Multivariate pattern classification of brain white matter connectivity predicts classic trigeminal neuralgia

Area of Science:

• Neuroimaging
• Neurology
• Machine Learning

Background:

• Trigeminal neuralgia (TN) is severe chronic neuropathic facial pain.
• Neuroimaging studies reveal central nervous system gray and white matter (WM) abnormalities in TN.
• Univariate statistics have limitations in analyzing complex neuroimaging data.

Purpose of the Study:

• To test if a multivariate pattern classification method can distinguish abnormal WM connectivity in classic TN from healthy controls (HCs).
• To explore the potential of machine learning in identifying neuroanatomical features of neuropathic pain.

Main Methods:

• Diffusion-weighted scans from 23 right-sided TN patients and matched controls were analyzed.
• Whole-brain interregional streamlines were extracted.
• A linear support vector machine algorithm classified normalized streamline counts to differentiate TN from HCs.

Main Results:

• The algorithm achieved 88% accuracy in differentiating TN from HCs.
• Abnormal WM connectivity patterns involved regions associated with sensory, affective, and cognitive pain dimensions.
• Normalized streamline counts correlated with pain duration and WM metric abnormalities.

Conclusions:

• Machine learning algorithms can detect characteristic structural alterations in TN.
• Structural brain imaging, analyzed with machine learning, can identify neuroanatomical features of neuropathic pain disorders.