Multivariate Pattern Analysis in Identifying Neuropathic Pain Following Brachial Plexus Avulsion Injury: A PET/CT Study.

BACKGROUND

Neuropathic pain following brachial plexus avulsion injury (BPAI) induces plastic changes in multiple brain regions associated with somatosensory function, pain, or cognition at the group level. The alternation of the whole pattern of resting-state brain activity and the feasibility of a brain imaging, information-based diagnosis of pain following BPAI is poorly investigated.

OBJECTIVES

To investigate whether brain pattern alternation can  identify neuropathic pain from healthy controls at an individual level and the specific regions that can be used as diagnostic neuroimaging biomarkers.

STUDY DESIGN

Controlled animal study.

SETTING

The research took place in the school of rehabilitation science of a university and affiliated hospitals.

METHODS

A total of 48 female Sprague-Dawley rats weighing 180 g-200 g were randomly assigned to either the BPAI group (n = 24) or normal control group (n = 24). A neuropathic pain rat model following BPAI was established in the BPAI group and a mechanical withdrawal threshold (MWT) test was performed to verify the presence of neuropathic pain. Micro-positron emission tomography with [Fluorine-18]-fluoro-2-deoxy-D-glucose (18F-FDG-PET) was used to obtain the whole brain metabolic activity scans. Multivariate pattern analysis (MVPA) was performed with a linear support vector machine (SVM) analysis both in PRoNTo toolbox (based on regions of interests) and SearchlightSearchlight approach (based on voxels within the region).

RESULTS

Compared with baseline status, MWT of the left (intact) forepaw was significantly reduced in the BPAI group (P < 0.001). The accuracy of a whole brain image that correctly discriminated BPAI from normal controls rats was 87.5% with both the PRoNTo toolbox and SearchlightSearchlight method. Pearson's correlation analysis revealed significant positive correlations (P < 0.05) between MWT and the standard taken values of brain regions including the left olfactory nucleus, right entorhinal cortex in the PRoNTo toolbox, and bilateral amygdala, right piriform cortex and right ventral hippocampus in Searchlight method.

LIMITATIONS

The alternation of metabolic connectivity among regions and functional connectivity among different networks were not investigated in the present study.

CONCLUSIONS

Our study indicated that MVPA based on the PET scans of rats' brains  could successfully identify neuropathic pain from health condition at the individual level and predictive regions could potentially be provided as neuroimaging biomarkers for the neuropathic pain following BPAI.