Granger Causality-Based Pain Classification Using EEG Evoked by Electrical Stimulation Targeting Nociceptive Aδ and C Fibers

Classification of pain levels from evoked electroencephalography (eEEG) has been achieved for the purpose of objective assessment of pain. However, differentiation of acute and chronic pain resulting from the activation of nociceptive nerves, the $\text{A}\delta $ and C fibers, respectively, has remained unsolved, mainly due to the lack of effective features. Granger causality (GC) was applied to show the role of specific brain areas in the differentiation between pain and tactile sensation but yet was used to differentiate nociceptive fiber activations. The purpose of this study is two-fold: first, to develop pain perception classification systems using GC as features, and second, to identify causal connectivity for effective differentiation of nociceptive fiber activations. The eEEG responding to three stimulation waveforms, 1 Hz square waves, 5 Hz sine waves, both targeting C fibers, and 250 Hz sine waves, targeting $\text{A}\delta $ , were recorded. Three GC categories based on channel areas, frequency bands, and stimulation waveforms were computed from the eEEG and used as features to classify nociceptive fiber activations and pain levels. It was found that GCs in the alpha band produced the highest average classification accuracy of 99.71%, followed by the gamma band, with 98.66%. Meanwhile, the GC feature contributing most to differentiating $\text{A}\delta $ and C was found in the alpha band. The GC flow computed from the results of 5 Hz and 250 Hz stimulation had an important role in the prediction of mild pain and maximum pain, respectively. These results demonstrated the high potential of GC as features of pain-related information.