Machine Learning-Based Identification of Functional Dysregulation Characteristics in Core Brain Networks of...

Area of Science:

• Neuroscience
• Psychiatry
• Medical Imaging

Background:

• Adolescent bipolar disorder (BD) shares symptoms with other psychiatric conditions, complicating early diagnosis.
• Identifying unique neuroimaging markers is crucial for early differentiation and future research in adolescent BD.
• Task-based functional magnetic resonance imaging (fMRI) offers a potential avenue for detecting neural differences.

Purpose of the Study:

• To identify distinctive neuroimaging markers for adolescent bipolar disorder (BD) using task-based fMRI.
• To explore the utility of machine learning in classifying adolescent BD versus healthy controls.
• To investigate neural alterations across different mood states in adolescent BD.

Main Methods:

• Cross-sectional study involving adolescents with BD and matched healthy controls.
• Emotional Go/NoGo task-based fMRI acquired during depression, mania, and remission states.
• Machine learning classifiers applied to fMRI data to identify predictive neuroimaging features.

Main Results:

• Machine learning, particularly Random Forest (RF), achieved high accuracy (94.29%) in discriminating BD from controls using remission state fMRI data.
• Task-evoked functional alterations were detectable in adolescents with BD even during remission.
• Aberrant activation patterns in the limbic system, prefrontal cortex, and default mode network correlated with clinical and behavioral measures.

Conclusions:

• Adolescents with BD, even in remission, exhibit distinct neural activity in key brain regions.
• These findings suggest potential candidate neuroimaging signatures for adolescent BD.
• Task-fMRI may serve as a valuable tool for understanding and potentially differentiating adolescent BD.