Study On Mechanism And Treatment Outcome Prediction Of The Transcutaneous Auricular Points Vagus Nerve Stimulation In Primary Insomnia Based On Brain Functional Features

BackgroundPrimary insomnia(PI),the second-largest mental disorder,has become a risk factor for many diseases and seriously affected the phsysical and mental health of patients.Although drug therapy and cognitive-behavioral therapy(CBT)were recommended as the first-line treatment options for PI in the clinical guidelines,about 40%of patients fail to obtain continuous remission through these treatments due to many adverse reactions of drug therapy,long treatment cycle of CBT,high medical costs,and lack of therapists.In recent years,simple and effective traditional Chinese medicine suitable technologies and Western medicine physical therapies have gradually risen,such as auricular acupuncture and transcutaneous vagus nerve regulation technology.Transcutaneous auricular vagus nerve stimulation(taVNS)is effective to treat insomnia,however,individual efficacy varies greatly and the neural mechanism is still unclear,which limits its application.Therefore,exploring the mechanism and objective markers of individual efficacy differences in PI patients is of great significance to improve clinical efficacy.However,there is no mechanism study on the efficacy difference of taVNS in PI.In recent years,resting-state functional magnetic resonance imaging(rs-fMRI)has become a powerful tool to explore the neural mechanism of neuropsychiatric disorder.The studies of the classification and prediction of mental diseases by using brain functional features combined with machine learning methods have gradually become popular,which provides a new method for studying the neural basis of mental diseases.Therefore,in this study,the brain functional indexes of PI patients before treatment were extracted as features,and the machine learning methods were used to construct the prediction and classification model of the curative effects.Then,the features with the most distinguishing ability to the curative effect were screened out through the weight,F-score feature selection methods,and 10-fold cross validation method,so as to explore the central mechanism and objective markers of the curative effects differences of taVNS in the treatment of insomnia.Part Ⅰ Central mechanism of primary insomnia based on regional homogeneity ObjectiveThe differences in autonomic nervous function and brain function were observed between primary insomnia(PI)patients and healthy controls(HC),and the central mechanism of PI was analyzed based on regional homogeneity(ReHo),to lay the foundation for the subsequent efficacy prediction research of taVNS.MethodsA total of 100 patients with PI and 100 healthy people were recruited.The baseline demographic data,clinical symptom scales,heart rate variability(HRV)and rs-fMRI data were collected,and the HRV data were collected also during 5 minutes of continuous taVNS,to compare the differences in sleep,emotion,quality of life,autonomic nervous function,brain function and response to taVNS stimulation between PI patients and healthy people,The clinical symptom scales included Pittsburgh Sleep Quality Index(PSQI),Insomnia Severity Index(ISI),Self-rating Anxiety Scale(SAS),Self-rating Depression Scale(SDS),Composite Autonomous Symptom Score-31(COMPASS-31),and 36-item Short from Health Survey(SF-36),as an evaluation index of sleep,emotion,quality of life,and autonomic nerve function.HRV indexes include the total heart rate,inter-beat interval(NN or RR),the mean square successive difference(MSSD),the percentage of successive normal sinus RR intervals more than 50ms(PNN50),high frequency(HF),low frequency(LF),and peak rate(LF/HF),as evaluation indexes for the functional activities of the autonomous nervous system(ANS).The rs-fMRI index used ReHo to measure the degree of local brain signal synergy to evaluate the brain function.ResultsThere was no significant difference in demographic data(including age,gender,BMI,blood pressure,and education degree),living habits(such as smoking,drinking,and coffee),and chronic diseases history(hypertension,diabetes,and hyperlipidemia)between the patients and the healthy control group(P>0.05),indicating comparability.In terms of clinical symptoms,compared with the healthy control group,the PSQI(Z=-10.421,P=0.000)、ISI(Z=-10.383,P=0.000)、SAS(Z=-8.400,P=0.000),and SDS(Z=-8.838,P=0.000)scores were significantly increased in the patient group,and the SF36(Z=-8.613,P=0.000)was significantly decreased.The COMPASS-31 score was slightly higher,but no difference between groups(Z=-1.864,P=0.062).HRV analysis results1.Comparison results between groups:there was no significant difference in the baseline total heart rate(t=0.073,P=0.942),RR(Z=-0.068,P=0.946),MSSD(Z=0.702,P=0.483),PNN50(Z=-0.835,P=0.404),LF(Z=-0.064,P=0.949),HF(Z=0.753,P=0.452),and LF/HF(Z=-0.872,P=0.383)between the two groups.During the taVNS,compared with the control group,the difference of the total heart rate(Z=-2.857,P=0.004)in the patient group was significantly decreased,and the difference of RR(Z=3.898,P=0.000)and PNN50(Z=-1.977,P=0.048)were significantly increased.There was no significant difference in MSSD difference(Z=-0.551,P=0.582),LF difference(Z=-0.562,P=0.574),HF difference(Z=-0.520,P=0.603)and LF/HF difference(Z=0.530,P=0.596)between the two groups.2.Intra-group comparison results:compared with baseline,the total heart rate(t=5.198,P=0.000)was significantly decreased during the taVNS stimulation,while RR(Z=-4.977,P=0.000),MSSD(Z=-2.242,P=0.025)and PNN50(Z=-4.062,P=0.000)were significantly increased,and there was no significant difference in LF(Z=-0.914,P=0.361),HF(Z=-1.713,P=0.087)and LF/HF(Z=-0.533,P=0.594);in the control group,the total heart rate decreased significantly(t=2.722,P=0.008),MSSD increased significantly(Z=-2.704,P=0.007),RR(Z=-1.270,P=0.204),PNN50(Z=-1.805,P=0.071),LF(Z=-0.353,P=0.724),HF(Z=-1.522,P=0.128),and LF/HF(Z=-1.618,P=0.106)showed no significant difference.The results of brain function analysis showed that,compared with healthy controls,the ReHo values were significantly increased in the bilateral precuneus,right angular gyrus,bilateral medial orbitofrontal gyrus,and left superior orbitofrontal gyrus,bilateral calcarine,bilateral lingual gyrus,left cerebellum,right middle temporal gyrus,right superior temporal gyrus,and right middle occipital gyrus.Part II Predicting therapeutic effects of taVNS based on the fractional amplitude of low-frequency fluctuations in PIObjectiveTo observe the individual efficacy differences of the PI patients after taVNS treatment,predict the PSQI score difference based on the fractional amplitude of low-frequency fluctuations(fALFF)and support vector regression(SVR)methods,explore the central mechanism of individual efficacy difference,and find the objective markers of efficacy in PI patientsMethodsThe patients with PI were treated with taVNS for 4 weeks.The clinical symptom scales,HRV indexes and rs-fMRI data were collected before and after treatment.Firstly,the PSQI score differences were calculated as the efficacy evaluation index.The fALFF of the whole brain voxels before treatment were calculated by Matlab and DPABI 3.1 software as the features.The prediction model of PSQI score difference was constructed based on SVR algorithm.The correlation analysis between the predicted values and the real values was carried out to obtain the correlation values r and P.Then,the weight method of feature selection and 10-fold cross-validation method were used to select the features with the best prediction ability.The permutation test was used to evaluate the contingency of r-value to Pvalue.Finally,the voxels with the top 30%of the predicted weight values were taken and visualized via MRICron software.ResultsThe results of clinical scales analysis showed that,PSQI(Z=-6.780,P=0.000),ISI(Z=-6.606,P=0.000),SAS(t=4.534,P=0.000),SDS(t=3.787,P=0.000)and COMPASS-31(Z=-2.380,P=0.017)were significantly decreased and SF-36 was significantly increased(t=-7.108,P=0.000)after treatment.The HRV analysis resultsResults of the comparisons before and during continuous taVNS:Before treatment,compared with the baseline,the total heart rate of the patients was significantly decreased(t=5.198,P=0.000)during taVNS stimulation,while RR(t=-5.608,P=0.000),MSSD(Z=-2.242,P=0.025),and PNN50(Z=-4.062,P=0.000)were significantly increased,no significant difference in LF(Z=-0.914,P=0.361),HF(Z=-1.713,P=0.087),and LF/HF(Z=-0.533,P=0.594);after treatment,compared with baseline,the total heart rate was significantly reduced(t=-4.215,P=0.000)during the taVNS,and the RR was significantly increased(t=-4.250,P=0.000),no significant difference in MSSD(Z=-0.302,P=0.763),PNN50(Z=-0.862,P=0.388),LF(Z=-1.010,P=0.313),HF(Z=-1.421,P=0.155),and LF/HF(Z=-0.074,P=0.941).Results of the comparisons before and after treatment:Compared with before treatment,the baseline total heart rate of the patients was significantly decreased(t=2.507,P=0.015)after treatment,no significant difference in baseline RR(t=-1.723,P=0.090),MSSD(Z=-0.751,P=0.453),PNN50(Z=-1.487,P=0.137),LF(Z=-1.437,P 0.151),HF(Z=0.179,P=0.858),and LF/HF(Z=-0.515,P=0.606).The total heart rate difference was significantly decreased(Z=-3.043,P=0.002),while the RR difference(Z=-2.640,P=0.008),MSSD difference(Z=-2.447,P=0.014),and PNN50 difference(Z=-2.545,P=0.011)were significantly increased,no significant difference in LF difference(Z=-0.236,P=0.813),HF difference(Z=-0.631,P=0.528),and LF/HF difference(Z=-0.717,P=0.473).The prediction results showed that using the baseline fALFF could effectively predict the PSQI differences of the PI patients after taVNS treatment(r=0.3166,P=0.0067).The top 30%of brain regions in weight ranking include the bilateral orbitofrontal cortex,medial frontal cortex,and middle frontal gyrus(DMN),left middle and inferior orbitofrontal gyrus,right inferior occipital gyrus,bilateral caudate,left amygdala,left putamen,left pallidum,left insula,and bilateral parahippocampal gyrus(AN),bilateral lingual gyrus,bilateral calcarine,right cuneus,and left fusiform(VIN),bilateral cerebellum(CEN).left superior temporal gyrus,bilateral middle temporal gyrus,left temporal gyrus,and left Heschl gyrus.Part III Predicting the classification of PI after taVNS treatment based on functional connectivityObjectiveTo construct a predictive model for the classification of curative effect in patients with PI after taVNS treatment by using the resting-state functional connectivity(rsFC)matrix as features and logistic regression(LR)method,exploring the central mechanism of the individual differences in curative effect,and finding reliable objective markers,to screen out the target patients in advance.MethodsAccording to the efficacy score of PI patients after taVNS treatment,the outcomes of patients were divided into the effective group and the ineffective group.Calculating all the baseline rsFC of the regions of interest(ROI)in DPABI 3.1(Data Processing&Analysis for Brain Imaging,V3.1)software,and the features with distinguishing ability were screened out as the feature through the F-score and 10-fold cross-validation method.Based on Matlab(Matrix Laboratory)software,the LR algorithm was used to construct the prediction model of efficacy classification of PI patients after taVNS treatment,and the contingency degree of accuracy was evaluated by permutation test.The performance of the model was evaluated by using the area under the receiver operator characteristic curve(AUROC),accuracy,sensitivity,and specificity.Finally,the edges and important brain regions with classification significance were visualized via Circoss and Brain Net Viewer software.ResultsSince there were only 20 patients in the ineffective group,the top 20 patients in the effective group were included in the analysis.There was no significant difference in age,gender,weight,height,blood pressure,education level,history of hypertension,diabetes,hyperlipidemia,and chronic diseases between the effective group and the ineffective group(P>0.05).Comparison results of the clinical scales between groups:Before treatment,there was no significant difference in these clinical scale scores between the effective group and the ineffective group(P>0.05),and the two groups were comparable.After treatment,the PSQI difference(Z=-5.440,P=0.000),ISI difference(t=5.483,P=0.000),SAS difference(Z=-3.333,P=0.001),and SDS difference(t=3.103,P=0.004)were significantly decreased in the two groups,no difference between COMPASS-31 difference(t=1.834,P=0.074)and SF-36 difference(t=1.306,P=0.199).Intra-group comparison results of the clinical scales:Compared with baseline,PSQI(t=15.884,P=0.000),ISI(t=10.379,P=0.000),SAS(t=7.258,P=0.000),SDS(t=4.860,P=0.000),and COMPASS-31(t=3.497,P=0.002)were significantly decreased and SF36 was significantly increased(t=-6.308,P=0.000)in the effective group after treatment;in the in effective group,ISI decreased significantly(t=-2.806,P=0.005),SF-36 increased significantly(t=-6.876,P=0.000),PSQI(t=1.891,P=0.074),SAS(t=0.746,P=0.465),SDS(t=0.626,P=0.539),and COMPASS-31(t=0.712,P=0.485)no significant difference.The prediction results showed that the classification effect was the best when the top 300 F scores of the FCs were used as features,the prediction model had the highest accuracy[average accuracy 80%(5000 permutation tests,P<0.0002),sensitivity 80%,specificity 80%,AUC 0.7875].There are 64 edges of the consensus functional connections mainly distributed in the default mode network(DMN),the affective network(AN),the visual network(VIN),and the cerebellar network(CEN).There were 12 main regions with distinguishing ability,including the bilateral anterior cingulate and paracingulate gyrus,right posterior cingulate gyrus,right amygdala,right middle and inferior orbitofrontal gyrus,right calcarine,right angular gyrus,and cerebellum.ConclusionPI patients have impaired autonomic nervous function and abnormal brain function.The brain areas with abnormal function are mainly distributed in the DMN,AN,VIN,and CEN.Functional activities of these brain networks are potential markers for insomnia-related symptoms.The taVNS can improve the sleep,emotion,and living quality of patients with PI,via regulating their autonomic nerve balance and brain function.Using the baseline brain functional features of PI patients can predict the efficacy of taVNS.The functional integration within and between the networks of DMN,AN,VIN,and CEN may be the central mechanism of the difference in the efficacy of taVNS in the treatment of PI,which can be used as an objective marker of efficacy.