Using deep learning to predict internalizing problems from brain structure in youth.

Area of Science:

• Neuroscience
• Developmental Psychology
• Machine Learning

Background:

• Internalizing problems, such as anxiety and depression, are linked to negative outcomes.
• Predictors are known, but biological markers for internalizing problems remain unclear.
• Neurodevelopmental (ND) conditions frequently co-occur with internalizing problems.

Purpose of the Study:

• To utilize deep learning for identifying complex brain-behavior relationships.
• To predict cross-sectional and longitudinal trajectories of internalizing problems.
• To explore brain structure as potential biomarkers for internalizing problems.

Main Methods:

• Deep learning models were developed using brain structure measures (thickness, surface area, volume).
• Models predicted clinically significant internalizing problems cross-sectionally (N=14,523) and worsening longitudinal trajectories (N=10,540).
• Data from four large-scale datasets (ABCD, HBN, HCP-D, ONPN) were analyzed using stratified cross-validation and AUC evaluation.

Main Results:

• The cross-sectional model achieved an AUC of 0.80, indicating good predictive performance.
• The longitudinal model showed sub-optimal performance in the general population (AUC=0.66) but good performance in external test sets of ND conditions (AUC=0.80) and across all ND conditions (AUC=0.73).

Conclusions:

• Deep learning models incorporating brain structure show potential as biomarkers for internalizing problems.
• These biomarkers may be particularly valuable for individuals with neurodevelopmental conditions.
• Further research is warranted to refine these models for clinical application.