Mechanism And Efficacy Prediction Of Brain Effect Of TaVNS For Major Depressive Disorder Based On FMRI And Machine Learning Technology

BackgroundMajor depressive disorder(MDD)is a common clinical psychiatric disorder with high incidence,recurrence and disability rates,and in severe cases,suicidal tendencies,which brings a heavy burden to patients,families and society.Currently,the treatment of MDD is mainly based on pharmacotherapy,cognitive-behavioral therapy and physical therapy.Although pharmacotherapy is effective,it has the shortcomings of adverse drug reactions and dependence.Cognitive-behavioral therapy is time-consuming,costly,and relatively short of therapists,which still does not meet clinical needs.Therefore,a safe,effective and convenient non-pharmacological therapy is urgently needed.Our team’s previous research has shown that transcutaneous auricular vagus nerve stimulation(taVNS)is clinically effective in treating MDD,and has the advantages of being safe,convenient,and economical.However,there are shortcomings such as fewer control groups and a single study on the efficacy mechanism.In addition,although taVNS is effective in the treatment of MDD,the individual efficacy varies greatly,which further limits its clinical application and promotion.Therefore,it is important to strengthen the in-depth study on the efficacy mechanism of taVNS for MDD,explore the individual differences in its efficacy,and explore its objective neuroimaging markers to improve the clinical efficacy.However,there are no predictive studies on the efficacy of taVNS in the treatment of MDD.In recent years,resting-state functional magnetic resonance imaging(rs-fMRI)has been gradually applied to the research of neurological and psychiatric diseases.In particular,the integration of machine learning(ML)methods for disease classification and efficacy prediction is a hot topic of current research.Several studies have demonstrated the feasibility and effectiveness of fusing neuroimaging features with ML methods for individualized prediction of treatment effects,which also provides a new approach for conducting efficacy prediction studies for the treatment of MDD.Therefore,in this study,we observed the clinical efficacy and safety of taVNS for MDD,further observed its immediate and long-term brain effect mechanisms and their relationships,and used ML methods to construct an efficacy prediction model to predict the response level of patients with MDD treated with taVNS,and screened key baseline predictive features to evaluate the appropriate population for taVNS treatment of MDD and help acupuncture precision.This will provide clinical reference value for evaluating the suitable population for taVNS treatment of MDD and helping acupuncture to treat accurately.Objective(1)The clinical efficacy and safety of taVNS for MDD were observed in a randomized controlled study to provide a reliable basis for the clinical application and further promotion of taVNS for MDD.(2)Based on the rs-fMRI technique,a randomized controlled study was conducted to observe the mechanism of immediate and long-term brain effects in MDD patients treated with taVNS,to explore the relationship between immediate and long-term brain effects,and to further observe the correlation between changes in behavioral data and differential brain areas in MDD patients at baseline and before and after treatment.(3)The functional features of amplitude of low frequency fluctuation(ALFF)and regional homogeneity(ReHo)based on rs-fMRI were used to discover the mechanisms of abnormal brain function in MDD patients,and the ML method was fused to observe the baseline functional.We also investigate whether the baseline functional characteristics can effectively predict individual differences in patients with MDD treated with taVNS,and explore objective imaging markers associated with efficacy.MethodsPart Ⅰ:Clinical efficacy and safety study of transcutaneous auricular vagus nerve stimulation for MDDA total of 84 patients with MDD were included in this study,and were divided into 42 cases each in the treatment and control groups using the randomized control principle.The treatment group was treated with taVNS,and the main stimulation area was the auricular cavity.The control group mainly stimulated the superior auricular navicular region at the outer ear margin.Both groups were treated for 30 minutes each time,twice a day,7 days a week,for a total of 8 weeks.Before and after treatment,17-items of Hamilton Rating Scale for Depression Scale(HAMD-17),14-items of Hamilton Rating Scale for anxiety scale(HAMA-14),Self-rating Depression Scale(SDS)and Self-rating Anxiety Scale(SAS)were evaluated.Patients were evaluated for efficacy based on their HAMD-17 score reduction rate before and after treatment,and observed for adverse effects during treatment.Part II:Study on the mechanism of fMRI brain effect of transcutaneous auricular vagus nerve stimulation for MDDThe patients in this study were 84 patients from the first part of the study,and the grouping method and treatment protocol were the same as in the Part Ⅰ.The immediate effect was treated for 30 minutes and the long-term effect was treated for 8 weeks.Rs-fMRI scans were performed before and after the immediate treatment and after 8 weeks in both groups.Differences in local brain functional activity of ALFF and ReHo were observed before and after immediate and long-term treatment in both groups,and correlations between changes in behavioral data and differential brain areas at baseline and before and after treatment in MDD patients were observed using biased correlation analysis.Part III:Predictive study of the efficacy of transcutaneous auricular vagus nerve stimulation in the treatment of MDDA total of 87 patients with MDD and 96 healthy controls(HCs)were included in this study.MDD patients received 8 weeks of taVNS treatment and the HC group was not treated.Both groups underwent rs-fMRI scans before treatment.The ALFF and ReHo characteristics of the subjects in both groups were extracted,and to exclude the influence of factors other than disease on the characteristics of brain regions,three factors,namely sex,age,and education,were also included as covariates.The study first obtained the abnormal brain regions of MDD patients by a two-layer nesting method.In the inner nesting,t-test+LASSO-logistic regression was used to screen out the significant features;in the outer nesting,the most critical features were selected based on the retention frequency of the features in all models,and the best retention frequency was determined by the classification performance of the logistic regression model constructed with the corresponding feature set.Finally,a taVNS efficacy prediction model was constructed using SVM based on the functional characteristics of abnormal brain regions.And an additional 1000 repetitions of ten-fold cross-validation was used to calculate the mean and variance of the 1000-times efficacy prediction model to assess the stability of key features in predicting efficacy.ResultsPart I:Clinical efficacy and safety study of transcutaneous auricular vagus nerve stimulation for MDDThere was no statistical difference between the two groups in terms of sex,age,education and disease duration(P>0.05),and there was no statistical difference between the two groups in terms of baseline HAMD-17,HAMA-14,SDS and SAS scores(P>0.05).There were statistical differences in HAMD-17,HAMA-14,SDS,and SAS scores between the two groups after treatment compared with those before treatment(P<0.05).There was a statistical difference in HAMD-17,HAMA-14,SDS,and SAS scores between the two groups after treatment(P<0.05).The total effective rate was 80.00%in the treatment group and 62.16%in the control group,which were statistically diffeirent(P<0.05).And there was no significant adverse reaction in the MDD group during the treatment period.Part Ⅱ:Study on the mechanism of fMRI brain effect of transcutaneous auricular vagus nerve stimulation for MDDChapter 1 Immediate Effects Research:Compared with the immediate pre-treatment period,ALFF was increased in the left/right middle frontal gyrus and posterior cerebellar lobe and decreased in the right inferior temporal gyrus in the treatment group after immediate treatment,and increased in the right middle frontal gyrus and posterior cerebellar lobe and decreased in the right inferior temporal gyrus in the control group after immediate treatment.In comparison between the two groups after treatment,ALFF in the left middle frontal gyrus and posterior cerebellar lobe was significantly increased and ALFF in the right inferior temporal gyrus was significantly decreased in the treatment group compared with the control group.ReHo was increased in the left middle cingulate gyrus and decreased in the right inferior orbital frontal gyrus and right precuneus in both groups after immediate treatment compared to before immediate treatment.Compared between the two groups after treatment,ReHo in the right precuneus was significantly decreased in the treatment group compared to the control group.Bias correlation analysis showed that the ALFF values of the right inferior temporal gyrus in the MDD group were positively correlated with the HAMD-17 scale score(r=0.257,P=0.016),while the ALFF values of the left middle frontal gyrus,the ReHo values of the left middle cingulate gyrus and the HAMA-14 scale scores were negatively correlated(r=-0.237,P=0.038;r=-0.225,P=0.049).Chapter 2 Long-term effects studies:Compared with the pre-treatment period,ALFF in the right middle frontal gyrus,left superior frontal gyrus and right lingual gyrus were increased and decreased in the left precuneus and right angular gyrus in the treatment group,while ALFF in the left superior frontal gyrus and right lingual gyrus were increased and decreased in the left precuneus and right angular gyrus in the control group after treatment.Compared between the two groups after treatment,ALFF in the treatment group was significantly increased in the right middle frontal gyrus and the right lingual gyrus,and significantly decreased in the left precuneus and the right angular gyrus compared to the control group.Compared with the pre-treatment group,ReHo was increased in the left/right insula,left hippocampus,left/right middle occipital gyrus and posterior cerebellar lobe and decreased in the right orbital inferior frontal gyrus in the treatment group,while ReHo was increased in the left/right insula,left hippocampus,right middle occipital gyrus and posterior cerebellar lobe and decreased in the right orbital inferior frontal gyrus in the control group after treatment.When compared between the two groups after treatment,ReHo was significantly increased in the right insula,left hippocampus,and left middle occipital gyrus,and significantly decreased in the right inferior orbital frontal gyrus in the treatment group compared to the control group.Bias correlation analysis showed that in the treatment group,the change value of ALFF in the left precuneus was positively correlated with the improvement value of SDS score(r=0.421,P=0.010),the change value of ReHo in the right inferior orbital frontal gyrus was positively correlated with the improvement value of HAMD-17 score(r=0.378,P=0.021),and the change value of ReHo in the right middle occipital gyrus was negative correlation(r=-0.404,P=0.013).In the control group,the change in ALFF in the left precuneus was positively correlated with the improvement in HAMD-17 score(r=0.388,P=0.023),and the change in ReHo in the left hippocampus was negatively correlated with the improvement in HAMD-14 score(r=-0.398,P=0.020).Part III:Predictive study of the efficacy of transcutaneous auricular vagus nerve stimulation in the treatment of MDDThere were no statistical differences between the MDD and HC groups in terms of sex,age and education(P>0.05).There was a statistical difference in the MDD group compared to the pre-treatment group in terms of HAMD-17 scores after treatment(P<0.05).The study found the best classification performance of the logistic model constructed with the corresponding feature set(AUC=0.853)by double-layer nesting when the retention frequency was taken as 80%,and the corresponding key feature set included 19 functional features.The study further constructed support vector regression models for HAMD-17 score differences based on key features with better predictive performance(r=0.890,P<0.001)and support vector classification models predicted valid or invalid predictive performance with AUC=0.969.Conclusion(1)TaVNS is clinically effective in the treatment of MDD,and its efficacy is better than that of the control group,which can significantly improve patients’ clinical symptoms and provide quality of life.And no significant adverse effects were seen,which is worthy of further clinical promotion.(2)In terms of immediate effects,taVNS had immediate modulatory effects on some brain regions in the cognitive control network(CCN),default mode network(DMN),reward network,limbic system and sensorimotor network(SMN)in MDD patients.In terms of long-term effects,taVNS had long-term modulatory effects on some brain regions in the CCN,DMN,emergent network,reward network,visual processing cortex,limbic system,and SMN in MDD patients.The immediate effect may be the basis of action and potential target for the long-term effect,and the activation of brain regions was more extensive in the treatment group than in the control group.(3)In terms of efficacy correlation,some brain areas in the DMN,reward network and visual processing cortex were correlated with clinical efficacy in MDD group,which may be important brain networks and brain areas targeted for modulation by taVNS for MDD treatment.(4)The integrating ML method using abnormal baseline brain functional features was able to predict individual efficacy of taVNS for MDD with relative accuracy,and 19 quantitative functional features,including the left precuneus,left orbitofrontal middle gyrus and right precentral gyrus,were more reliable in predicting taVNS efficacy and may be the imaging markers for predicting efficacy.(5)The DMN,reward network,and SMN play important roles in the mechanisms of immediate and long-term brain effects and efficacy prediction models of taVNS for MDD,and are the brain networks that screen out a higher concentration of imaging markers.