Prediction of anxious depression using multimodal neuroimaging and machine learning

Area of Science:

• Neuroscience
• Psychiatry
• Medical Imaging

Background:

• Anxious depression, a subtype of major depressive disorder (MDD), is linked to poor outcomes and impaired social functioning.
• Understanding the neurobiology of anxious depression is crucial for diagnosis refinement and patient stratification.

Purpose of the Study:

• To investigate the associations between anxiety symptoms and brain structure/function in patients with MDD.
• To develop a neuroimaging-based model for distinguishing anxious from non-anxious MDD patients.

Main Methods:

• Structural and resting-state functional MRI were used to compare gray matter volume (GMV), fractional amplitude of low-frequency fluctuation (fALFF), regional homogeneity (ReHo), and functional connectivity.
• A random forest model was employed to predict anxiety status in MDD patients using neuroimaging data.
• 260 MDD patients and 127 healthy controls participated.

Main Results:

• Anxious MDD patients exhibited significant differences in GMV and ReHo compared to healthy controls.
• Compared to non-anxious MDD patients, anxious MDD patients showed altered GMV, fALFF, ReHo, and functional connectivity in specific brain regions.
• The random forest model achieved an Area Under the Curve (AUC) of 0.802 in distinguishing anxious from non-anxious MDD.

Conclusions:

• Anxious depression is characterized by dysregulation in brain regions involved in emotion regulation, cognition, and decision-making.
• The developed diagnostic model shows potential for accurate and objective clinical diagnosis of anxious depression.
• Neuroimaging biomarkers offer a promising avenue for stratifying MDD patients with anxiety.