Dynamic functional connectivity measures using fcMRI

Introduction: Communication between brain areas and the formation of brain networks is key to understanding how the brain functions; furthermore perturbed functional connectivity (fc) within networks is thought to be responsible for some pathologies. Therefore it is important to better understand the processes underlying fc. In many fcMRI studies, fc is measured from the temporal correlation between BOLD timecourses from separate locations over long time windows (~5 min) [e.g. 1]. However, other modalities have shown that fc is non-stationary over time with significant modulation even in the resting state [2], and that this modulation occurs on a timescale accessible to fMRI. This suggests that measurement of fc on a shorter timescales using fcMRI could yield important new information. Indeed recent work suggests that the temporal variability in wavelet transform coherence can be used to asses non-stationarity [3]. Here three specific issues are addressed: 1) An empirical technique is proposed to derive the statistical significance of fc maps created for different time windows. 2) Dynamic changes in motor network (MN) and default mode network (DMN) connectivity over time on a 20 s timescale are demonstrated. 3) It is shown that fc information exists on both short (20s) and long (300s) timescales by filtering out fast fluctuations for long timescale analyses. Methods: Six subjects took part in the study which was approved by the local ethics committee. A BOLD localiser experiment comprising a visually cued finger-tapping task (30s on; 30s off; 5 trials) was performed to plan subsequent slice placement and locate a motor seed region. This was followed by a ‘Motor’ experiment in which 720s of data were recorded: during the first 300s subjects were instructed to remain at rest with their eyes open, for 30060s) timescale. Future work in this area will look to exploit these analysis methods to assess how dynamic connectivity changes during a task positive state. References: [1] Biswal B et al. (1995) MRM 34: 537-541; [2] De pasquale, PNAS; [3] Chang and Glover, [4] Hale et al, MAGMA, 2010 in press. Acknowledgements: MRC, the Wellcome trust and the Leverhulme trust for funding. Fig. 3 Normal vs smoothed timecourse (seed top, ROI bottom) and corresponding correlation maps Fig. 2 Correlation timecourse for the motor network showing correlation maps at different time points for a 20s window