Supervised Multimodal Fusion And Its Application In Autisms Spectrum Disorders

As a heterogeneous neurodevelopmental disorder,autism spectrum disorder is characterized by severe and pervasive deficits in reciprocal socialization,communication,and repetitive or unusual behaviors.Studies have shown that the core symptoms shared by various sub-types of autism spectrum disorders are social disorders and communication disorders,mainly result of social motivation and social cognitive deficits and impairments.Most of the existing studies on autism were based on single modal or unsupervised data mining.Compelling evidence shows that,comparing with single modality analysis,multimodal magnetic resonance imaging fusion can capture brain properties from multiple aspects,and provide complementary inf'ormation.Multimodal fusion can extract the co-variability between different modalities in brain diseases.Numerous studies have been reported using data driven method to excavate mental disorders alteration pattern,including schizophrenia,major depression disorder.However,study about multimodal image fusion in autism has not been carried out yet.Therefore,we aimed to identify the multimodal brain regions in autism particularly associated with Social reciprocity deficits,through a novel supervised,fusion with reference method..The main contents and work of this paper are as follows:Based on spc8,data was preprocessed in order to check whether the image quality meet the experiment criterion,remove noise and improve the accuracy of the variant brain reugions.Three input characteristics were calculated based on functional and structural magnetic resonance image data:fMRI(fractional amplitude of low-frequency fluctuation),fMRI(regional homogeneity),sMRI(cortical thickness);Then,based on a supervised multimodal fusion method,multivariate canonical correlation analysis with ref'erence+independent component analysis(MCCAR?jICA),three features were separated.The covariant multimodal brain areas were significantly group different between autism and healthy control,and were significantly related to social impairment.Finally,same experiments were performed on independent validation data set,processing,statistical analysis,and multimodal fusion analysis,in order to verify the stability of the brain areas.