| Posttraumatic stress disorder(PTSD)is a prevalent neuropsychiatric disorder which delayed or sustained over a long period after a person experienced or witnessed an exceptional physical or threatening catastrophic shock.With the rapid economic development and car ownership rising,posttraumatic stress disorder patients caused by the motor vehicle accident account for a high proportion of the total incidence.According to previous studies,it is important to have a targeted psychological assessment and early intervention in the diagnosis and post-treatment of post-traumatic stress disorder.However,early diagnosis cannot be achieved through the clinical symptoms and the corresponding diagnostic scale.On the other hand,some researches had also reported that not all traumatic event witnesses or experiencers will eventually suffer from PTSD.In fact its pathogenesis and early diagnosis remains to be further studied.Within our acknowledgments,early alterations could provide evidence for early diagnosis and treatment and early diagnosis of PTSD plays an important role during the treatment.Until now,there are few studies on the diagnosis and prognosis of the patients with posttraumatic stress disorder using the structural and functional alterations of whole brain simultaneously in the acute stage.Therefore,we extract discriminant features from multi-modal magnetic resonance(MR)images onset and implement classification-based prediction between PTSD patients and trauma control subjects in this work.Specifically,discriminant features are a collection of measures derived from grey matter(GM)and white matter(WM).We choose cortical thickness of GM and other three descriptions of WM connection which are fiber count,fractional anisotropy(FA),and mean diffusivity(MD).The cerebral cortex was constructed by using the Desikan-Killiany atlas to divide the brain into 68 regions-of-interest(ROIs)to obtain the average thickness of the cortex as gray matter features.For DTI images,after applying automated anatomical labeling(AAL)to parcellate the whole brain into 90 regions-of-interest(ROIs),the descriptions can be quantified.Then,a weighted clustering coefficient of every ROI connected with the remaining ROIs is extracted as white matter feature.Then the grey and white matter measures are merged into a complete feature matrix.The feature selection method based on support vector machine(SVM)is used to automatically select the most discriminative classification features subset,and the cross-validation is used to evaluate and ensure the generalization of classifier’s performance.The final classification accuracy was approximately 79.86% and the area under receiver operating characteristic curve(AUC)is 0.816 evaluated via dual leave-one-out cross-validation.In addition,the experimental results reveal that either white matter features or grey matter features have certain distinguishing ability,and the combination of those two types of features has a better classification performance than single type.At the same time,consistent and satisfactory classification performance indicates that the proposed method and tricks,as well as the quantified features can be used as relatively reliable early prediction methods and predictors.Besides,we found some brain regions with discriminative ability,and these brain regions are consistent with the existing literature reports.It’s a clear evidence that brain responses to traumatic events at acute phase have great relevance to the evolution of posttraumatic stress disorder. |
PDF Full Text Request