Functional Magnetic resonance imaging (fmri) is a special modality of MRI that reveals brains function in addition to its structure by measuring the hemodynamic response of blood. This signal can either be acquired while doing a special task or during resting state. In this thesis, resting state fMRI data has been utilized to extract feature for classification of autistic and healthy individual. After preprocessing step including brain extraction, slice timing correction, motion correction and registration to MNI standard space, the group independent components were extracted using group concatenated ICA method. A linear regression was used to extract the time course of all networks for each person. In addition, the time course for important brain voxels identified by Power et al. were extracted. Two kinds of features showed proper distinguishing property. First the correlation of brain networks, and the second correlation of brain network with 264 most informative brain voxels. One-sample t-test was used for choosing the most significant features. Using this method for feature selection, 5 features from network-network correlation and 47 features from network-voxel correlation were chosen and fed to linear classifier. The method used in this thesis yields 99.7 accuracy, 99.6 specificity and 99.8 sensitivity which significantly improves classification performance comparing to similar researches
عنوان اصلی به زبان دیگر
عنوان اصلي به زبان ديگر
Using Resting State fMRI for Diagnosis of Autism Spectrum Disorder (ASD)
نام شخص به منزله سر شناسه - (مسئولیت معنوی درجه اول )