Research On Brain-heart Non-linear Interaction And The Relevant Algorithms For Depression Disorder

Of all the organs in the human body,the heart is one of the most extensive connection with the nerves of the brain.More than 150 years ago,Claude Bernard has recognized the complex interactions between brain and heart.In specific physiological states or diseases,brain-heart interactions play a crucial role in determining the coupling between the Central Nervous System(CNS)and the Autonomic Nervous System(ANS).Depression with the predominant manifestation of affective and cognitive disorder,as the high incidences of mental disorder in global,is critical to its inchoate detection and intervention in the diagnosis and treatment by considering the underlying brain-heart interactions.This thesis focused on the research of brain-heart interactions based on bio-electricity signals,orients the pivotal problem in signal interaction analysis and processing,investigates the corresponding solution and neural measurement instruments,depends on the major project of Science and Technology Innovation 2030 “Brain Science and Brain-like Research”,has validated the effectiveness of the proposed algorithms and measurement instruments in the analysis of brain-heart interactions in population with mental depression.Bioelectrical signals are the unstable and weak signals emanated by organisms,which are characterized by weak signals and susceptible to interference.With the development of non-invasive wearable biosensor,EEG and ECG collected by the autonomous operation of users in a pervasive environment are more susceptible to noise interference.Therefore,the improvement of the signal-to-noise ratio between EEG and ECG signals is required to be focused on as a key prerequisite for brain-heart interactions analysis.At the same time,the quantitative representation of the interaction between heart and brain has also pose a challenge for understanding brain-heart interactions.For the characteristics of multichannel EEG and unscrambled ECG signal acquisition,the frameworks of noise removal algorithms and processing are proposed to address in this thesis respectively,to provide a signal quality assurance for brainheart interactions analysis.Based on this,different strategies and methods are employed to perform quantitative measurements of signal interactions.The innovative research results of this thesis include the following four main aspects:(1)A novel thinking of improving moment matching to remove electrooculographic artifacts is proposed for the removal of electrooculographic artifacts in multichannel EEG signals.With the demand for high-throughput and highprecision recording of EEG in brain science research,it is of practical importance to quickly remove EEG artifacts with large amplitude and frequency bands that overlap with EEG.An improvement algorithm considering the similarities and differences between EEG artefacts in multichannel EEG and strip noise in hyperspectral images is proposed in this thesis.Without separate EEG reference acquisition,moment matching filtered reference data are generated based on a regression algorithm.Meanwhile,for the bias differences among different channels caused by conduction effects in multichannel EEG acquisition,the compensation term in the original moment matching is improved,and the banding effect is eliminated by adding empirical coefficients to finally obtain the moment matching formula for multichannel EEG ocular artifacts removal.By comparing with the existing methods on different datasets,the results show that the proposed method in this thesis is capable of obtaining the highest signal-tonoise ratio of EEG signals and has obvious advantages of real-time performance in the analysis of numerous data volume.Experimental analysis shows that high quality EEG signals for brain-heart interactions analysis is provided by this method.(2)To address the signal quality and detection problems in non-perturbative ECG recording,a framework for the overall processing of ballistocardiogram(BCG)signals based on non-perturbative ECG acquisition seats is presented in this thesis.The analysis of heart rate variability is able to be performed using non-contact BCG signals.A holistic processing framework,including signal quality detection,J-peak detection of BCG signal and beat-by-beat estimation of cardiac period in the noise segment of the signal,is proposed to improve the quality of BCG signal.Firstly,the signal quality detection based on constant false alarm rate(CFAR)is proposed to determine the analyzability of the signal,retain the analyzable signal and discard the insufficient quality signal.Secondly,the threshold of CFAR and the maximum difference between local adjacent peaks and valleys are combined to detect the J-peak detection of the available BCG signals.Finally,an improved LSTM(Long Short-Term Memory)model is proposed to perform the beat-by-beat estimation of the cardiac cycle for the local noise segment in the retained signal,and high-quality heart rate variability data are obtained in the end.The validation results show that more than 95% noise marks are achieved by the proposed method,the sensitivity and accuracy of J-peak detection reach98.9% and 95.7%,respectively.The cardiac cycle estimation is satisfied with the signal analysis requirements and the computational and storage load are greatly reduced by the overall processing procedure.The framework has able to be extended to ECG signals to improve the usability of ECG signal measurements in unsupervised environments.(3)To address the issue of cortical processing representation of heartbeat signals from different emotional stimuli in depressive disorders,Heartbeat evoked potentials(HEP)as a neural measure for the cortical processing of cardiac afferent signals in brain-heart interactions is utilized to investigate whether audio stimulation with different emotional polarity leads to the changes of HEP and brain-derived activation in depressed patients.It is found that depression and control in different time windows of HEP present the significant differences in response to neutral and negative affective stimuli,and that scalp differences were reflected in the left hemisphere during neutral stimulation and in the right hemisphere during negative stimulation.Source localization analysis shows that brain regions involved in emotional processing,such as the insula,superior frontal gyrus and parahippocampal gyrus,as well as limbic lobe,which is related to brain regions involved in the regulation and control of visceral activity,are significantly enhanced in the depressed group during negative and positive emotional stimuli,suggesting that HEP is able to be utilized as a reliable neurological measure to study brain-heart interactions during different emotional processes in depressed patients and thus help to understand clinical symptoms during depression.(4)To address the lack of research on the directionality of brain-heart interactions in patients with depressive disorders,this paper uses Heart Rate Variability(HRV)as an indicator of ANS changes and convergent cross mapping to investigate the nonlinear directional interactions between HRV and specific components of EEG in depressed populations in the context of different emotional audio stimuli.The experimental results show that the effect of HRV on delta and alpha waves are significantly higher in depressed patients than in healthy controls in the condition of no audio stimulation at rest.The intensity of alpha and beta waves on HRV is significantly lower in the depressed group than in healthy controls under neutral affective stimulation.For the research individuals,the changes in the interaction of theta,alpha and beta waves with ECG in the depressed group are mainly reflected in the negative affective stimulation condition.The study demonstrates that brain-heart directional interaction measure based on physiological signal time series is conducive to understanding the complex relationship between depression and brain-heart system interaction behaviors and provide a support for considering potential brain-heart interactions in the diagnosis and treatment of depression.To sum up,brain-heart interactions in depressed patients during different emotional processes are investigated and discussed in this thesis from the research of signal preprocessing algorithms before brain-heart interactions analysis to the cortical processing of cardiac signals and brain-heart nonlinear directional interactions.The proposed signal preprocessing technique is conducive to performing a preferably analysis between autonomous cardiac control and brain dynamics.The investigation of brain-heart interactions behavior by different measurements has implications for understanding the information transfer between CNS and ANS in depressed patients,which has ability to provide a novel supplementary support for the diagnosis and personalized medical treatment of depression in a pervasive environment.