Analysis And Recognition Of Abstinent Heroin Addicts Based On The Features Of Event-related Potential And Behavior

Heroin is one of the most addictive opioid drugs.After abstinence,heroin addicts have withdrawal symptoms such as nausea,tremors,chills or trouble sleeping.They crave heroin in psychology and behavior.Chemical detection methods can accurately identify people who have taken heroin recently,but it is not effective in the diagnosis of middle and long term abstinent heroin addicts(AHAs).At present,researches on AHAs have found that reward processing and attention bias are abnormal in their brains,and they show different behavior patterns in risky decision-making tasks.Therefore,we use a gambling task with high and low risk options.Based on the analysis of eventrelated potential(ERP)and behavior differences between the two groups,we combine with different feature selection algorithms and optimization algorithms to build a classification model.Finally,we realize the classification of AHAs and healthy controls(HCs).The main conclusions are as follows:(1)In the task state,we find that AHAs have more statistically significant data on Fz electrode.It indicated that the reward related circuits in their forehead were damaged and difficult to recover.In the resting state,there is no significant difference in the time domain or time and frequency domain between AHAs and HCs.Through the further analysis of the Fz electrode,the time domain results of ERP show that the amplitude of N100 and P300 decrease and the latency of P300 shorten in AHAs.The differences in ERP components show that there are abnormal cognitive and processing processes of money stimulation in AHAs.The decrease of ERP mean and variance indicate that they are less sensitive to monetary stimulation,which is consistent with the characteristics of their behavior.The results of ERP in time and frequency domain show that the power of δ in N200 is significantly different under four kinds of stimuli.The power of δ in P300 is significantly different under the stimulation of gain or loss a lot.When the two groups are affected by positive stimulation,the power of δ and θ in N100 are significantly different.In behavior,addicts are are not sensitive to money stimulation,they are in a more impulsive behavior pattern.Specifically,they prefer high-risk options and have shorter decision-making time.(2)In order to achieve the goal of fast and accurate classification of AHAs and HCs,this paper uses behavioral data and ERP data from Fz electrode to structural features.We use Relief F and SVM-RFE to sort the features,then particle swarm optimization algorithm(PSO)and genetic algorithm(GA)are used to find the optimal classification model of support vector machine(SVM).The results show that the optimal classification performance can be obtained by using Relief F and PSO,and the classification accuracy can reach 85.22%.For different types of stimuli,the best accuracy can reach 88.63% by using the data when they get 99 points.Through the above research,this paper proves that there are still differences in ERP and behavior between AHAs and HCs under task state.And it can be used to identify AHAs by using appropriate feature selection algorithm and optimization algorithm.These results can help us to study the abnormal reward circuit,analyze the changes of cognitive processing and attentional bias.This study makes up for the shortcomings of chemical detection,and may become an auxiliary recognition method for AHAs in the future.