Discriminating Bipolar Disorder From Major Depression Based on SVM-FoBa: Efficient Feature Selection With Multimodal...

Area of Science:

• Neuroscience
• Psychiatry
• Medical Imaging

Background:

• Differentiating bipolar disorder (BD) and major depressive disorder (MDD) is clinically challenging due to a lack of biomarkers.
• Understanding the underlying pathophysiology and brain alterations in BD and MDD is crucial.
• Neuroimaging studies face challenges with high dimensionality and model interpretability.

Purpose of the Study:

• To develop and validate a novel feature selection method for neuroimaging data.
• To improve the classification accuracy between BD and MDD using multimodal brain imaging.
• To identify brain regions and networks critical for differentiating BD from MDD.

Main Methods:

• A novel feature selection approach, Support Vector Machine with Forward-Backward search (SVM-FoBa), was developed.
• Structural and resting-state functional magnetic resonance imaging (MRI) data from 21 BD, 25 MDD, and 23 healthy controls were analyzed.
• Multimodal data analysis combined anatomical and functional imaging features.

Main Results:

• The SVM-FoBa method achieved a high classification accuracy of 92.1% for distinguishing BD from MDD.
• Discriminative features were identified from both structural and functional MRI data.
• The inferior frontal gyrus, default mode network, and cerebellum were highlighted as key regions for BD-MDD differentiation.
• Functional MRI data provided more discriminative information than structural MRI data.

Conclusions:

• The SVM-FoBa approach is effective for feature selection in neuroimaging studies.
• Multimodal analysis enhances the ability to differentiate between BD and MDD.
• The findings suggest potential neuroimaging biomarkers for differentiating BD and MDD, aiding in clinical diagnosis.