Multi-Channel EEG Analysis For Chronic Pain

Chronic pain is an extremely high degree of neural diseases. This kind of persistent pain is closely connected to the injury of cognitive areas of brain so that reduces the patients'attention, executive, and cognitive performance. Chronic pain makes patients suffer great pain and becomes a serious social problem. Recently, studies to the EEG of chronic pain have been one of the hot and hard issues in the fields of brain and cognitive sciences and clinical care. For the limitation and macroscale in information extraction of EEG by current methods, it is hard to reveal the pathological of the neural systems of chronic pain patients, and can't provide effective treatments to the chronic pain patients. To improve this problem, this thesis investigates the multi-channel EEG signal processing of chronic pain. The main research work is introduced as follows:First, analyzing the current methods of EEG signal processing and the limitation and macroscale of information extraction, this thesis proposes the Parallel Factor Analysis (PARAFAC) algorithms based on multi-channel EEG analysis. Comparing with the current methods, the proposed algorithms can extract more information simultaneously in time-frequency-channel domain from the multi-channel EEG signals. This information would provide more evaluation for the brain and cognitive research and clinical care. Some simulations demonstrate the effective of the algorithms.Second, this thesis proposes the PARAFAC algorithm based on wavelet transform to study the different characteristics of EEG signals between chronic pain rats and control rats under the laser stimuli. By this algorithm, we find that it is effective to process the evoked-EEG and verifies the previous study results, while it conquers the limitation of insufficient information in previous ERP analysis.At last, we study the evoked-EEG of chronic pain patients and healthy people in clinic. By the method of PARAFAC, the information in time-frequency-location domain is extracted. Analyzed by mathematical statistics of t-test, this work finds that the somatosensory cortical responses occurre around 250 ms in both patients and healthy people, and patients have lower neural response frequency than healthy people. We also find that the central and prefrontal regions are predominant in pain processing. Moreover, this thesis investigates the influences of different ages to the response frequency by multiple linear regression. Simulations and analysis demonstrate that PARAFAC is an effective method to reveal the pathologies of chronic pain in clinic.