Predicting depression risk in early adolescence via multimodal brain imaging

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Depression risk increases during adolescence, with a family history significantly elevating risk in children.
• Early identification of at-risk pre-adolescent children is critical for timely intervention and prevention strategies.

Purpose of the Study:

• To apply advanced machine learning to predict depression risk in pre-adolescent children at a two-year follow-up.
• To identify neuroimaging features that best predict future depression in at-risk youth.

Main Methods:

• Utilized a large longitudinal sample (N=2658) from the Adolescent Brain Cognitive Development (ABCD) Study.
• Employed multimodal neuroimaging data (structural MRI, diffusion tensor imaging, task and rest fMRI) and machine learning.
• Implemented a rigorous leave-one-site-out cross-validation for prediction accuracy assessment.

Main Results:

• All brain features significantly outperformed chance in predicting depression risk.
• Resting-state functional MRI (fMRI) features demonstrated superior classification performance, especially in the high-risk group (N=625) with parental depression history.
• Functional connectome features from rest-fMRI showed better prediction than individual brain region measures.

Conclusions:

• Neuroimaging, particularly resting-state functional connectivity, holds significant potential for identifying biological markers of depression risk in early adolescence.
• Interacting elements within the brain's connectome are crucial for capturing individual psychopathology variability.
• This study advances the identification of early biological risks for depression in population-based samples.