Research Of Neural Oscillatory Information Interaction Of Lies And A Polygraph Detection System

Studies of the functional and structural networks of the brain have suggested that deception involves a number of complex cognitive processes such as executive control,mental decision-making and working memory.Deception involves the cooperation of numerous brain networks,but few researches has been done on the direction of these network connections.To achieve this goal,our research used the event related potential(ERP)signals and phase transfer entropy(PTE)to construct weighted directed functional brain network(WDFBN)in the deception state.The aim is to study the effective connections between brain regions under neural oscillations and to explore the patterns of information interaction in the human brain during deception.In this study:First,40 participants were randomly assigned to either the honest or deception group using the standard guilt knowledge test experimental model and instructed to tell the truth or lie in response to certain stimuli.At the same time,all subjects’ EEG signals were recorded simultaneously using a 64-lead EEG cap and the data were pre-processed to obtain ERP signals.Second,the WDFBN was constructed based on the subject’s ERP signal,with the network nodes being 62 electrode channels and the connected edge weights being the normalized phase transfer entropy(d PTE)between each pair of electrodes.PTE is a method for measuring the effective connectivity of a network based on phase relations and is robust to noise and linear mixing.As it would be more accurate to perform instantaneous phase calculations on narrowband signals,the d PTE effect connection characteristic matrix of the signal was calculated in four bands delta(0 to 3 Hz),theta(4to 7 Hz),alpha(8 to 13 Hz)and beta(14 to 30 Hz).Furthermore,statistical tests were used to assess the differences in the feature matrix between the two states.Differences in entropy values between groups were analysed for the channel combination features in the matrix.The entropy values on the electrode pairs with significant differences were subjected to minimum-redundancy-maximum-relevance before being used as classification features for the machine learning Catboost classifier.Statistical analysis of the classification results showed that the all bands were able to distinguish deception from truth-telling ERPs with predictive accuracy of 77.43%,81.43%,75.96% and 82.14%,respectively,based on 4 features,and thus d PTE can be considered a powerful feature for lie detection.Finally,in order to decode the information interaction patterns of neural oscillations in the lying state.Features of lead pairs with large differences in significance in each of the four bands were obtained and WDFBNs were constructed based on the classification results.Analysis of functionally activated brain regions based on effector connectivity networks showed that deceptive responses in the delta bands elicited stronger information interactions in the frontoparietal network than real responses,and that the temporal lobe in theta band and the occipital lobe in alpha were activated during the execution of the deception task.In addition,in the beta band the activity of the frontoparietal network reduced.At the same time,the delta band was found to flow mainly from the superior parietal lobule to the prefrontal and temporal lobes;the theta band flowed mainly from the superior temporal gyrus to the angular gyrus;the alpha band flowed mainly from the dorsolateral prefrontal cortex to the occipital lobe;and the beta band flowed from the the precuneus to premotor cortex.In addition,the prefrontal cortex was activated in all frequency bands and the information interaction between the prefrontal cortex and other brain regions better reflected the difference between lying and honesty,implying that the prefrontal lobe plays a key role in deception processing.These results suggested that deception causes activation of multiple brain regions,that connections between these brain regions were enhanced,and that information exchange became more frequent during this process.The results further confirmed that cognitive processes during deception require the joint involvement of the executive control network,the working memory network and the psychosocial network.The results of this study showed the pattern of information interaction in the brain during deception,providing new insights into the neural activity mechanisms of the brain during executive deception and providing a new basis for the development of brain network-based lie detection.