Altered Structural Networks Of Obsessive-compulsive Disorder And Machine Learning Analysis With Multiple Neuroimaging Indices

Objectives:Obsessive-compulsive disorder(OCD)is a common mental disorder characterized by recurrent and intrusive thoughts and/or repetitive uncontrollable behaviors.However,the pathogenesis of OCD remains unclear.Numerous neuroimaging studies and our previous findings suggested that OCD patients exhibited widespread brain structural abnormalities.Structural network topological organization provides a better perspective to clarify the pathological mechanism of OCD.Moreover,neurobiological markers for clinical diagnosis of OCD are of great importance.Therefore,this study aims to carry out three parts of research on the basis of previous work:1.To investigate white matter(WM)integrity abnormalities in OCD.2.To explore changes in topological organization of structural network in drug-naive OCD patients.3.Finally,machine learning approach was performed based on multimodal and multiple neuroimaging indices to confirm the best biological marker of OCD.Methods:1.We acquired diffusion tensor imaging(DTI)scans from 52 OCD patients and 46 healthy controls(HC).Fractional anisotropy(FA)and mean diffusivity(MD)derived from DTI were compared using tract-based spatial statistics(TBSS)by FMRIB Software Library(FSL).The disease severity was evaluated by score of the Yale-Brown Obsessive-Compulsive Scale(Y-BOCS).For those brain regions exhibiting altered structure,correlations between alterations and clinical symptoms severity were analyzed in all patients(n=52)and medication-naive patients(n=27),respectively.2.We acquired DTI datasets from 28 drug-naive OCD patients and 28 well-matched HC.A deterministic fiber tracking approach was used to construct the whole-brain structural connectome.Group differences in global and nodal topological properties as well as rich-club organizations were compared by using graph theory analysis.The relationships between the altered network metrics and the total Y-BOCS score,obsession score,and compulsion score were calculated.3.Structural magnetic resonance imaging(sMRI)and diffusion tensor imaging(DTI)data were acquired from 48 OCD patients and 45 well-matched HC.sMRI images were preprocessed using voxel-based morphometry(VBM)to obtain gray matter volume(GMV)and white matter volume(WMV).FA and MD were extracted the same way as part ?.Above four features were examined using support vector machine(SVM).Ten brain regions which contributed the most to the classification of each feature were also estimated.Results:1.Comparison of white matter integrity:OCD patients exhibited significantly decreased FA values in the genu and body of corpus callosum(CC,P<0.05,FWE corrected).FA values in the genu of the CC and the body of the CC did not significantly correlate with total Y-BOCS score,obsession score,or compulsion score in either the whole OCD group(n=52)or the medication-naive OCD group(n=27).2.Comparison of structural networks:(1)Both OCD and HC groups presented small-world characteristics.Compared with controls,OCD patients exhibited a significantly decreased small-worldness(?),normalized clustering coefficient(?)and shortest path length(Lp),as well as an increased global efficiency(Eglob).(2)The nodal efficiency(Enodal)was found to be reduced in the left middle frontal gyrus,and increased in the right parahippocampal gyrus and bilateral putamen in OCD patients.(3)OCD patients showed increased rich-club,feeder and local connection strength,and the connection strength of the rich-club was positively correlated with the total Y-BOCS score.3.Machine learning analysis by multiple indices:(1)Using different algorithms,the classifier achieved accuracies of 72.08%,61.29%,80.65%,and 77.42%for GMV,WMV,FA,and MD,respectively(permutation P<0.05).The DTI indices perform better than the sMRI indices.(2)The most discriminative gray matter regions that contributed to the classification included right anterior cingulate gyrus(ACG),right angular gyrus,right inferior parietal,bilateral paracentral lobule,left inferior frontal gyrus,and bilateral cerebellum regions.For WMV feature and the two feature sets of DTI,the shared regions contributed the most to the discrimination mainly included the uncinate fasciculus,the cingulum in the hippocampus,corticospinal tract,as well as cerebellar peduncle.Additionally,external capsule,fornix and stria terminalis,anterior corona radiate,cerebral peduncle were among the informative regions for classification.Conclusions:1.OCD patients demonstrated altered WM integrity in the genu and body of CC.CC might be a vital region in the neuropathological mechanism of OCD.2.Medication-naive patients group showed significantly decreased ?,indicating the disruption of functional segregation and functional integration.OCD patients congruously exhibited increased Eglob,Enodal in various regions,rich-club connection strength,feeder connection strength and local connection strength.The disruptions of the core nerve fiber bundles might lead to a compensatory increase in connectivity between different brain regions,and also across the whole brain,which may be the mechanism of recurrent and intrusive thoughts and/or repetitive uncontrollable behaviors.Rich-club brain regions were more likely to be related with the severity of symptoms directly because of the key role they played.3.Based on whole-brain volumetry and DTI images,SVM algorithm revealed high accuracies for distinguishing OCD patients from healthy subjects at the individual level.DTI indices performed better than sMRI indices during the classification.OCD patients might be characterized by abnormalities in extensive and distributed brain regions,and machine learning methods were more sensitive to detect differences in spatial distribution.The uncinate fasciculus,the cingulum in the hippocampus,corticospinal tract,as well as cerebellar peduncle reserved the highest value for identifying OCD.This might provide a new perspective for future clinical diagnosis of OCD.