Using carbogen for calibrated fMRI at 7 Tesla: Comparison of direct and modelled estimation of the M parameter
Introduction
Blood oxygenation level-dependent (BOLD) contrast is widely used in functional magnetic resonance imaging (fMRI) to map neuronal activity in the brain (Buxton, 2012, Ogawa et al., 1990). BOLD contrast relies on a complex interplay between changes in cerebral blood flow (CBF), cerebral blood volume (CBV) and cerebral metabolic rate of oxygen (CMRO2). The quantitative relationship between these parameters has not, however, been fully elucidated (Aguirre et al., 1998, Buxton, 2010, Turner, 2002). Additionally, since the BOLD response is a relative change from an unknown baseline, differences in baseline physiology may influence BOLD signal changes, precluding truly quantitative BOLD response comparisons across subjects, or within the same subject across days (Gauthier et al., 2011, Lu et al., 2008). However, more physiologically-specific MRI-based methods have been proposed to detect functional activity in the brain. These are typically based on measuring a single physiological quantity rather than an ambiguous combination of several (Davis et al., 1998, Lu et al., 2003, Williams et al., 1993). Of these physiological quantities, CMRO2 is believed to provide one of the most direct indices of neural activity mapping since it reflects oxidative energy consumption of the tissue (Hoge et al., 1999).
Calibrated fMRI techniques used to quantify CMRO2 in the brain commonly rely on the combined acquisition of CBF and BOLD signals during a gas-breathing manipulation. Either hypercapnic (CO2) (Davis et al., 1998, Hoge et al., 1999) or hyperoxic (O2) (Chiarelli et al., 2007c) breathing challenges are used to elevate venous saturation and thereby the BOLD signal. The BOLD signal increase measured during this gas breathing manipulation is extrapolated to its theoretical maximum M via a biophysical calibration model. The recently introduced generalized calibration model (GCM) extends the hyperoxia calibration model (Chiarelli et al., 2007c), making it valid for combined hypercapnia and hyperoxia breathing challenges (Gauthier and Hoge, 2013).
Common to all models is a dependence on accurate estimation of the calibration parameter M to obtain valid CMRO2 estimates. Previous studies using hypercapnia, hyperoxia or a combination of both have yielded a large range of estimated M-values (Ances et al., 2008, Ances et al., 2009, Bulte et al., 2009, Bulte et al., 2012, Chen and Parrish, 2009, Chiarelli et al., 2007b, Chiarelli et al., 2007c, Driver et al., 2012, Gauthier et al., 2012, Gauthier et al., 2013, Gauthier and Hoge, 2013, Hall et al., 2012, Huber et al., 2013, Ivanov et al., 2012, Leontiev et al., 2007, Leontiev and Buxton, 2007, Lin et al., 2008, Mark et al., 2011, Mark and Pike, 2012, Mohtasib et al., 2012, Perthen et al., 2008). Some of the variance between studies may be explained by differences in haemodynamic and physiological properties across brain regions and subjects (Aguirre et al., 1998, Ances et al., 2008, Chiarelli et al., 2007a, Lu et al., 2008) or the dependence of M on specific measurement parameters, such as echo time (TE) or static magnetic field strength (B0) (Hoge et al., 1999, Uludağ et al., 2009). However, some uncertainty in M is also expected due to the relatively high probability of extrapolation error when starting from noisy measurements or poorly estimated parameters (Chen and Pike, 2010, Chiarelli et al., 2007b).
A recently introduced technique uses carbogen mixtures with high CO2 and O2 content (10% CO2, 90% O2) to saturate the BOLD signal by elevating venous O2 saturation close to 100% (Gauthier et al., 2011). This approach can be used to directly measure the calibration constant M and circumvent the disadvantages associated with M-value modelling. However, CO2 concentrations at 10% can lead to a significant level of discomfort (Banzett et al., 1996, Gauthier et al., 2011) and are therefore not a viable alternative for routine calibrated BOLD studies. In this study we have applied the direct M-value estimation approach at 7 Tesla using a moderate CO2 concentration of 7% with simultaneous visuo-motor activation, based on previous work in our lab (Ivanov et al., 2012). Since breathing discomfort is known to be related to CO2 concentration (Banzett et al., 1996, Gauthier et al., 2011, Gauthier and Hoge, 2013), this lower CO2 concentration is expected to be more tolerable for participants. The directly estimated results are compared to GCM-based M-value modelling also based on carbogen-7 inhalation. Furthermore, this work is the first to quantitatively assess the direct and GCM-based M-value estimation techniques at ultra-high field in the same experiment. Benefits and drawbacks of both techniques are analysed and compared to assess their feasibility for future calibrated fMRI studies.
Section snippets
Methods
Data were acquired in 15 healthy participants (7 women, 26 ± 3 years) on a Siemens Magnetom 7T whole body MRI scanner (Siemens Medical Solutions, Erlangen, Germany) equipped with a 24 channel head coil (NOVA Medical Inc, Wilmington, MA, USA). All participants gave written informed consent for participation in this study. Ethics approval was obtained from the local review board (Ethics Commission, Leipzig University).
Results
Baseline O2 and CO2 partial pressures were comparable for all subjects. The average PetCO2 and PetCO2 at rest were 107.8 ± 2.4 mm Hg and 40.2 ± 2.2 mm Hg, respectively. End-tidal partial pressures of CO2 during carbogen inhalation showed an average of 53.8 ± 2.3 mm Hg. The average manipulated PetCO2 was 600.5 ± 64.5 mm Hg. Participants' end-tidal partial pressures of O2 (PetCO2) and CO2 (PetCO2) measured during rest and gas manipulation periods are given in Supplementary Table 1.
Visually-evoked ROI-based
Discussion
In this study we applied and compared two concurrent calibrated BOLD techniques at 7 Tesla. The direct calibration method was used under the assumption that complete venous oxygen saturation could be reached using a combined hyperoxia–hypercapnia breathing challenge. To assess whether complete saturation of venous blood was reached using this challenge, subjects in this study also performed a visuo-motor task during gas inhalation. Venous oxygenation levels of 91.1% were obtained during gas
Conclusion
In this calibrated fMRI study, we compared the direct estimation of calibrated BOLD M-values to the generalized calibration model-based approach. Calibrated BOLD modelling suffers from important drawbacks, such as the reliance on assumed parameters whose values have not been fully validated. Additionally, M-value modelling is vulnerable to errors in CBF signal quantification. While the direct method does not suffer from these sources of error, it relies on complete venous oxygen saturation
Acknowledgements
We thank Elisabeth Wladimirow and Domenica Wilfing for helping with the data acquisition, subject handling and scanning organisation. We would also like to thank Robert Trampel for helping with the ethics application and Robert Trampel, Gabriele Lohmann and Jenny von Smuda for the helpful discussions.
Conflict of interest statement
We wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work
References (67)
- et al.
The variability of human, BOLD hemodynamic responses
NeuroImage
(1998) - et al.
Regional differences in the coupling of cerebral blood flow and oxygen metabolism changes in response to activation: implications for BOLD-fMRI
NeuroImage
(2008) - et al.
Stimulus-response characteristics of CO2-induced air hunger in normal subjects
Respir. Physiol.
(1996) - et al.
A general analysis of calibrated BOLD methodology for measuring CMRO2 responses: comparison of a new approach with existing methods
NeuroImage
(2012) - et al.
Quantitative measurement of cerebral physiology using respiratory-calibrated MRI
NeuroImage
(2012) Dynamic models of BOLD contrast
NeuroImage
(2012)- et al.
Modeling the hemodynamic response to brain activation
NeuroImage
(2004) - et al.
Caffeine dose effect on activation-induced BOLD and CBF responses
NeuroImage
(2009) - et al.
MRI measurement of the BOLD-specific flow-volume relationship during hypercapnia and hypocapnia in humans
NeuroImage
(2010) - et al.
Sources of systematic bias in hypercapnia-calibrated functional MRI estimation of oxygen metabolism
NeuroImage
(2007)