Abstrakt

Time series analysis and identification of rsn using GLM-ICA two class classifier

SP Thaiyalnayaki, O Uma Maheswari

The functionally receptive regions of the brain had always been concealing themselves even from the functional MRIs. This induced the quest of finding an efficient algorithm to address the issue of functional localization of activated regions and classification of the active functional areas of the brain. A sample of 10 Healthy controls and 3 shaky hand septuagenarians, nine acquisitions was considered for this study. The fMRI of three motor imbalance subjects whose scan size was 128 × 128 × 23 and a total of 110 volumes accompanied by three acquisitions for each, who are on an average of 65 ± 3 years were subjected to examination. The conclusion derived from this commendable study was that the number of voxels awakened in the sensory motor network was more in tremulous subjects. Besides the activation of the sensory motor network, tightly localized spatial variations were observed in default mode network (DMN), auditory network, sensory motor and visual networks (medial, occipital), right and left executive control networks of these people. This lead to GLM-ICA two class classifier for decision support which performs pre-processing each rsfMRI scan, encompassing realignment, coregistration, normalisation and smoothing, Extracting independent components from smoothed output, Selecting highest Eigen vectors for retrenching the high-dimensionality of the gathered neuroimages time series, Evolving 2-class classifier using k-means algorithm. The proposed algorithm GLM-ICA two class classifier aided the classification of about 87.5% of the functional localization.