Fraguas D, et al. Autism spectrum disorder: does neuroimaging support the DSM-5 proposal for a symptom dyad? A systematic review of functional magnetic resonance imaging and diffusion tensor imaging studies.. Journal of autism and developmental disorders. 2012. doi:10.1007/s10803-011-1360-4

2. Lau WK, Leung M, Lau B. Resting-state abnormalities in Autism Spectrum Disorders: A meta-analysis. Scientific Reports. 2019. doi:10.1038/s41598-019-40427-7

3. Frye RE, Vassall SG, Kaur G, Lewis C, Karim M, Rossignol DA. Emerging biomarkers in autism spectrum disorder: a systematic review. Annals of Translational Medicine. 2019. doi:10.21037/atm.2019.11.53

4. Szkodo MORPM, Micai M, Caruso A, Fulceri F, Fazio M, Scattoni ML. Technologies to support the diagnosis and/or treatment of neurodevelopmental disorders: A systematic review.. Neuroscience and biobehavioral reviews. 2023. doi:10.1016/j.neubiorev.2022.105021