Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Magn Reson Imaging. Author manuscript; available in PMC 2011 September 1.
Published in final edited form as:
PMCID: PMC2936716

Functional connectivity in BOLD and CBV weighted resting state fMRI in the rat brain



To directly compare functional connectivity and spatiotemporal dynamics acquired with BOLD and CBV-weighted fMRI in anesthetized rats.

Materials and Methods

A series of BOLD images were acquired in 10 rats followed by CBV-weighted images created by injection of ultra-small iron oxide particles. Functional connectivity, spectral information, and spatiotemporal dynamics were compared for the BOLD and CBV-weighted resting state scans.


BOLD scans exhibited higher cross-correlation values compared to CBV-weighted scans, but the spatial patterns of correlation were similar. The BOLD spectrum contains power evenly distributed throughout the low frequency range while the CBV power spectrum exhibited a high power peak localized to ~0.2 Hz. Both BOLD and CBV resting state scans showed similar propagating waves of activity along the cortex from the SII toward MI; however, these waves were detected more often in BOLD scans than in CBV scans.


While the power spectrum of the CBV signal is different from that of the BOLD signal, both connectivity maps and spatiotemporal dynamics are similar for the two modalities. Further experiments should address the relationship between spontaneous neural activity, local changes in metabolism, and hemodynamic fluctuations to elucidate the origins of the BOLD and CBV signals.

Keywords: BOLD, CBV, functional connectivity, low frequency fluctuations, resting state, rat cortex

Background and Introduction

Clinical interest in mapping functional connectivity with MRI continues to grow as the technique has demonstrated the ability to detect alterations in patients with disorders such as Alzheimer’s (1), schizophrenia (2), and depression (3). Despite this promising evidence of sensitivity to clinically relevant changes, the interpretation of functional connectivity data remains limited by our incomplete understanding of the interactions between the local changes in neural activity, metabolism, and hemodynamics that lead to the low frequency BOLD fluctuations.

To elucidate the origins of functional connectivity, several groups have turned to animal models (48). Their work has demonstrated that functional connectivity similar to that observed in awake humans can be detected in other species such as rodents, even though anesthesia is typically required to facilitate imaging. These animal models provide a platform for investigations of the relationship between spontaneous neural activity, local metabolic changes, variations in blood flow, and MRI signal fluctuations (9,10). The rodent brain in particular has been well-characterized by neuroscientists through recording techniques, selective lesioning, and behavioral studies, providing an extensive framework for the design and interpretation of functional connectivity experiments. Rodent models also offer the advantage of high inter-subject homogeneity, and the use of high-field dedicated small animal MRI systems provides excellent spatial and temporal resolution.

In humans, functional connectivity studies are performed almost exclusively with BOLD contrast (1114). In animal models, however, both BOLD and cerebral blood volume (CBV) weighting have been used to map correlated signal fluctuations (68), raising the issue of whether BOLD and CBV-weighted studies supply comparable measurements of functional connectivity. The BOLD signal comprises a complicated combination of several hemodynamic and metabolic properties including CBV, cerebral blood flow (CBF), and the local rate of oxygen consumption (CMRO2). It is not yet clear if spontaneous fluctuations in each of these parameters are equally linked to spontaneous fluctuations in neural activity. For example, 0.1 Hz oscillations in CBF and CBV have been observed with multiple modalities and are often attributed to vasomotion, and some studies have not found a clear link between these oscillations and electrical activity (1517). By measuring hemodynamic parameters that contribute to the BOLD signal, such as CBV, it may be possible to determine if particular contrasts are more closely related to neural activity than others.

Previous work with stimulus-induced activation suggests that CBV-weighted imaging may offer enhanced sensitivity and increased functional localization compared to BOLD (18,19), partially due to reduced contributions from large vessels. It is possible that this same property will improve localization in functional connectivity studies. In this study, BOLD and CBV-weighted data are acquired sequentially from the same rat to determine the relative sensitivity and specificity of the two techniques.

In addition to examining the steady-state characteristics of the BOLD and CBV functional connectivity maps, this study also compares the spatiotemporal dynamics of the signal fluctuations. Functional connectivity scans are conventionally analyzed using seed region-based cross-correlation techniques (11,12,14,20); however, a recent article by Majeed et al. (21) describes a method for visualizing the spatiotemporal dynamics of the low frequency fluctuations. The data presented by Majeed et al. challenges standard interpretations of functional connectivity maps acquired in the anesthetized rodent, as waves of BOLD signal propagate along the cortex, connecting areas that exhibit little correlation in their time courses.

Dynamic spatiotemporal analysis of resting state functional connectivity (RSFC) scans has only been conducted using the BOLD signal to date. At short TRs, the BOLD signal is heavily weighted toward CBF due to inflow effects, which could be partially responsible for the spatiotemporal dynamics observed in Majeed’s work (21). CBV-weighted imaging using ultra-small paramagnetic iron oxide (USPIO) particles to provide contrast is not typically susceptible to inflow effects because the signal from the blood is diminished due to the presence of iron oxide, so the presence of propagating waves in the CBV-weighted signal would suggest that they are not primarily due to inflow effects.

The purpose of this study is twofold; first, to determine whether measurements of BOLD and CBV RSFC provide comparable information for functional connectivity mapping and second, to provide insight into the relative contribution of CBV information to the BOLD signal. BOLD and CBV-weighted data from the same rats will be examined using spectral analysis, traditional cross correlation analysis, and dynamic spatiotemporal visualization. The results of this study show that BOLD and CBV provide similar maps of functional connectivity and demonstrate that the propagating waves previously observed in the BOLD signal can be detected with CBV-weighted imaging, suggesting that these dynamics are a general hemodynamic phenomenon widely observed in anesthetized rodents.

Materials and Methods

Animal preparation and physiological monitoring

All experiments were performed in compliance with guidelines set by the Institutional Animal Care and Use Committee (IACUC). Ten male Sprague-Dawley rats weighing 200–300g were initially anesthetized with 2.5% isoflurane (mixed with 1:1 oxygen and room air) and maintained at 1.5% isoflurane throughout the experimental setup. Heart rate and blood oxygen saturation were recorded with a pulse oximeter placed on the hind paw. Body temperature was monitored with a rectal thermometer and maintained at 36° C − 38° C using an adjustable temperature water circulating pad. The respiratory rate was monitored using a pressure sensitive balloon placed under the rat’s chest. To provide forepaw stimulation for activation studies, two needle electrodes were inserted underneath the skin of the rat’s left forepaw between digits 2 and 3 and digits 3 and 4. The rat’s tail vein was catheterized to allow for injection of iron oxide contrast agent later in the experiment. Finally the rat was placed in the MRI cradle and the head was secured with a bite bar and ear bars.

Once experimental setup was complete, the rat was given a subcutaneous bolus injection of 0.05 mg/kg medetomidine (Domitor, Pfizer, Karlsruhe, Germany). Three minutes after the medetomidine bolus, isoflurane was discontinued. Fifteen minutes post bolus, subcutaneous infusion of 0.1 mg/kg/hr medetomidine was initiated to maintain anesthetic depth for the duration of the experiment (22).

Image acquisition

Images were acquired on a 9.4 T / 20 cm horizontal bore Bruker BioSpec magnet interfaced with an AVANCE (Bruker, Billerica, MA) console. The magnet was equipped with an actively shielded gradient coil capable of producing a maximum gradient strength of 20 G/cm with a rise time of 120µs. A two-coil actively decoupled imaging setup was used (2cm surface coil for reception and 7 cm diameter volume coil for transmission; Bruker, Billerica, MA) to achieve maximal SNR over the cortical areas of interest. Shimming was initially performed on a 6 mm × 6 mm × 6 mm region visually centered over the primary somatosensory cortex using FASTMAP. A single-shot, multi-slice gradient echo EPI sequence (TR=1500ms, TE = 15ms, FOV = 2.56 cm × 2.56 cm, matrix size = 64 × 64, slice thickness = 2 mm, effective bandwidth = 200 kHz, flip angle = 31°) was used to acquire a time series of 180 BOLD weighted images for the duration of a forepaw stimulation paradigm (A.M.P.I. Master-8 Stimulator; 4mA, 9 Hz; 30 imaging volumes off - 20 on - 30 off - 20 on - 30 off - 20 on - 30 off). Activated voxels in the primary somatosensory cortex were identified using STIMULATE (23) which correlates the timecourses of voxels in the acquired image set with a box car function representing on and off times for stimulation. Following the forepaw stimulation scan, a BOLD weighted dataset was acquired without stimulation with the following parameters: TR = 300ms, TE = 15ms, number of repetitions = 1200, FOV = 2.56 cm × 2.56 cm, matrix size = 64 × 64, slice thickness = 2mm, 1 slice centered over the forepaw regions of the primary somatosensory cortex (located using data from the preceding stimulation scan). Manual shimming was then performed on the single slice to remove any local field inhomogeneities. The same imaging setup was repeated two more times to obtain a total of three sets of BOLD weighted image series. After BOLD scans, 2% isoflurane was administered to further sedate the rat to allow for tail vein infusion of 5 mg/kg of USPIO particles (Molday Ions, BioPal, Worcester, MA) over a period of ~1 minute. Isoflurane was discontinued after USPIO infusion was completed. Thirty minutes after USPIO injection, three sets (1 stimulation scan and 1 resting state scan) of images were acquired with identical parameters to the resting state BOLD images but with cerebral blood volume (CBV) weighting due to USPIO particles.

Data analysis

All data processing was performed using MATLAB (MathWorks, Natick, MA). Each time course used for analysis was de-meaned (mean value of timecourse subtracted from each data point), quadratically detrended, and normalized by converting the data to percentage difference relative to mean voxel intensity. Image signal to noise ratio (mean brain signal / standard deviation of the background noise), temporal variation (variance of demeaned and detrended timecourse from SI), and activation percent changes were calculated for the CBV and BOLD data sets.

The primary somatosensory cortex was localized by cross-correlating time courses from the forepaw stimulation scan with a box car reference function representing stimulus on and off times. A seed region for cross correlation analysis of the resting state data was constructed from the nine voxels which were most highly correlated with the stimulus reference function.

Power spectra were obtained for the time courses of selected ROIs by using the Welch method (8 sections with 50% overlap, Hamming window). Visual inspection of the power spectra revealed that peak power was shifted to higher frequencies for CBV data as compared to BOLD. Time courses were bandpass filtered based on initial inspection of the power spectra using a 3rd order butterworth filter to retain contributions from a 0.2 Hz frequency window (0.05 Hz – 0.25 Hz for BOLD, 0.1 Hz – 0.3 Hz for CBV). The window was shifted higher by 0.05 Hz for CBV data to prevent attenuation of the peak typically seen at 0.2 Hz. Nine voxels in SI activated by forepaw stimulation were used as the seed region for cross-correlation analysis of resting state data. The average time course representing the voxels in the selected ROI was correlated with the time course from voxels throughout the image and a correlation map was obtained.

Additional 3×3 voxel ROIs were selected based on anatomy in the secondary somatosensory cortex (SII) and the caudate/putamen (CP) complex directly from the EPI images (no visible distortion was apparent in the EPI images which would negatively affect the manual selection of ROIs; Figure 2 contains the actual EPI images with locations of seed ROIs) for further cross-correlation analysis [in accordance with the Paxinos and Watson rat brain anatomical atlas (24)]. A 6×6 average correlation matrix representing data from all rats was calculated from the average time courses from six seed regions (left and right SI, SII, and CP) in order to visualize the strength of connectivity between the regions.

Figure 2
Cross correlation maps for CBV and BOLD resting state scans. Correlation maps are overlaid on the EPI image used to create the map. Locations of the seed regions are shown in the first row. Bilateral connectivity is evident for BOLD and CBV for all three ...

Spatiotemporal dynamics of the low frequency signal fluctuations were analyzed using image by image visualization of data normalized and filtered to contain only low frequencies (21). The resulting images were displayed as a movie in order to locate visually detectable spatiotemporal patterns or events. If spatiotemporal patterns in the BOLD and CBV signals contain similar time and spatial signatures this would suggest that the BOLD and CBV MRI signals are affected similarly by coordinated neural activity.


SNR, temporal variance, and signal change during stimulation

SNR, temporal variance, and average percent change in signal during stimulation were calculated for CBV and BOLD (Table 1). The use of USPIOs to create CBV contrast leads to dose dependent (USPIO dose) decreases in image signal-to-noise ratio and a dose dependent increase in the absolute value of the signal percent change during activation (25). As expected in this study SNR decreased significantly with the addition of USPIOs (p<0.01). The absolute value of the average percent change during stimulation was found to be significantly greater for BOLD as compared to CBV (p<0.01). The average temporal variance for time courses representing spontaneous activity in SI was similar for CBV and BOLD (p = 0.36); however the difference in temporal variance between CBV and BOLD was significant in the SII and CP regions (p <0.01). The USPIO concentration of 5 mg/kg used in this experiment was chosen based on preliminary experiments to approximately match absolute maximum signal percentage change ( ~2% from baseline) observed in the BOLD images during forepaw stimulation.

Table 1
BOLD and CBV temporal variance, SNR, and percentage signal change during forepaw simulation.

Spectral analysis

Figure 1 shows the individual power spectra for two rats and the average power spectra of all ten rats for time courses from the SI for BOLD and CBV resting state scans. The power spectra from the BOLD scans had maximum power in the very low frequency range with a broad distribution of power across the low frequencies (< 0.3 Hz). There was less power in the very low frequency range in the CBV scans, and the distribution of power was more localized than in the BOLD scans, with a distinct peak often appearing near 0.2 Hz. The center of mass for the low frequency power (< 0.3 Hz) for BOLD occurred at 0.13 Hz. The center of mass for all low frequency power (< 0.3 Hz) in the CBV scans was localized to 0.16 Hz. For both CBV and BOLD power spectra, just before 0.3 Hz power was reduced abruptly to a baseline level (low power noise at all frequencies) in a nearly identical pattern.

Figure 1
Average CBV and BOLD power spectra are plotted for two representative rats. The average BOLD and CBV power spectra for all ten rats is shown in the bottom plot. BOLD data exhibits higher power in the very low frequencies (less than 0.1 Hz) and a broader ...

Seed-based correlation analysis

Representative cross-correlation maps of CBV and BOLD for all three seed regions (SI, SII, and CP) are shown in Figure 2. CBV cross correlation was significantly more localized than BOLD in the cortex contralateral to the seed region in band passed SI and SII data (for the SI region). BOLD connectivity based on a seed region in the left SI resulted in an average of 45 ± 17 voxels exceeding half of the maximum cross correlation value in the bilaterally symmetric area as compared to 26 ± 13 for CBV (p<0.01).The average cross correlation value (of voxels greater than half of the maximum correlation value) for voxels in the SI BOLD data was 0.50 ± 0.15 in the 3×3 region bilaterally symmetric to the ROI while the average cross correlation in the SI CBV data was 0.25 ± 0.08 (p< 0.01). Connectivity and extent values for SI, SII, and CP are located in Table 2.

Table 2
BOLD and CBV functional connectivity statistics.

To determine whether the 0.2 Hz peak observed in the CBV data had different properties from the lower frequency range typically used in functional connectivity studies, connectivity maps were created using only the low frequency range (< 0.1 Hz) for CBV and BOLD (Figure 2; first column) in SI. BOLD and CBV connectivity strength for the lower frequency data resulted in an increase in average connectivity values (0.55 ± 0.12 and 0.34 ± .14 respectively) which was significant (p = 0.04 and p < 0.01 respectively) . The spatial extent of connectivity in the contralateral cortex was reduced slightly for BOLD (voxels greater than half max connected to seed = 43 ± 18) and increased slightly for CBV (voxels greater than half max connected to seed region = 26 ± 15); the change was not significant for BOLD (p = 0.57) or for CBV (p = 0. 84). Although the spatial extent of connectivity in the contralateral cortex does not change significantly, the number of voxels connected to the seed region outside of the contralateral cortex greatly increases for CBV weighted scans. The low pass filter images in Figure 2 show increased connectivity in subcortical brain regions for CBV weighted images while the low pass BOLD connectivity maps look almost identical to the band pass maps. The low pass CBV images also contain several voxels located in the skull weakly connected to the seed region. Low frequency connectivity in CBV scans based on a seed region in SI resulted in widespread, but relatively weak cross correlation throughout cortical and subcortical areas which was not seen in the higher frequency (0.1Hz – 0.3 Hz) cross correlation analysis. Low frequency BOLD correlation did not exhibit the widespread connectivity pattern seen in the CBV scans (Figure 2).

Averaged cross correlation matrices (all rats) showing the connectivity between averaged time courses in six regions of interest (right SI, SII, CP and left SI, SII, and CP) are shown in Figure 3. The pattern of connectivity between contralateral analogues is similar for BOLD and CBV (highest for SI and lowest for CP); however, the correlation values for BOLD scans is significantly higher for all regions. The average value of correlation between left and right SI was 0.66 for BOLD and 0.31 for CBV. Anatomically selected seed regions in the SII region resulted in average bilateral correlation values of 0.58 for BOLD and 0.22 for CBV, and seed regions selected in the CP resulted in average bilateral cross correlation strength of 0.37 for BOLD and 0.12 for CBV. Weak correlation is observed between non-analogous areas for both contrast mechanisms.

Figure 3
Average cross correlation matrix showing the strength of connectivity between six areas (left and right SI, SII, and CP) for CBV and BOLD. For both contrast mechanisms, the highest correlation values are between left and right SI, followed by left and ...

Spatiotemporal dynamics

Spatiotemporal analysis was conducted using image to image visualization of low frequency fluctuations in the filtered, normalized BOLD and CBV resting state data sets following the method described by Majeed et al. (21). Both data sets showed well organized bilateral waves of increased signal in the optimal frequency range (BOLD - 0.05 Hz – 0.25 Hz; CBV - 0.1 Hz – 0.3 Hz) that propagated along the cortex from SII towards the primary motor cortex (MI) (Figure 4). These waves occurred often within each 6 minute resting state data set. Movies displaying the propagating waves are available online in the supplemental material for this article. The very low frequency range for BOLD and CBV (<0.1Hz) was also analyzed for propagating waves. Propagating waves in this low frequency range were seen sparsely for BOLD and never for CBV.

Figure 4
Propagating waves of activity from SII towards SI in BOLD and CBV resting state scans. a.) 0.05 Hz – 0.25 Hz filtered BOLD scan. Propagation from SII to MI takes approximately four seconds. b.) 0.1 Hz – 0.3 Hz filtered CBV scan. Propagation ...

The waves occurred spontaneously and typically occurred in groups of two or three repeating waves before ceasing for several seconds or several minutes. The time from the increased signal inception in SII to its disappearance in MI is defined as the wave propagation time. Propagation of the waves from SII to MI took approximately four seconds for the 0.05 Hz – 0.25 Hz BOLD data and approximately three seconds for the 0.1 Hz – 0.3 Hz CBV data. The increased signal intensity moved toward MI and eventually fades away (Figure 4; last frame for BOLD and CBV). These propagation times are similar to those measured by Majeed et al. in the α-chloralose anesthetized rat (21).


The work described here directly compares two types of contrast (BOLD and CBV) used in previous functional connectivity rodent studies. Spontaneous fluctuations of the BOLD signal arise from increasing and decreasing metabolic demand as a function of local neural activity. The vasculature responds to this metabolic demand by increasing the volume (CBV) and flow (CBF) of blood to an area. Changes in oxygen metabolism (CMRO2), CBV, and CBF can be measured independently with MRI, and much research has focused on how changes in these parameters in response to stimulation give rise to the task-related BOLD signal. However, the relationship between the recorded BOLD and CBV signals in the absence of a task has not been explored.

Comparison with previous functional connectivity studies in rodents

The functional connectivity maps obtained with BOLD and CBV weighted imaging in this study are similar to those reported in previous rodent studies. BOLD studies performed by Pawela and Zhou also found strong connectivity between bilaterally symmetric SI, SII, and CP regions (4,7). Lu et al. observed similar connectivity between bilaterally symmetric SI regions using CBV (8). While these functional connectivity studies were performed with a longer TR (1 – 1.5 s) and no correction for physiological noise, the results were similar to those obtained in this study, suggesting that respiratory noise may not have a significant impact on functional connectivity in rodent models. Majeed et al. utilized a TR of 100 ms, short enough to resolve the primary cardiac contribution as well as the primary respiratory peak (21). The functional connectivity maps were similar to those presented here suggesting that cardiac noise contamination does not alter correlation patterns in rodents; however, Majeed’s rats were anesthetized with alpha-chloralose and were mechanically ventilated. We have mapped the location of contributions from the respiratory signal in the freely breathing medetomidine anesthetized rats from this study (Figure 5). It does not appear that respiratory noise corrupts the functional connectivity data because of its localization to the base and edges of the brain for both BOLD and CBV. At the TR used in this study (300ms), we are unable to resolve the contributing signal from cardiac noise.

Figure 5
Map of peak in power spectrum repesenting respiratory noise. The majority of the noise is localized to the base and edge of the brain. This should not influence functional connectivity mapping.

Most functional imaging studies conducted in rodent models are performed in anesthetized animals, and the choice of anesthesia can impact the results. α-chloralose has traditionally been the primary anesthesia for functional imaging in the rat, but it requires intubation and mechanical ventilation, cannulation of the femoral artery for blood gas monitoring, and subsequent sacrificing of the animal at the end of the experiment (2527). In this experiment a continuous infusion of medetomidine anesthesia is used, which allows for longitudinal studies (22). The results are similar to functional connectivity experiments carried out in animals anesthetized with α-chloralose (8,21).

Isoflurane was administered to each rat for ~2 minutes during tail vein injection of the USPIOs for CBV imaging. Isoflurane could potentially alter the vascular response and could be partially responsible for the change in the LFFs. Two control rats were imaged to ensure that the brief administration of isoflurane nor the injection process was responsible for the change in the power spectrum observed between BOLD and CBV scans. Each of the control rats were administered isoflurane while injecting a dose of saline equal to the dose of USPIOs they would have otherwise received. Neither control rat exhibited a change in functional connectivity, power spectra, or SI activation to a forepaw stimulation as compared to the previous BOLD scans after administration of the isoflurane and saline.

BOLD vs CBV weighting for functional connectivity mapping

Similar spatial patterns of connectivity were observed for both BOLD and CBV-weighted data using seed regions in SI, SII, and CP. The strongest correlation was observed between bilaterally symmetric regions, with very little correlation between non-analogous regions in either the ipsilateral or contralateral hemisphere. These findings are in agreement with previous studies of functional connectivity in the rodent, which predominantly found bilateral patterns of correlation (4,6,7).

Cross correlation analysis conducted with the lower frequency spontaneous fluctuations (< 0.1 Hz) resulted in a global increase in the number of voxels connected to the seed region for CBV scans; however, there was very little change in the connectivity patterns for the BOLD data. The widespread connectivity in the low frequency CBV data is not unexpected considering there is less power located in the low frequency (< 0.1 Hz) portion of the spectra as compared to the BOLD spectra. Correlation of the seed region with such a widespread group of voxels including several voxels in the skull (Figure 2; CBV - low pass) suggests an increased contribution from noise sources as opposed to correlated CBV signals. The high power in the low frequency BOLD signal may represent other physiological processes, such as metabolic fluctuations that are also tied to neural activity, so the functional connectivity maps appear unchanged.

We chose the dose of USPIO based upon preliminary experiments so that approximately equivalent temporal variance and percent signal change during stimulation were obtained in the BOLD and CBV-weighted scans. Similar temporal variance was considered a priority since it provides some measurement of the relative amplitude of the spontaneous signal fluctuations. While the temporal variance was well-matched for BOLD and CBV scans in this study, the absolute percent change during activation was significantly higher for the BOLD scans (p<0.01). This may indicate that the fluctuations in the CBV-weighted scans contained relatively lower contributions from hemodynamic processes and relatively stronger influences from system noise, in agreement with the observation that SNR was reduced in CBV weighted images. This reduction in SNR may account for the lower correlation values observed in the CBV data.

A USPIO concentration of 5 mg/kg was used for CBV scans. In order to optimize dosing for functional connectivity studies, several combinations of USPIO concentrations and TRs were tested in preliminary experiments. USPIO doses between 1 mg/kg and 15 mg/kg were tested with TRs ranging from 100 ms to 1000 ms. A TR of 300 ms was chosen to allow the sampling and removal of the primary respiratory component while maintaining acceptable SNR. With this TR, a USPIO concentration of 5mg/kg provided approximately the same percentage change during stimulation and level of temporal variance observed in BOLD studies. The ideal dose of iron oxide for functional connectivity studies may be different from the dose used for functional studies, particularly when the available SNR is limited due to the use of a short TR. Higher percent changes during stimulation in CBV-weighted scans can be achieved by using higher doses of USPIO (25,27), but the resulting signal loss may degrade the quality of functional connectivity maps. Previous CBV functional connectivity studies in rodents employed larger doses of USPIO than were used in the present study, but cross-correlation values were higher in our study.

The high degree of similarity between the functional connectivity maps created with BOLD and CBV suggests that results from studies using different contrast mechanisms should be readily comparable. Because CBV-weighted scans have a lower SNR, BOLD is potentially a better choice for studies with high temporal resolution and limited SNR. The use of BOLD contrast also eliminates the need for exogenous contrast agent.

Spectral differences between BOLD and CBV

Recent work in the α-chloralose-anesthetized rat reported two low frequency peaks with different characteristics (21), motivating the further exploration of frequency-dependent differences in BOLD and CBV data in this study. Resting state data collected in the rat model using the BOLD signal reveals LFFs with relatively uniform power within a frequency range of 0.01 Hz – 0.30 Hz, with one or two localized low frequency peaks, similar to previous experiments in α-chloralose-anesthetized rats (6,8,21). CBV weighted data collected from the same animals exhibit different spectral properties, with a strongly localized peak at ~ 0.2 Hz. The differences in the two power spectra are related to the frequency signatures of the components contributing to the measured signal. The BOLD spectrum is derived from a combination of several signals (CBF, CBV, CMRO2) with independent frequency contributions which results in a relatively uniform power band between 0.0 Hz – 0.3 Hz. The CBV spectrum ideally contains only one signal, blood volume changes, which oscillate at approximately 0.2 Hz during resting state. Despite the differences in the power spectra obtained with CBV and BOLD weighting, the resulting connectivity maps are very similar, suggesting that the processes contributing to different frequency bands share a common origin.

In addition to fluctuations arising from spontaneous variations in neural activity, low frequency oscillations in these signals may also arise from sources that may or may not be closely tied to neural activity. Oscillations of approximately 0.1 Hz have been observed in CBF using laser-Doppler flowmetry (15), and in combination CBV, CBF, and oxygen saturation data using reflective light imaging (16). While some studies report a relationship between these oscillations and neural activity (15), others do not (17). It appears that these fluctuations can be decoupled from neural activity, as they have also been observed in isolated arteries (28). It is therefore possible that particular hemodynamic processes exhibit slow oscillations that are partially related to neural activity and partially due to other factors. If so, it may be possible to identify frequency ranges and/or hemodynamic contrasts that are more specific to neural activity than the broadband BOLD signal typically used. Combined with Majeed et al.’s study showing that low and high frequency peaks in the BOLD signal from α-chloralose-anesthetized rats have different spatiotemporal dynamics, our finding that the CBV power spectra exhibits a localized peak in the higher frequencies as compared to the broader BOLD signal suggests that perhaps the higher frequency fluctuations are linked to ongoing regulatory processes. In support of this idea, the patterns of spatiotemporal dynamics linked to the higher peak in both types of scans are highly reproducible, both within and across animals. The regulatory processes could be vascular, neural, or a combination of both. An intriguing possibility is that functional connectivity is mediated by both a low frequency regulatory ‘driver’, possibly in the brainstem (29) and by communication between strongly connected areas, which may preferentially contribute to different frequency ranges of the BOLD fluctuations. However, in this study little difference was observed between functional connectivity mapped with BOLD as compared to CBV. This would suggest that the regulatory processes either dominate the fluctuations used to map functional connectivity in the anesthetized rodent, or that any vasoregulatory processes are tightly tied to neural activity and provide nearly identical information about functional connectivity. Further work in human subjects or unanesthetized animals would assist in addressing these issues.

Spatiotemporal dynamics

The detection of a spatiotemporal pattern of propagation from lateral to medial cortical areas reproduces the results recently reported by Majeed et al., but at a lower field strength (9.4 T rather than 11.7 T) and using a different anesthetic (medetomidine instead of alpha-chloralose). The number of occurrences of the waves in this study was lower than that previously observed, possibly due to a change in magnetic field strength or an effect of altered vascular response or neuro-vascular coupling in response to the anesthesia being used. These waves of signal were also observed in CBV-weighted images for the first time. These findings demonstrate that the patterns of spatiotemporal propagation are common to multiple preparations in the anesthetized rat.

In the previous work by Majeed et al., a TR of 100 ms was used, making the images sensitive to the effects of inflowing blood (and thus CBF fluctuations). This study also utilized a relatively short TR (300 ms; chosen to maximize signal while allowing the primary contribution from respiration to be removed by filtering), but the use of USPIO should reduce or eliminate the signal from the inflowing blood, suppressing inflow effects. Nevertheless, a well defined peak at 0.2 Hz was observed, and the spatiotemporal dynamics of the CBV fluctuations were similar to those observed in the previous study with BOLD, suggesting that inflow effects are not the primary source of the spatiotemporal waves observed in the cortex.

The presence of propagating waves in the cortex illustrates the limitations of traditional functional connectivity studies. In both BOLD and CBV data, little correlation is observed between SI and SII even though some type of connectivity is present, since the signal waves clearly move between the two areas. The lack of correlation is likely due to the time lag (2–3 s) between the two areas, rather than a lack of relationship between their time courses.

In conclusion, it is important to further explore the relationship between neural activity and the physiological processes (CBF, CBV, and CMRO2) with low frequency components contributing to the BOLD RS signal. We have imaged the CBV signal, one contributor to the BOLD signal, independently to determine its contribution to the overall BOLD signal. The results show that while the power spectrum of the CBV signal is different from that of the BOLD signal, both connectivity maps and spatiotemporal dynamics are similar for the two modalities. This may indicate that FC is primarily reflective of coordinated fluctuations within the vasculature which may be fully or partially driven by neural activity. The links between neural activity and vasomotion are still unclear and further experiments will be necessary to better understand the origin of the spontaneous MRI signal oscillations (30).

Supplementary Material

Supplementary Movie 1


Averaged spatiotemporal dynamics of low frequency fluctuations in the resting state band pass filtered (0.05 Hz – 0.25 Hz) BOLD data in one rat. Propogating waves can be seen moving from SII towards MI.

Supplementary Movie 2


Averaged spatiotemporal dynamics from band pass filtered (0.1 Hz – 0.3 Hz) CBV data from one rat Propogating waves moving from SII towards MI can clearly be obsterved.


We would like to thank the NIH (NIH T32EB005969 and NIH 1R21NS057718-01) for supporting Matthew Magnuson throughout the duration of this work.

Grant Support:

NIH T32EB005969

NIH 1 R21NS057718-01


1. Wang K, Liang M, Wang L, et al. Altered functional connectivity in early Alzheimer's disease: a resting-state fMRI study. Hum Brain Mapp. 2007;28(10):967–978. [PubMed]
2. Hoptman MJ, D'Angelo D, Catalano D, et al. Amygdalofrontal Functional Disconnectivity and Aggression in Schizophrenia [PMC free article] [PubMed]
3. Cullen KR, Gee DG, Klimes-Dougan B, et al. A preliminary study of functional connectivity in comorbid adolescent depression [PMC free article] [PubMed]
4. Pawela CP, Biswal BB, Cho YR, et al. Resting-state functional connectivity of the rat brain. Magn Reson Med. 2008;59(5):1021–1029. [PMC free article] [PubMed]
5. Shmuel A, Leopold DA. Neuronal correlates of spontaneous fluctuations in fMRI signals in monkey visual cortex: Implications for functional connectivity at rest. Hum Brain Mapp. 2008;29(7):751–761. [PubMed]
6. Williams K, LaConte S, Peltier S, Keilholz S. MRI evidence of RS in rodent model. Proceeding of the International Society for Magnetic Resonance in Medicine. 2006;14:2116.
7. Zhao F, Zhao T, Zhou L, Wu Q, Hu X. BOLD study of stimulation-induced neural activity and resting-state connectivity in medetomidine-sedated rat. Neuroimage. 2008;39(1):248–260. [PMC free article] [PubMed]
8. Lu H, Zuo Y, Gu H, et al. Synchronized delta oscillations correlate with the resting-state functional MRI signal. Proc Natl Acad Sci U S A. 2007;104(46):18265–18269. [PubMed]
9. Logothetis NK, Pauls J, Augath M, Trinath T, Oeltermann A. Neurophysiological investigation of the basis of the fMRI signal. Nature. 2001;412(6843):150–157. [PubMed]
10. Shmuel A, Augath M, Oeltermann A, Logothetis NK. Spontaneous fluctuations in functional MRI signal reflect fluctuations in the underlying neuronal activity. Neuroimage: 13th Annual meeting of the organization for human brain mapping; 2007.
11. Biswal B, Yetkin FZ, Haughton VM, Hyde JS. Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn Reson Med. 1995;34(4):537–541. [PubMed]
12. Cordes D, Haughton VM, Arfanakis K, et al. Mapping functionally related regions of brain with functional connectivity MR imaging. AJNR Am J Neuroradiol. 2000;21(9):1636–1644. [PubMed]
13. Hampson M, Peterson BS, Skudlarski P, Gatenby JC, Gore JC. Detection of functional connectivity using temporal correlations in MR images. Hum Brain Mapp. 2002;15(4):247–262. [PubMed]
14. Lowe MJ, Mock BJ, Sorenson JA. Functional connectivity in single and multislice echoplanar imaging using resting-state fluctuations. Neuroimage. 1998;7(2):119–132. [PubMed]
15. Golanov EV, Yamamoto S, Reis DJ. Spontaneous waves of cerebral blood flow associated with a pattern of electrocortical activity. Am J Physiol. 1994;266(1 Pt 2):R204–R214. [PubMed]
16. Mayhew JE, Askew S, Zheng Y, et al. Cerebral vasomotion: a 0.1-Hz oscillation in reflected light imaging of neural activity. Neuroimage. 1996;4(3 Pt 1):183–193. [PubMed]
17. Vern BA, Leheta BJ, Juel VC, LaGuardia J, Graupe P, Schuette WH. Interhemispheric synchrony of slow oscillations of cortical blood volume and cytochrome aa3 redox state in unanesthetized rabbits. Brain Res. 1997;775(1–2):233–239. [PubMed]
18. Mandeville JB, Marota JJ. Vascular filters of functional MRI: spatial localization using BOLD and CBV contrast. Magn Reson Med. 1999;42(3):591–598. [PubMed]
19. van Bruggen N, Busch E, Palmer JT, Williams SP, de Crespigny AJ. High-resolution functional magnetic resonance imaging of the rat brain: mapping changes in cerebral blood volume using iron oxide contrast media. J Cereb Blood Flow Metab. 1998;18(11):1178–1183. [PubMed]
20. Xiong J, Parsons LM, Gao JH, Fox PT. Interregional connectivity to primary motor cortex revealed using MRI resting state images. Hum Brain Mapp. 1999;8(2–3):151–156. [PubMed]
21. Majeed W, Magnuson M, Keilholz SD. Spatiotemporal dynamics of low frequency fluctuations in BOLD fMRI of the rat. J Magn Reson Imaging. 2009;30(2):384–393. [PMC free article] [PubMed]
22. Weber R, Ramos-Cabrer P, Wiedermann D, van Camp N, Hoehn M. A fully noninvasive and robust experimental protocol for longitudinal fMRI studies in the rat. Neuroimage. 2006;29(4):1303–1310. [PubMed]
23. Strupp JP. Stimulate: A GUI based fMRI analysis software package. Neuroimage. 1996;3(S607)
24. Paxinos G, Watson C. The rat brain in stereotaxic coordinates. fifth edtion. Academic Press: San Diego; 2005.
25. Lu H, Scholl CA, Zuo Y, Stein EA, Yang Y. Quantifying the blood oxygenation level dependent effect in cerebral blood volume-weighted functional MRI at 9.4T. Magn Reson Med. 2007;58(3):616–621. [PubMed]
26. Keilholz SD, Silva AC, Raman M, Merkle H, Koretsky AP. Functional MRI of the rodent somatosensory pathway using multislice echo planar imaging. Magn Reson Med. 2004;52(1):89–99. [PubMed]
27. Keilholz SD, Silva AC, Raman M, Merkle H, Koretsky AP. BOLD and CBV-weighted functional magnetic resonance imaging of the rat somatosensory system. Magn Reson Med. 2006;55(2):316–324. [PubMed]
28. Osol G, Halpern W. Spontaneous vasomotion in pressurized cerebral arteries from genetically hypertensive rats. Am J Physiol. 1988;254(1 Pt 2):H28–H33. [PubMed]
29. Drew PJ, Duyn JH, Golanov E, Kleinfeld D. Finding coherence in spontaneous oscillations. Nat Neurosci. 2008;11(9):991–993. [PubMed]
30. Sirotin YB, Das A. Anticipatory haemodynamic signals in sensory cortex not predicted by local neuronal activity. Nature. 2009;457(7228):475–479. [PMC free article] [PubMed]