Gray Matter Differences in Adolescent Psychiatric Inpatients: A Machine Learning Study of Bipolar Disorder and Other...

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Bipolar disorder (BD) is frequently misdiagnosed, leading to delayed treatment.
• Objective biomarkers are needed to improve BD diagnosis and differentiate it from other psychopathology.
• Machine learning on brain structure offers a potential solution for diagnostic accuracy.

Purpose of the Study:

• To develop objective biomarkers for bipolar disorder (BD) using whole-brain gray matter volumes (GMVs).
• To discriminate between BD Type I/II, other specified BD (OSB), other psychopathology (OP), and healthy controls (HC) in adolescents.
• To identify specific GMVs associated with clinical symptoms in pediatric psychiatric populations.

Main Methods:

• Utilized five support vector machine (SVM) models for classification.
• Analyzed GMVs in inpatient adolescents (13-17 years) across BD-I/II (n=34), OSB (n=106), OP (n=52), and HC (n=27) groups.
• Examined discriminative GMVs and their correlation with clinical symptoms.

Main Results:

• Whole-brain classifiers achieved significant accuracy in distinguishing diagnostic groups (e.g., BD-I/II vs. OSB: 79%, AUC=0.70).
• Key discriminative GMVs were identified in brain regions related to movement, sensory processing, and cognitive control.
• Correlations were found between these GMVs and self-reported mania, negative affect, and anxiety across inpatient groups.

Conclusions:

• Pattern recognition models using GMVs can effectively differentiate pediatric BD-I/II from other specified BD and psychopathology.
• GMVs in specific brain regions serve as potential objective biomarkers for BD.
• These findings support enhanced diagnostic accuracy and targeted interventions for pediatric BD.