hide
Free keywords:
-
Abstract:
Introduction:
Resting state fMRI studies have become more and more popular over the last years. Here we investigated the consistency of these measurements by acquiring data from a single subject in multiple sessions spread over a period of 2 years.
Methods:
Up to now 43 time points have been measured on a 3T scanner (Trio a Tim System, Siemens, Erlangen, Germany) using a 12 channel and a 32 channel head coil. Day-time was between 11 am and 7 pm (Fig. 1). In each session 305 whole brain volumes were acquired with open eyes using a TR of 2s (10 min, EPI, TE 30ms, resolution 3mm isotropic).
Data were pre-processed with SPM8 including normalization to the MNI template and removing of signals correlated with realignment parameter and mean white matter signals (rest toolbox, Xiao-Wei 2010). This allows to apply predefined parceling according to anatomical structures (AAL, Tzourio-Mazoyer 2002) for calculating ROI-to-ROI correlation matrices. Mean and standard deviation maps were calculated for the z-transformed correlation values over all repetitions.
In addition, Eigenvector Centrality Maps (ECM, Lohmann 2010) were calculated for both coils by using the program package LIPSIA (Lohmann 2010). Again, mean values and standard deviation were evaluated for all repetitions.
In a second step, a one sample t-test with 5 covariates for linear time, sin and cos of time of the year, as well as sin and cos of day-time was performed to investigate the influence of the different time scales (Fig. 1).
Results:
Analysis of the ROI-to-ROI correlation matrices showed very stable pattern. The biggest standard deviation was found in correlations between fusiform and superior parietal cortex and between middle and inferior temporal cortex, irrespective of the different preprocessing methods (Fig. 2).
With regard to standard deviation for the ECMs the biggest differences were present in the fusiform gyri, the insula and the putamen (Fig. 3). These effects were similar for both head coils.
In the ECM analysis we found significant effects with a two-sample t-test between the two coils due to the different sensitivity profiles (p<0.05, FEW corrected, k>5). The 32 channel head coil is more sensitive near to the surface (frontal) whereas the 12 channel coil showed higher connectivity in lower structures. But we only found significant effects of variation with linear time, time of the year and day-time with a lower threshold of p<0.001 (uncorrected, k>5, Fig. 4). The right precuneus showed a decrease of connectivity for the linear time. Variations with the time of year were most prominent in the SMA and right precuneus with high values in summer and low values in winter.
The analysis of the day-time dependency yielded visual structures (calcarine, cuneus), some (pre-) motor areas in the paracentral lobule and a region in the left insula.
Conclusions:
The high number of repetitions of single subject resting state examinations allows a detailed analysis of changes in the different commonly used measures of resting state fcMRI. Although the main patterns were very stable, we found some differences in several regions systematically varying over time of year and day-time. These within subject variability will be compared with between subject variability found in the HCP database in a further investigation.
