Exploring The Effects Of Acupuncture On Brain Function In Patients With Stroke Numbness Based On Multivariate Pattern Analysi

Objective:Limb numbness is one of the common sequelae of stroke,which is belongs to the category of sensory disturbance.The course of the disease is often lingering and difficult to restore,seriously affecting the quality of life of the patients.The western medicine treatment mainly including the nerve nutrition,the sense stimulation and the rehabilitation therapy,but the clinical curative effect is unsatisfactory.According to the theory of traditional Chinese medicine,the core pathogenesis of limb numbness after stroke is the damage of spirit,and the deficiency of Qi and blood.The acupuncture method of "Co-regualtion of the body and spirit",which is centered on supplementing Qi,nourishing blood and regulating physical body,as well as regulating spirit,has been put forward,and achieved remarkable effect in the acupuncture treatment of limb numbness after stroke.Resting state functional magnetic resonance imaging（rs-fMRI）provides technical support for the study of the central mechanism of acupuncture treatment for various diseases,and multivariable pattern analysis based on machine learning can extract neuroimaging markers of related diseases.In this study,rs-fMRI was combined with machine learning to classify and predict the patients with limb numbness after stroke at the individual level and identify the relevant brain functional markers.Then the mechanism of acupuncture of "Co-regulation of the body and spirit" was preliminarily explored by comparing the changes of brain functional markers before and after acupuncture treatment.Methods:A total of 25 stroke sequelae patients with unilateral limb numbness were enrolled as experimental group and 24 age-and gender-matched healthy subjects as control group.The patients in the experimental group were treated with the acupuncture of "Coregulation of the body and spirit" 3 times a week for 4 weeks.The acupuncture points including Baihui,Sishencong,Shuigou,Quchi,and Zusanli on the affected side.Moreover,before and after acupuncture,the numbness degree of the limbs,the scores of the whole sensory function and the emotion were evaluated in experimental group.Then,the patients were scanned with rs-fMRI before and after acupuncture while the healthy subjects were scanned only at baseline.After the rs-fMRI data pre-processing,the brain functional activity indexes were calculated:fALFF,DC,Reho,and the differences in brain functional activity between the experimental group and control group were analyzed based on the general linear model.Based on this,the feature extraction and the classification of patients with machine learning were carried out.At last,the differences of brain functional markers before and after acupuncture were compared and the correlation between the brain functional activities and clinical scales were analyzed.Results:（1）The baseline data of the experimental group and control group were balanced,and there was no significant difference in general data such as sex and age,which was comparable（P>0.05）.After acupuncture treatment for 4 weeks,the scores of VAS,SDS and SAS in the experimental group were significantly lower than those in the baseline（P<0.05）,while the scores of NSA were significantly higher（P<0.05）.（2）Based on general linear model analysis of differences in brain functional activity,stroke sequelae patients with unilateral limb numbness compared with healthy controls,the brain regions with enhanced Reho were located in the right caudate nucleus,the left hippocampus,the left inferior temporal gyrus and the right insula（P<0.05）.Reho decreased in the bilateral postcentral gyrus,bilateral precuneus,left angular gyrus,right superior occipital gyrus,left cuneus,right middle frontal gyrus,bilateral paracentral lobules,right lingual gyrus,and left middle occipital gyrus（P<0.05）.And compared with healthy controls,the patients showed enhanced fALFF values in the left orbital part of superior frontal gyrus,left fusiform gyrus,right dorsolateral superior frontal gyrus,right insula,left supplementary motor area gyrus,and right putamen（P<0.05）.The fALFF attenuated regions were located in the right cerebelum₇b3 bilateral angular gyrus,right postcentral gyrus,left precuneus,right cerebelum_crusl,left lingual gyrus,bilateral inferior parietal angular gyrus,left cingulate gyrus,right superior marginal gyrus,left precentral gyrus,right cuneus,and left middle occipital gyrus（P<0.05）.The brain regions with enhanced DC were located in right orbital part of inferior frontal gyrus,the vermis,right angular gyrus,left insular inferior frontal gyrus,right superior frontal gyrus,left orbital part of inferior frontal gyrus and left middle frontal gyrus（P<0.05）.The areas with decreased DC values were located in the left middle occipital gyrus,left precuneus,right superior occipital gyrus,right cerebelum₈,left inferior parietal angular gyrus,right postcentral gyrus,left lingual gyrus,right paracentric lobule and right cerebelum₄₅（P<0.05）.（3）Based on the general linear model analysis,16 brain regions with abnormal Reho values were obtained,and the Reho values were extracted as features.The recursive feature elimination（RFE）method was used for feature selection.When the selected features were 7,the SVM model can achieve the highest classification accuracy.The 7 features and their corresponding weights were right insula（0.841）,left hippocampus（0.838）,right postcentral gyrus（0.519）,left angular gyrus（0.517）,right lingual gyrus（0.439）,right caudate nucleus（0.414）and right middle frontal gyrus（0.404）.Meanwhile,20 brain regions with abnormal fALFF values were selected and their fALFF values were extracted as features.When the selected features were 8,the SVM model can achieve the highest classification accuracy.Including right cerebelum_crusl（1.146）,the right inferior parietal angular gyrus（0.823）,the left supplementary motor area gyrus（0.796）,the right angular gyrus（0.687）,the left angular gyrus（0.613）,the right supramarginal gyrus（0.509）,the right postcentral gyrus（0.487）and the left middle occipital gyrus（0.378）.Based on the general linear model analysis,16 brain regions with abnormal DC values were obtained,When the DC values were extracted as features,the SVM model can achieve the highest classification accuracy with 9 features.Of which,the right superior frontal gyrus（0.504）,the vermis（0.499）,the left precuneus（0.468）,the right postcentral gyrus（0.436）,the left inferior parietal angular gyrus（0.428）,the left middle occipital gyrus（0.410）,the left lingual gyrus（0.395）,the right paracentric lobule（0.375）,and the left middle frontal gyrus（0.363）.The SVM model based on fALFF had the best performance,and its prediction accuracy,precision and sensitivity were 92%,90%,100%,respectively,it is suggested that the fALFF value can distinguish the patients with limb numbness after stroke from the healthy controls more accurately,and the fALFF value in the related brain regions may be neuroimaging markers for the patients with limb numbness after stroke.（4）After acupuncture treatment,the fALFF value of left angular gyrus,right postcentral gyrus and right supramarginal gyrus in the patients with limb numbness after stroke increased significantly compared with that before acupuncture treatment（P<0.05）,and there were no significant difference in the remaining brain areas（P>0.05）.In correlation analysis,the fALFF value of left angular gyrus was negatively correlated with SAS score（r=-0.285,P=0.049）and positively correlated with NSA score（r=0.307,P=0.034）.The fALFF value of the right postcentral gyrus was negatively correlated with VAS score（r=-0.618,P=0.000）and positively correlated with NSA score（r=0.438,P=0.002）.While the fALFF value of the right supramarginal gyrus was negatively correlated with VAS score（r=-0.388,P=0.007）and with SAS score（r=-0.376,P=0.009）,and there was a positive correlation with NSA score（r=0.359,P=0.012）.Conclusions:（1）The abnormal brain functional activity in stroke sequelae patients with limb numbness are mainly related to the sensorimotor network,in addition to the default mode network and the basal ganglia network.（2）The fALFF values of cerebelum_crus1,inferior parietal angular gyrus,the supplementary motor area gyrus,angular gyrus,supramarginal gyrus,postcentral gyrus and middle occipital gyrus,selected by machine learning based on recursive feature elimination,can be used to accurately distinguish the patients with limb numbness after stroke from the healthy people,which may be the neuroimaging markers related to limb numbness after stroke.（3）The "Co-regualtion of the body and spirit" acupuncture can significantly relieve the symptoms of limb numbness,relieve anxiety,depression,improve the overall sensory and motor function of stroke patients.The targets of acupuncture were the brain regions closely related to sensorimotor,including left angular gyrus,right postecentral gyrus and right supramarginal gyrus.And the mechanism of acupuncture effect may be through regulating the intensity of spontaneous neural activity in the above-mentioned brain regions.