The relationship between phase and correlation is shown more clearly in the bottom plots, where cos is in pink and correlation is in black. From viewing these sliding window time courses, it can be seen that phase offsets between the BOLD time courses approach 180 degrees during the post-dose section, and appear to be responsible for the large drop in correlation. This suggests that larger phase differences, rather than variations in joint signal power, during the post-dose state may be responsible for the caffeine-induced increase in correlation variability. Caffeine has been previously shown to reduce stationary measures of the correlation between spontaneous BOLD signal fluctuations in the motor cortex . However, because of the complexity of caffeine’s interaction with both the neural and vascular systems, it remains unclear how BOLD signal correlation is disrupted. In this study, we examined temporal variations in correlation before and after a 200mg dose of caffeine to gain more insight into the physiological mechanisms of caffeine’s effect on functional connectivity. We found that correlations between left and right motor cortex BOLD signals showed significantly more temporal variability following a caffeine dose. Furthermore, these variations appear to be driven by phase differences between the signals. Fluctuations in BOLD correlation, particularly those dependent on phase differences, may reflect the underlying dynamic nature of neural activity coherence. Transient episodes of inter-regional phase-locking of neural activity have been theoretically simulated in models based on the known anatomical connectivity of the primate brain and demonstrated in vivo .
In addition, transient periods of strong correlations were found between magnetoencephalography power fluctuations within distributed resting-state networks,vertical grow system revealing the presence of non-stationary neuronal dynamics in the human brain . While a relationship between BOLD fluctuations and electrical power fluctuations has been shown using simultaneous electroencephalography and fMRI , BOLD fluctuations have generally produced stronger and more stationary correlations within functional networks than those found with EEG or MEG. It has been hypothesized that the hemodynamic response function smoothes out this temporal variability. However, very slow modulations in local field potential power have been shown to exhibit large and variable phase differences between the cat homologues of the DMN and task positive network , which would not be masked by HRF smoothing. Furthermore, previous work has shown variable phase differences between BOLD time courses in the DMN and TPN . These variations are also found in the data presented in this study. Caffeine caused greater variability in the correlation between resting BOLD time courses in the left and right motor cortices. The mode of interhemispheric coordination of spontaneous oscillations in neural activity is not known for certain. One possibility is that homologous activity in the left and right sensorimotor cortex is primarily mediated by the corpus callosum . However, it is also possible that the thalamus, which serves as a relay center for both sensory and motor mechanisms , coordinates spontaneous activity between the motor cortex hemispheres .
Previous studies have shown that caffeine stimulates motor activity by counteracting the inhibitory control exerted by adenosine receptors on striatal dopamine transmission, which will ultimately disinhibit thalamocortical projection neurons . Caffeine’s direct impact on the pathway between the thalamus and the cortex may increase variability in the coordination of neural activity between these two sites, which could lead to the correlation variability between hemispheres observed in this study. As caffeine antagonism of adenosine receptors produces both neural and vascular effects, which in turn both influence the BOLD signal, the increased temporal variability in BOLD correlation might also reflect changes in the vasculature that reduce the BOLD signal’s sensitivity to underlying neural activity. This could be ac-complished either through caffeine’s inhibition of adenosine-induced dilation of blood vessels or its reduction of the ratio between blood flow changes and oxygen metabolism changes in response to neural activity . If the BOLD signal is less sensitive to underlying neuronal fluctuations, then a larger proportion of noise of non-neural origins could be present in the resting-state BOLD time courses. This decrease in signal to noise ratio could lead to the observed reductions in stationary correlation and increased temporal variability in correlation without an accompanying change in the coherence of spontaneous neural activity. However, it is unlikely that increased physiological noise in the BOLD signal is primarily responsible for the findings presented here. This is because we find extended periods of strong correlation in the post-dose condition that are unlikely to be present if there was an overall decrease in signal-to-noise ratio .
Furthermore, preliminary work by our group with MEG measures, which are insensitive to caffeine’s vascular effects, has shown that caffeine also reduces stationary measures of correlation in the motor cortex . Future work with simultaneous electroencephalographic and fMRI measures will be useful in elucidating whether caffeine directly increases phase variability between neural activity in the left and right motor cortices. In conclusion, we find that correlation between the BOLD signals in the left and right motor cortices varies over time, and that this variability is significantly increased by caffeine. The predominant source of these variations appears to be the non-stationarity of the phase differences between the two signals. These results suggest that caffeine causes greater variability in the underlying coherence of neural activity. As caffeine is a widely consumed stimulant, its effects on functional connectivity should be considered as a potential confound for other resting-state disease and drug studies. Furthermore, future studies assessing changes in functional connectivity caused by other pharmaceutical agents or diseases will benefit from considering not just stationary measures of functional connectivity, but dynamic properties as well.The number of studies applying resting-state functional magnetic resonance imaging to explore intrinsic brain activity is growing rapidly. In the absence of an explicit task, low-frequency fluctuations in the blood oxygenation level dependent signal are believed to reflect spontaneous neural activity . Correlations between resting BOLD fluctuations have been used to assess functional connectivity between different brain regions , and comparisons of inter-subject differences in functional connectivity have shown potential in predicting cognitive performance and diagnosing disease. For example, measures of resting-state functional connectivity have been found to correlate with performance on working memory tasks and intelligence in healthy subjects. In addition, several studies have demonstrated that resting-state functional connectivity is altered in patients with cognitive disorders such as Alzheimer’s disease, schizophrenia, multiple sclerosis, epilepsy, and attention deficit hyperactivity disorder . In these studies, it is often assumed that differences in the correlation of BOLD fluctuations correspond to differences in the coherence of underlying neural activity between subjects. In addition to functional connectivity measurements, there has been increasing interest in the amplitude of resting BOLD fluctuations. This metric has been shown to vary across brain region, resting condition, and disease state, and thus may provide additional information about underlying neural activity .Although resting-state fMRI studies are finding increase use, the interpretation of inter-subject differences in spontaneous BOLD fluctuation correlation or amplitude is complicated by the BOLD signal’s dependence on both neural and vascular factors. The BOLD signal reflects deoxyhemoglobin levels in the blood and is a complex function of changes in the cerebral metabolic rate of oxygen , cerebral blood flow , and blood volume that follow neural activity . Factors that alter metabolism or the vasculature can therefore produce changes in the BOLD response, even if neural activity remains the same.
Previous work has shown that inter-subject differences in the task-evoked BOLD response amplitude depend on a number of hemodynamic factors,cannabis grow equipment including vascular reactivity, baseline CBF, and venous oxygenation . It is possible that spontaneous BOLD fluctuations will be similarly modulated by hemodynamic differences. In fact, resting-state BOLD fluctuation amplitudes have been used to normalize task-related BOLD response amplitudes in order to reduce variability . In these studies, the magnitudes of resting BOLD fluctuations were shown to correlate with vascular reactivity measured with hypercapnia , suggesting that spontaneous BOLD fluctuations are particularly sensitive to vascular factors. Accounting for hemodynamic factors that affect resting-state fMRI measures is critical, particularly when studying patients with diseases like Alzheimer’s that are known to alter both the neural and vascular systems. For this study, we acquired simultaneous measures of BOLD and CBF to determine their relationship during the resting state. As the task-evoked BOLD response amplitude exhibits an inverse dependence on baseline CBF across subjects , we hypothesized that a similar dependence might exist for resting BOLD fluctuations. Instead, we found that smaller fluctuations in CBF and tighter coupling between CBF and oxygen metabolism during the resting state weakens the dependence of both BOLD amplitude and connectivity on baseline CBF.Seventeen healthy subjects participated in this study after providing informed consent. The data from 9 of these subjects were previously used as the control data for a separate study examining the effect of a caffeine dose on resting-state functional connectivity measures . The imaging protocol for each subject included a high-resolution anatomical scan, a bilateral finger-tapping block design scan, and three five-minute resting state scans. Bilateral finger tapping was self-paced and the block design consisted of 20s rest followed by 5 cycles of 30s tapping and 30s rest. Subjects were instructed to tap while a flashing checkerboard was displayed and then to rest during the display of a control image, consisting of a white square situated in the middle of a gray background. During resting-state scans, the control image was displayed for the entirety of the scan and subjects were asked to maintain attention on the white square. Images from each scan section were co-registered using AFNI software . Data from the first 10 s of each scan were discarded to allow magnetization to reach a steady state. In addition, voxels from the edge of the brain were excluded from the analysis in order to minimize the effects of motion. CBF-weighted and BOLD weighted time series were computed as the running difference and average of the first and second echo data, respectively . To avoid the confounds of a circular analysis , we used the first two cycles of the finger tapping data to define regions-of-interest and used the last three cycles to estimate activation amplitudes. It should be noted that using the entire finger-tapping dataset to define ROIs and then again to estimate response amplitudes did not change our overall conclusions. Activation maps were generated using a general linear model approach on the BOLD and CBF time courses. The stimulus-related regressor was produced by the convolution of the square wave stimulus pattern with a gamma density function . Constant and linear trends and a respiratory noise term were included in the GLM as nuisance regressors. We defined ROIs for the left and right motor cortices using the intersection of CBF and BOLD activation, which limits BOLD signal bias from the inclusion of large draining veins. For BOLD activation, we used a correlation threshold of r > 0.60, corresponding to p < 0.001. A less conservative threshold of r > 0.32, corresponding to p < 0.05, was used for CBF activation to account for the lower signal to noise ratio in the ASL CBF data. A combined motor cortex ROI was defined as the union of the left and right motor ROIs, i.e. excluding the supplementary motor area. Baseline CBF images were calculated from the average difference between the first echo control and tag images in the resting ASL scan. The mean ASL images were corrected for coil sensitivity and radio frequency field in homogeneities with the minimum contrast scan , and converted to physiological units, ml/, using the CSF reference scan . Average CBF0 values were calculated from each subject’s motor cortex ROI. Nuisance terms were removed from all functional data through linear regression. Nuisance regressors included constant and linear trends, the 6 motion parameters obtained during image co-registration, physiological motion effects , low frequency variations in cardiac and respiratory rate , and a global signal term. The global signal term was calculated as the mean signal from the anterior portion of the brain in order to minimize bias that can be introduced when motor cortex fluctuations are included in the regression . Note that heart rate related regressors were not removed from the task data because the pulse oximeter was not worn during bilateral finger tapping. All time courses were also temporally low-pass filtered using a finite impulse response function with a cutoff frequency of 0.08 Hz. This cutoff frequency was chosen for consistency with previous functional connectivity studies .