Data Mining And Application Of Pain-evoked Responses

Pain is defined as "an unpleasant sensory and emotional experience associated with actual or potential tissue damage, or described in terms of such damage" by International Association for the Study of Pain (IASP). Thus, pain can be a major symptom in lots of medical conditions, and varies with the appearance, aggravating, ease and freeing of medical conditions, pain assessment act as a significant clinical indicator for monitoring patients and measuring therapy performance. When pain persists despite apparent healing of the body, protective and warning system will translate into pernicious chronic pain, and significantly interfere with individual general functioning and quality of life. At this time, pain assessment act as a main criteria for pain management decision. In fact, pain is a subjective feeling which can be affected by sensation, attention, emotion, conscious state, experience and other factors, its subjectivity impedes objective assessment and scientific diagnosis of pain conditions.Electroencephalography (EEG) is a technique of recording the small electric signal along the scalp, which is usually used for clinical diagnosis and laboratory investigation. Moreover, it is well known for advantages of non-invasive and extremely high spatial density. It is able to acquire direct and reliable electrophysiological index of neurological activity in a brain, and is already widely used in basic and clinical neuroscience research, to understand the neural mechanism of pain processing. As laser beams can selectively activated nociceptor Aδ and C fibers, laser evoked potentials (LEPs) is deemed to be the best tools for character investigation of nociceptive pain.In our research, abundant data was collected from a large sample subjects. We adopted a multivariate linear regression (MVLR) analysis using partial least squares to explore how baseline correction mode affects the time-frequency distribution (event-related desynchroniztion [ERD] and event-related synchroniztion [ERS]) of laser-evoked potentials, with the purpose of finding a non-bias baseline correction method. ERD/ERS of electrocortical signals (e.g., electroencephalogram [EEG] and magnetoencephalogram [MEG]) reflect important aspects of sensory, motor, and cognitive cortical processing. The detection of ERD and ERS relies on time-frequency decomposition of single-trial electrocortical signals, to identify significant stimulus-induced changes in power within specific frequency bands. Typically, these changes are quantified by expressing post-stimulus EEG power as a percentage of change relative to pre-stimulus EEG power. However, expressing post-stimulus EEG power relative to pre-stimulus EEG power entails two important and surprisingly neglected issues. First, it can introduce a significant bias in the estimation of ERD/ERS magnitude. Second, it confuses the contribution of pre-and post-stimulus EEG power. Taking the human electrocortical responses elicited by transient nociceptive stimuli as an example, we demonstrate that expressing ERD/ERS as the average percentage of change calculated at single-trial level introduces a positive bias, resulting in an overestimation of ERS and an underestimation of ERD. This bias can be avoided using a single-trial baseline subtraction approach. Furthermore, given that the variability in ERD/ERS is not only dependent on the variability in post-stimulus power but also on the variability in pre-stimulus power, an estimation of the respective contribution of pre-and post-stimulus EEG variability is needed. This can be achieved using a multivariate linear regression (MVLR) model, which could be optimally estimated using partial least square (PLS) regression, to dissect and quantify the relationship between behavioral variables and pre-and post-stimulus EEG activities. We combined single-trial baseline subtraction approach with partial least square (PLS) regression, and found it could achieve correct detection and quantification of event-related desynchronization (ERD)/synchronization (ERS).Then, we step by step denoised single-trial laser-evoked potential using advanced technique of EEG processing, such as band-pass filter, independent component analysis (ICA), common spatial pattern (CSP) analysis, single-trial extraction and multivariate linear regression (MVLR) analysis. Specifically, first, we canceled irrelevant artifacts in frequency domain. Second, we decomposed the band-filtered64-channel epochs into many independent components, and further denoised them by distinguishing and deleting components of blinking and eye movement. Third, we adopted CSP on baseline group and LEPs group to calculate spatial patterns according to which two groups can be ultimately separated. Processed using these spatial filters, single-trial waveforms furthest reserved LEP and decreased background noise, increased SNR. Fourth, we adopted MVLR analysis using averaged waveforms as templates, to substantially enhance SNR of single-trial LEP.As I mentioned earlier, pain is a subjective first-person experience, and self-report is the gold standard for pain assessment in clinical practice. However, self-report of pain is not available in some vulnerable populations (e.g., patients with disorders of consciousness), which leads to an inadequate or suboptimal treatment of pain. Therefore, the availability of a physiology-based and objective assessment of pain that complements the self-report would be of great importance in various applications. At the end of the fourth step of denoising, MVLR analysis produced single-trial LEP features which captured the variability of amplitudes and latencies. With these features, we adopted a Naive Bayes classifier to discretely predict low and high pain and a multiple linear prediction model to continuously predict the intensity of pain perception from single-trial LEP features, at both within-and cross-individual levels. Our results showed that the proposed approach provided a binary prediction of pain (classification of low pain and high pain)with an accuracy of86.3±8.4%(within-individual) and80.3±8.5%(cross-individual), and a continuous prediction of pain (regression on a continuous scale from0to10)with a mean absolute error of1.031±0.136 (within-individual) and1.821±0.202(cross-individual). Thus, the proposed approach may help establish a fast and reliable tool for automated prediction of pain, which could be potentially adopted in various basic and clinical applications.In short, our research firstly explored the effect of baseline correction method on time-frequency distribution of laser-evoked potential, with a large sample of EEG data, to obtain a non-bias method of baseline correction during time-frequency analysis. Then, we adopted many advanced denoising techniques of EEG processing, further mined features of pain evoked EEG responses. On the basis of this effort, we predicted pain intensity using single-trial LEP features. We may be able to establish an objective in clinical diagnosis to assess the intensity of pain sensation and evaluated the effect of pain treatment, thus helping release patients from pain.