Structural Features Related to Affective Instability Correctly Classify Patients With Borderline Personality...

Area of Science:

• Neuroimaging
• Machine Learning
• Psychiatry

Background:

• Previous morphometric studies on Borderline Personality Disorder (BPD) yielded inconsistent results due to methodological limitations.
• Voxel-level analyses and region of interest approaches may not capture complex structural alterations differentiating BPD from other conditions.

Purpose of the Study:

• To apply Multiple Kernel Learning (MKL), a whole-brain multivariate method, to identify objective structural biomarkers for BPD.
• To differentiate BPD from healthy controls (HC) and patients with bipolar disorder (BD) using structural neuroimaging data.

Main Methods:

• Multiple Kernel Learning (MKL) was applied to structural MRI data from patients with BPD and matched HC.
• MKL was also used to contrast BPD with BD to ensure specificity of identified biomarkers.
• Correlation analysis was performed between identified brain circuits and affective symptoms measured by the Zanarini questionnaire.

Main Results:

• A specific cortico-subcortical circuit, including basal ganglia, amygdala, temporal lobes, and orbitofrontal cortex, correctly classified BPD against HC with 80% accuracy.
• This circuit showed a positive correlation with affective instability scores, suggesting its role in BPD-related affective disturbances.
• Contrasting BPD with BD helped refine the identified circuit, highlighting regions specific to BPD pathophysiology.

Conclusions:

• BPD is characterized by structural anomalies within a specific cortico-subcortical circuit linked to affective instability.
• This identified circuit serves as a potential objective biomarker to discriminate BPD from healthy individuals and other clinical populations like BD.
• The findings support the utility of advanced machine learning techniques in uncovering complex neurobiological patterns in psychiatric disorders.