15961

14254

كارشناسي ارشد

بيوالكتريك

اصفهان : دانشگاه صنعتي اصفهان

1399

هفت، 173ص. : مصور، جدول، نمودار

رسول اميرفتاحي ورنوسفادراني، فرزانه شايق بروجني

بهزاد نظري، امير اخوان

1399/08/14

كتابنامه

مهندسي برق

مهندسي برق و كامپيوتر

1399/08/14

2647029

Automatic diagnosis of autism spectrum disorder based on effective connectivity in the DMN and investigation of brain hemispheric asymmetry Soroor Shafieizadegan Isfahan s shafeizadegan@ec iut ac ir 2020 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language Farsi Supervisors Dr Rassoul Amirfattahi fattahi@iut ac ir and Dr Farzaneh Shayegh f shayegh@iut ac ir Abstract Approximately one percent of people are diagnosed with Autism Spectrum Disorder ASD in the world and the preva lence of this disease is similar across different countries This disorder defined a group of neurological signs that are usuallyrecognized by deficiencies in interactions and social communications Due to the similarity of autism disorder with a num ber of cognitive disorders in some symptoms diagnosis of this disease especially in children is usually delayed by severalyears The use of functional images especially resting state functional magnetic resonance imaging rs fMRI is a goodprocedure to find accurate biomarkers for the diagnosis of this disease due to its non invasive nature and its independenceof task performance during imaging To better understand the function of brain activity in this disease many studies haveused functional connectivity but these studies have reported inconsistent results The default mode network DMN is animportant brain network in ASD that has the most disruption among functional networks Furthermore brain asymmetryis one of the fundamental aspects of the human brain that changes in many psychiatric disorders In this study to reducethe effect of factors affecting the non convergence of results in brain connectivity studies in ASD effective connectivitiesin the core DMN were calculated using spectral dynamic causal modeling spDCM method these connectivity parametershave used for investigation of brain asymmetry in ASD and classification and diagnosis of this disorder with using machinelearning methods In this regard 20 sites from the ABIDE data set including 1194 people have been used Also in thisstudy the variability of the effective connectivity calculated with the spDCM method has been evaluated with regard tosome pre processing methods using machine learning algorithms The results of the asymmetry study in the DMN networkshow a significant reduction in the hemisphere asymmetry of the brain in the connections between the medial prefrontalcortex and the left and right inferior parietal cortex in ASD compared to typical control subjects In addition the maxi mum classification accuracy to diagnose this disorder according to the proposed methods using the random forest machinelearning algorithm among ABIDE datasets sites belonging to ABIDEII IP site with 96 accuracy ABIDEI Caltech sitewith 92 accuracy and ABIDEII USM site with 90 accuracy and the maximum average accuracy using random forestalgorithm for all sites is obtained 75 23 Also the results indicate effective connectivity with the spDCM procedure withregard to some preprocessing methods does not change which confirms the results of previous studies Key Words hemispheric asymmetry brain connectivity dynamic causal modeling DCM defaultmode network DMN resting state fMRI rs fMRI autism spectrum disorder ASD machine learn ing

رسول اميرفتاحي ورنوسفادراني، فرزانه شايق بروجني

بهزاد نظري، امير اخوان