Edited by: Pedro Antonio Valdes-Sosa, Centro de Neurociencias de Cuba, Cuba
Reviewed by: Baxter P. Rogers, Vanderbilt University, USA; Xin Di, New Jersey Institute of Technology, USA
*Correspondence: Michael P. Milham, Center for the Developing Brain, Child Mind Institute, 445 Park Avenue, New York, NY 10022, USA e-mail:
This article was submitted to Brain Imaging Methods, a section of the journal Frontiers in Neuroscience.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Blood oxygenation level dependent (BOLD)–based functional MRI (fMRI) is a widely utilized neuroimaging technique for mapping brain function. Hematocrit (HCT), a global hematologic marker of the amount of hemoglobin in blood, is known to impact task-evoked BOLD activation. Yet, its impact on resting-state fMRI (R-fMRI) measures has not been characterized. We address this gap by testing for associations between HCT level and inter-individual variation in commonly employed R-fMRI indices of intrinsic brain function from 45 healthy adults. Given known sex differences in HCT, we also examined potential sex differences. Variation in baseline HCT among individuals were associated with regional differences in four of the six intrinsic brain indices examined. Portions of the default (anterior cingulate cortex/medial prefrontal cortex: ACC/MPFC), dorsal attention (intraparietal sulcus), and salience (insular and opercular cortex) network showed relationships with HCT for two measures. The relationships within MPFC, as well as visual and cerebellar networks, were modulated by sex. These results suggest that inter-individual variations in HCT can serve as a source of variations in R-fMRI derivatives at a regional level. Future work is needed to delineate whether this association is attributable to neural or non-neuronal source of variations and whether these effects are related to acute or chronic differences in HCT level.
Blood oxygenation level dependent (BOLD) contrast functional magnetic resonance imaging (fMRI) is one of the most widely utilized non-invasive imaging techniques that indirectly measure brain functions. In particular, BOLD-fMRI relies on the magnetic properties of hemoglobin (Hb)—the metalloprotein in red blood cells that transport oxygen (Ogawa et al.,
In fact, the task-based fMRI literature has a number of studies investigating the impact of HCT level on magnitude of BOLD activation. For example, Levin and colleagues have reported a positive relationship between HCT and percent signal change of the BOLD signal within primary visual cortex in response to photic stimulation in males but not females (2001). Further, manipulating the HCT level in the same male participants by isotonic saline hemodilution resulted in a reduction in BOLD activation, demonstrating a causal relationship. Using a motor task, Gustard and colleagues replicated the positive linear relationship between HCT and BOLD signal within the motor cortex (2003). These findings are consistent with animal studies showing that manipulating HCT level through hemodilution can change the T2*-weighted signal intensity in anesthetized rats (Lin et al.,
Given the relationship between HCT and task-evoked BOLD activation, along with observations linking resting state phenomena (e.g., low frequency fluctuation amplitudes) with task-based activation (Mennes et al.,
Here, we address this gap, by testing for relationships between HCT and inter-individual differences in an array of commonly employed R-fMRI derivatives, including: (1) Degree Centrality (DC, the number of significant connections of a given voxel; Zuo et al.,
The initial data included 531 participants from the Enhanced Nathan Kline Institute (NKI)/Rockland lifespan sample. Written informed consent was obtained from all participants prior to participation, as approved by the Nathan Kline Institute Institutional Review Board. Since HCT level changes dramatically across the lifespan and the normal range are different for children, adults, and aged people, we only included adults between 18 and 50 years of age to limit variability. A total of 226 participants are within this age range (42.6%), of which 180 completed the 2-day protocol and have both HCT and imaging data. The blood sample was obtained on day 1 and the R-fMRI scans on day 2 (~1–2 weeks apart; Nooner et al.,
As HCT level is also affected by medication, disease status, and psychobiological factors, we carefully selected our sample to control for these factors by excluding participants who: (1) have a positive drug test (e.g., cocaine, cannabinoid); (2) have medical conditions (e.g., diabetes, hypertension); (3) are currently taking medication (see Supplementary Table
Positive drug test | 19 |
Medical condition | 52 |
Currently taking medication | 12 |
Currently diagnosed or had a history of psychiatric disease | 93 |
Females who had the most recent menstrual period the day before blood draw | 9 |
Females who had menopause before the age of 40 | 4 |
HCT value | 0 |
Over 50% of the volumes with frame-wise displacement (FD) > 0.2 mm | 0 |
Mean FD outside of 3 inter-quartile range | 1 |
A 5 ml venous blood sample was collected at study entry during the first visit and tested between 2 and 6 h for Hb concentration and HCT level.
Imaging data were acquired using a 3.0 Tesla Siemens TrioTim scanner at NKI. For each participant, a 10-min resting-state functional MRI scan was acquired using multiband echo-planar imaging (EPI) sequence (900 volumes;
Imaging data were preprocessed using an alpha version of the Configurable Pipeline for the Analysis of Connectomes (CPAC version 0.3.4,
Depending on the approach, spatial normalization and spatial smoothing happened either before or after the derivative was calculated (see next section for details). Spatial normalization included: (1) structural-to-standard registration using Advanced Normalization Tools (ANTs,
Based on the R-fMRI data, we computed the following six voxel-wise derivatives for each participant at the individual-level (DC was calculated in standard space and then smoothed. VMHC was calculated on smoothed data in standard space. ReHo, ALFF, fALFF, and DR was calculated in native space and then registered to MNI space and smoothed):
DC, a graph theory-based measure, identifies the most connected nodes (i.e., “cortical hubs”) within the whole-brain functional network (i.e., the functional connectome) (Zuo et al., ReHo, measures local coherence of intrinsic brain activities and is defined as the Kendall's coefficient of concordance (KCC) of the time series of a given voxel with those of its 26 nearest neighboring voxels (Zang et al., ALFF: the standard deviation of the bandpass filtered (0.01–0.1 Hz in the present study) fMRI signal, measures the intensity of low frequency oscillations (Zang et al., fALFF, the square root of the ratio of ALFF to the sum of amplitudes of the entire frequency range, measures the relative contribution of low frequency oscillations to the power of the whole detectable frequency range (Zou et al., VMHC, measures functional connectivity between each pair of symmetric inter-hemispheric voxels (Zuo et al., DR, measures the functional connectivity of large-scale networks (Filippini et al.,
Across participants, we used the GLM implemented in a toolbox for Data Processing and Analysis of Brain Imaging (DPABI;
Sex, age, and race were included as nuisance variables because these are known factors that have an impact on HCT. Mean FD was included to control for the residual effect of head motion (Yan et al.,
Group analyses were constrained within the same study-specific mask as the one used to calculate DC. The results were corrected for multiple comparisons using Gaussian random field theory (voxel threshold:
As expected, HCT and Hb are significantly positively correlated for both males (
We first examined whether HCT has a global effect on each of the derivatives. We found that the global means are not associated with either variables of interests (i.e., HCT and HCT × Sex), but mainly associated with age (for ReHo, DR_Executive control, DR_right_Frontoparietal) and motion (for DC, ALFF, fALFF, VMHC, DR_Visual pole, DR_Auditory. DR_right_Frontoparietal). See Supplementary Table
The brain areas within which the intrinsic properties are associated with HCT regardless of sex were detected by the main effect of HCT (Figure
DC | B Frontal pole/Paracingulate gyrus | 10/32 | Default | 3 | 58 | 17 | 1856 |
fALFF | L Central opercular cortex/Insula/Putamen | 48 | Somatomotor/VA | −34 | −2 | 9 | 847 |
VMHC | SPL/SMG | 7/40 | FP | −35 | −50 | 40 | 207 |
DR_Occipital Pole | R SPL/PCC/precuneus | 5/7/26 | DA/Default | 13 | −46 | 41 | 707 |
L SPL/LOC/ | 5/7/40 | DA | −22 | −55 | 51 | 710 | |
DR_Lateral Visual | B Frontal pole/Paracingulate gyrus | 10/32/48 | Default | −13 | 43 | 19 | 1454 |
L precuneus/LOC/cuneus | 7/18/19 | Visual | −14 | −73 | 34 | 692 | |
DR_Cerebellum | L Lingual gyrus/Fusiform gyrus | 18/19 | Visual | −22 | −69 | −6 | 652 |
DR_Sensorimotor | B Insula/Thalamus/Striatum/Brain stem | 48 | Somatomotor/FP/VA/Limbic | −6 | −14 | 6 | 1634 |
DR_Auditory | L Cerebellum | − | FP/Default | −22 | −73 | −34 | 685 |
R SMG/Angular gyrus/LOC | 39/40/48 | DA/VA/Default | 54 | −47 | 25 | 734 |
Besides the areas commonly associated with two measures, several areas were associated with HCT in one measure. fALFF within the left putamen is negatively associated with HCT. The connectivity between the occipital pole network and the right posterior cingulate cortex (PCC)/precuneus and between the cerebellum network and the left lingual and fusiform gyrus are negatively associated with HCT. The connectivity between the lateral visual network and the left precuneus/cuneus, between the sensorimotor network and the bilateral insula and bilateral subcortical areas (including thalamus, striatum, and brain stem), and between the auditory network and the left cerebellum and the right temporoparietal junction are positively associated with HCT. DR measures the functional connectivity of large-scale networks, including both within- and between-network connectivity. On the whole-brain DR map, the connectivity within the original spatial template is typically strong in all subjects and the connectivity outside of this core functional architecture is more variable across subjects. We found that connections outside of the original ICN are more associated with individual differences in HCT, suggesting a larger impact of HCT on these connections.
Consistent with the known sex differences in HCT, certain HCT effect appears to be sex-dependent (Figure
VMHC | IFG (pars opercularis)/Precental gryus | 44 | Default/FP/DA | −52 | 11 | 22 | 232 |
DR_Medial visual | B Frontal pole/Paracingulate gyrus | 9/10/32 | Default | −5 | 54 | 25 | 658 |
DR_Cerebellum | L LOC/Lingual gyrus/Fusiform gyrus | 18/19 | Visual | −30 | −74 | −6 | 809 |
We explored the possibility that variations in baseline HCT levels among individuals may have systematic associations with R-fMRI findings; this concern arises from task-based fMRI findings relating HCT and the BOLD percent signal change. Of the R-fMRI measures examined in the present work, ALFF is most similar to percent signal change (Kannurpatti et al.,
A key question that cannot be addressed by the current observational, cross-sectional design is whether the associations observed reflect current HCT level at the time of sampling, or neural adaptations to chronic differences in HCT between individuals. This latter possibility is not without precedent. Studies of the effects of chronic hypoxia in high altitude residents have demonstrated brain functional and structural differences, which directly related to oxygen transport (Yan et al.,
Given the close relationship between HCT level and oxygen metabolism (Ogawa et al.,
The observed associations between HCT and intrinsic brain indices may have implications for the study of inter-individual and population differences in a number of contexts. For example, a variety of experimental manipulations have been shown to impact HCT (e.g., induction of psychological stress, acute or long-term exercise training, and pharmacologic challenges) (Patterson et al.,
A few noteworthy limitations exist for the present study. First, although we started with a relatively large sample, our decision to limit variability by restricting our analyses to healthy individuals resulted in a moderate sample size (
Although the biophysical model predicted that HCT has an impact on BOLD signal, we did not observe an association between HCT and ALFF, the R-fMRI measure most similar to task-evoked BOLD percent signal change. We observed regional associations between baseline blood HCT level and four variance-normalized intrinsic brain measurements (i.e., DC, fALFF, VMHC, and DR), some of which are modulated by sex. Future studies that directly manipulate HCT, using hemodilution or otherwise, are necessary to differentiate whether these effects are acute or due to neural adaptations to chronic differences in HCT.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
This work was supported by grants from the National Institute of Mental Health (BRAINS R01MH094639 to Michael P. Milham), grants from the Child Mind Institute (1FDN2012-1 to Michael P. Milham), and gifts to the Child Mind Institute (Michael P. Milham) from Phyllis Green, Randolph Cowen, and Joseph P. Healey.
The Supplementary Material for this article can be found online at:
1Race was coded as “American Indian,” “Asian,” “Black,” “Native Hawaii,” “White,” and “Other” originally. Because the percentage of races other than “White” is relatively low, we combined them and recoded race as Caucasian (