【免费】Anovelapproachtopredictsubjectivepainperceptionfromsingle-triallaser-evokedpotentials资源-CSDN文库

首发论文

需积分: 0 52 浏览量 2020-01-03 00:41:07 上传评论收藏 1.65MB PDF 举报

资源详情

资源评论

http://www.paper.edu.cn

- 1 -

中国科技论文在线

A novel approach to predict subjective pain perception from

single-trial laser-evoked potentials

Xiao Ping

, Huang Gan

, Zhang Zhiguo

, Hu Li

1**

(1. Faculty of Psychology, Southwest University, ChongQing 400715; 5

2. Department of Electrical and Electronic Engineering, The University of Hong Kong,

Hong Kong, China)

Foundations: 高等学校博士学科点专项科研基金（20120182120002）

Brief author introduction:肖平，（1988-），男，硕士研究生在读，主要研究方向：认知神经科学。

Correspondance author: 胡理，（1982-），男，副教授，主要研究方向：认知神经科学. E-mail:

hulitju@gmail.com

Abstract: 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. Here, we aimed

to develop a novel and practice-oriented approach to predict pain perception from single-trial

laser-evoked potentials (LEPs). We applied a novel single-trial analysis approach that combined

common spatial pattern and multiple linear regression to automatically and reliably estimate single-trial

LEP features. Further, we adopted a Naïve 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 of 86.3±8.4 % (within-individual) and 80.3±8.5 % (cross-individual), and a continuous

prediction of pain (regression on a continuous scale from 0 to 10) with a mean absolute error of 1.031 ±

0.136 (within-individual) and 1.821 ± 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.

Keywords: Physiology; Pain; Laser-evoked potentials (LEPs); Pain prediction; Classification;

Regression

0 Introduction

Pain is an unpleasant multidimensional experience associated with real or potential tissue

damage [1] . Therefore, pain experience does not simply reflect sensory information but can be 30

substantially influenced by various psycho-social contexts (e.g., the gender of experimenter) [2]

and psycho-physiological factors (e.g., fluctuations in vigilance and attention). Since pain is a

subjective first-person experience, self-report (e.g., Visual Analogue Scales [VAS] and Numeric

Rating Scales) is the gold standard for the determination of the presence, absence, and intensity of

pain perception in clinical practice [3,4]. Lack or any inaccuracy of pain assessment can lead to 35

inadequate or suboptimal treatment of pain in these vulnerable patients, which may lead to various

additional clinical problems (e.g., psychological distress or depression, the development of chronic

pain) [5]. Therefore, the availability of a physiology-based and objective assessment of pain that

complements the self-report of pain would be of great importance in basic and clinical

applications. The strong relationship between the N2 and P2 amplitudes in LEPs and the intensity 40

of pain perception have been well characterized [6-9], and the correlations between N2 and P2

latencies and the intensity of pain perception were also reported [8].The aim of the present study

was to develop a novel approach to rapidly and reliably predict pain from single-trial LEP features

(Fig. 1), which can be achieved through the following two major steps. First, a novel method that

combines common spatial pattern (CSP) and multiple linear regression (MLR) was proposed to 45

achieve an automated and reliable single-trial estimate of LEP features. Second, a Naïve Bayes

http://www.paper.edu.cn

- 2 -

中国科技论文在线

classifier and a multiple linear prediction model were trained to respectively distinguish low and

high pain and predict the intensity level of pain perception from single-trial LEP features. Such

training and predicting were achieved at both within-individual level, where the classifier and

prediction model were trained on and applied to single-trial LEPs from the same individual, and 50

cross-individual level, where the classifier and prediction model were trained on a cohort of

individuals and applied to another individual.

1 Materials and Methods

1.1 Experimental design and EEG recording

Twenty-nine healthy participants (9 females and 20 males) aged 17-25 years (mean 22.2 ± 55

1.9), without a history of chronic pain, participated in the study. All participants gave written

informed consent, and the local ethics committee approved the experimental procedures.

Radiant-heat stimuli were generated by an infrared neodymium yttrium aluminium perovskite

(Nd:YAP) laser with a wavelength of 1.34 μm (Electronical Engineering, Italy). Laser pulses

activate directly nociceptive terminals in the most superficial skin layers [10,11]. Laser pulses 60

were directed at the dorsum of left hand on a squared area (5 x 5 cm) defined prior to the

beginning of the experimental session. A He-Ne laser pointed to the area to be stimulated. The

laser pulse was transmitted via an optic fibre and its diameter was set at approximately 7 mm (~38

mm2) by focusing lenses. The pulse duration was 4 ms, and four different energies (E1: 2.5 J; E2:

3 J; E3: 3.5 J; E4: 4 J) of stimulation were used. After each stimulus, the laser beam target was 65

shifted by approximately 1 cm in a random direction to avoid nociceptor fatigue or sensitization.

Prior to data collection, a small number of laser pulses with different stimulus energies were

imposed on the stimulated territory to ensure that the participants were familiar with the

stimulation. Ten laser pulses at each of the four stimulus energies (E1-E4) were delivered, in

random order, for a total of 40 pulses per participant (Fig. 2, top panel). The inter-stimulus 70

interval (ISI) was ranged between 10 and 15 s. An auditory tone was delivered 3~6 s after the

presentation of each laser pulse to remind the participants to rate the intensity of the painful

sensation elicited by the laser stimulus, using a VAS ranging from 0 (not pain) to 10 (pain as bad

as it could be).

Participants were seated in a comfortable chair in a silent, temperature-controlled room. They 75

wore protective goggles and were asked to focus their attention on the stimuli and relax their

muscles. The EEG data were recorded using a 64-channel Brain Products system (Brain Products

GmbH, Munich, Germany; pass band: 0.01–100 Hz; sampling rate: 1,000 Hz) using a standard

EEG cap based on the extended 10–20 system. The nose was used as the reference channel, and all

channel impedances were kept lower than 10 kΩ. To monitor ocular movements and eye blinks, 80

electro-oculographic signals were simultaneously recorded from 4 surface electrodes: one pair

placed over the upper and lower eyelids, and the other pair placed 1 cm lateral to the outer corner

of the left and right orbits.

1.2 Data analysis: single trial feature extraction

Since the magnitude of laser-evoked brain responses is often several factors smaller than the 85

magnitude of background EEG and non-cortical artifacts [12], the prediction of pain perception

from single-trial LEPs would rely on advanced signal processing techniques to (1) enhance greatly

the signal-to-noise ratio (SNR) and (2) estimate automatically and reliably single-trial LEP

features. Thus, we proposed to enhance the SNR of single-trial LEPs using the combination of (1)

bandpass filtering (BPF), (2) independent component analysis (ICA), and (3) common spatial 90

http://www.paper.edu.cn

- 3 -

中国科技论文在线

pattern (CSP), and to estimate single-trial LEP features using (4) multiple linear regression (MLR)

(Fig. 3, top panel). BPF and ICA were applied to perform a filter on EEG recordings in the

spectral and spatial domains respectively, to remove irrelevant noise and artifacts. CSP was

adopted to further enhance the SNR of LEP responses by retrieving stimulus-evoked EEG

responses from spontaneous EEG activity. Finally, MLR was used to automatically parameterize 95

single-trial LEP responses using MLR coefficients, which captured the variability of single-trial

N2 and P2 latency and amplitude. The four analytic steps were detailed in the following

subsections.

1.2.1 Bandpass filtering and independent component analysis

To provide a filter in the spectral domain, continuous EEG data were bandpass filtered 100

between 1 and 30 Hz (zero-phase digital filtering) [13]. EEG trials were extracted using an

analysis window of 1,500 ms (500 ms pre-stimulus and 1,000 ms post-stimulus), and baseline

corrected using the pre-stimulus time interval.

Following, EEG trials contaminated by eye-blinks and movements were corrected using the

ICA algorithm [13-15]. As described in [16], artifacts related to eye-blinks and movements could 105

be automatically identified based on the combination of stereotyped artifact-specific spatial and

temporal features. In all datasets, the independent components having a large EOG channel

contribution and a frontal scalp distribution were removed, and the remaining independent

components were used to reconstruct the denoised EEG trials.

1.2.2 Common spatial pattern 110

CSP is a mathematical tool to decompose two populations of multivariate signals into a set of

spatial patterns, which maximize their differences in terms of variance [17,18]. This method was

first applied in the field of EEG analysis to detect abnormalities [19,20], and has recently shown to

be a powerful technique in brain-computer interface research for discriminating different mental

intentions [21]. In the present study, one population consisted of EEG recordings before the 115

presentation of laser stimulation (pre-stimulus EEG activity: -500 to 0 ms), and the other

population consisted of EEG recordings after the presentation of laser stimulation of the same trial

(post-stimulus EEG activity: 0 to 500 ms, where the Aδ related LEP responses were dominantly

observed). The pre- and post-stimulus EEG activities over all channels of the same trial formed

two matrices

postpre

∈ RXX ,

, where N is the number of channels and T is the number of 120

samples in each trial. To maximize the difference between the variance of

pre

and

post

CSP is objective to estimate the generalized eigenvector or the projection vector

w by solving

the generalized eigenvalue problem as follows:

wXXwXX

prepre

postpost

(1)

where ۦ·ۧ is the averaging operator across trials for the same population and

is the

generalized eigenvalue. The matrix

∈= R],,[

wwW L

, where

ww ,,

are N 125

eigenvectors estimated from Equation (1), is the spatial filter, and

NN×−

∈= R

WA is the

spatial pattern represented as a weighting of EEG channels. Importantly, CSP provided an ordered

http://www.paper.edu.cn

- 4 -

中国科技论文在线

list of spatial patterns according to the discriminative power between two populations. As a result,

the spatial pattern with the maximal variance of

post

would capture the minimal variance of

pre

, and vice versa. Typically, only a few spatial patterns were sufficient for the discrimination 130

between two populations [17]. These few spatial patterns, which can isolate stimulus-evoked EEG

responses (contained only in

post

) from spontaneous EEG activity (contained in

post

as well

pre

), worked as an effective spatial filter. In the present study, three eigenvectors

corresponding to the largest eigenvalues of

post

and lowest eigenvalues of

pre

were

selected to reconstruct spatial-filtered single-trial EEG responses of all channels (see Section 1 in 135

the Supplementary materials for the reason of selecting three CSP components).

1.2.3 Multiple linear regression

To estimate single-trial LEP features automatically and reliably, we applied a multiple linear

regression method [22] to LEP trials recorded at Cz within the post-stimulus time interval (0 to

500 ms), which were filtered using the combination of BPF, ICA, and CSP. The MLR approach 140

takes into account the variability of both N2 and P2 latency and amplitude of LEPs, and models

single-trial LEP responses as follows:

(

)

(

)

(

)

=+++

NN N PP P

taytl aytl

(2)

where

(

)

t is a modeled single-trial LEP waveform that varies as a function of time t ,

)(ty

and

)(ty

are the templates of N2 and P2 waves, which can generally be obtained as

the average of all LEP trials of each participant,

and

are the weights of N2 and P2 145

templates, and

and

are the latency shift values of the N2 and P2 templates, respectively.

Since the N2 and P2 peaks of the LEPs reflect the activity of the different neural generators [5],

and their amplitudes can be differentially modulated by several experimental factors (e.g., spatial

attention and probability of perception) [23,24], we modeled the N2 and P2 waves separately, thus

avoiding the assumption that all generators contributing to the LEP responses covary linearly [25]. 150

Using the Taylor expansion, the MLR model can be written as follows:

(

)

(

)

(

)

(

)

(

)

() () () ()

ββ ββ β

≈+ ++ +

=+ ++ +

NN NN N PP PP P

1N 2 N 3P 4 P 5

ft ay t lay' t ay t lay' t

yt y't yt y't

(3)

where

(

)

and

()

y' t

are the temporal derivatives of N2 and P2 templates, respectively,

and

is the residual term. Thus the single trial LEP waveform is approximated using the sum of

the weighted averages of N2 and P2 templates and their respective temporal derivatives.

Considering that these weights (

) captured the single-trial variability of N2 155

and P2 latency and amplitude in LEPs, these coefficients could be closely related to the subjective

intensity of pain perception. Correlations between these estimated single-trial coefficients and the

corresponding single-trial ratings of pain perception were measured using Pearson’s correlation

coefficient for each participant [25]. The obtained correlation coefficients were transformed to Z

values using the Fisher R-to-Z transformation and were finally compared against zero using a 160

one-sample t-test.

1.2.4 Performance evaluation

To quantitatively assess the performance of each analysis step (BPF, ICA, CSP, and MLR)

剩余19页未读，继续阅读

评论收藏

内容反馈

A novel approach to predict subjective pain perception from sing...

评论0

最新资源

A novel approach to predict subjective pain perception from sing...

评论0

最新资源

相关推荐

Algorithm-Predict-external-corrosion-on-oil-and-gas-pipelines.zip

PyPI 官网下载 | adme_predict-0.9-py2-none-any.whl

Python库 | exo_predict-1.0.21-py3-none-any.whl

PyPI 官网下载 | adme_predict-0.8-py2-none-any.whl

PyPI 官网下载 | exo_predict-1.0.21-py3-none-any.whl

IntentNet- Learning to Predict Intention from Raw Sensor Data.pdf

ensemble approach to predict churn

annpb.rar_ANN predict_bp predict_bp-ann_jstann_rainfall

A weighted risk score to predict post-PTMC mitral regurgitation

Efficient Human Pose Estimation from Single Depth Images

3-D Depth Reconstruction from a Single Still Image

IntentNet- Learning to Predict Intention from Raw Sensor Data

机器学习kaggle竞赛Predict Student Performance from Game Play代码+数据集

LayoutNet: Reconstructing the 3D Room Layout from a Single RGB Image

NIMA Neural Image Assessment (TIP 2018).pdf

RefineDet论文

predict-loan-defaulters+

车窗升降控制器的防夹测控算法

基于树莓派&ZigBee网关的远程智能家居控制

基于三维栅格地图的移动机器人路径规划

基于LSTM的空气质量指数预测

基于微博数据的用户特征分析及行为预测

基于神经网络的时间序列预测方法

大数据平台智能运维系统

基于BP神经网络短期电力负荷预测研究

基于区块链技术的商品溯源系统的设计与实现