Spatial–temporal clustering analysis in functional magnetic resonance imaging

In recent years, the temporal clustering analysis (TCA) method has been introduced to analyze functional MRI (fMRI) data without prior information about the activation patterns or experimental paradigms. It has been successfully applied to situations under which the timing of events of interest is not known. However, useful information regarding the spatial correlation of activation pixels with their neighbors is not taken into account in the original TCA (OTCA) method. In this study, we propose a new method called 'STCA' (spatial-TCA) which incorporates spatial information with the TCA method to improve the sensitivity in detecting the time window. The spatial information is defined as the correlation coefficient of the time activity curve between each pixel and its neighbors. The inclusion of spatial information can effectively reduce the contribution from noisy pixels and enhance the sensitivity. Both simulated data and in vivo fMRI experiments are employed to verify the method. Preliminary results show that the proposed method has increased the sensitivity significantly for in vivo fMRI data in detecting the activation response time as compared to both OTCA and modified TCA (MTCA). The OTCA/MTCA was applied to spatially smoothed data for various contrast-noise ratios and compared to STCA. The SNR improvements of both OCTA/MTCA are obvious but blurring effects are also visible. The STCA does not have this artifact.