Journal of Cerebral Blood Flow & Metabolism (2007), 1–12 & 2007 ISCBFM All rights reserved 0271-678X/07 $30.00 www.jcbfm.com
Dynamic forcing of end-tidal carbon dioxide and oxygen applied to functional magnetic resonance imaging Richard G Wise1,2, Kyle TS Pattinson1,3,4, Daniel P Bulte1, Peter A Chiarelli1, Stephen D Mayhew1, George M Balanos5, David F O’Connor3, Timothy R Pragnell3, Peter A Robbins3, Irene Tracey1,3 and Peter Jezzard1 1
Centre for Functional Magnetic Resonance Imaging of the Brain (FMRIB), Department of Clinical Neurology, University of Oxford, John Radcliffe Hospital, Oxford, UK; 2CUBRIC, School of Psychology, Cardiff University, Park Place, Cardiff, UK; 3Department of Physiology, Anatomy and Genetics, University of Oxford, South Parks Road, Oxford, UK; 4Nuffield Department of Anaesthetics, University of Oxford, John Radcliffe Hospital, Oxford, UK; 5School of Sport and Exercise Sciences, University of Birmingham, Edgbaston, Birmingham, UK Investigations into the blood oxygenation level-dependent (BOLD) functional MRI signal have used respiratory challenges with the aim of probing cerebrovascular physiology. Such challenges have altered the inspired partial pressures of either carbon dioxide or oxygen, typically to a fixed and constant level (fixed inspired challenge (FIC)). The resulting end-tidal gas partial pressures then depend on the subject’s metabolism and ventilatory responses. In contrast, dynamic end-tidal forcing (DEF) rapidly and independently sets end-tidal oxygen and carbon dioxide to desired levels by altering the inspired gas partial pressures on a breath-by-breath basis using computer-controlled feedback. This study implements DEF in the MRI environment to map BOLD signal reactivity to CO2. We performed BOLD (T2*) contrast FMRI in four healthy male volunteers, while using DEF to provide a cyclic normocapnichypercapnic challenge, with each cycle lasting 4 mins (PETCO2 mean6s.d., from 40.961.8 to 46.461.6 mm Hg). This was compared with a traditional fixed-inspired (FICO2 = 5%) hypercapnic challenge (PETCO2 mean 6s.d., from 38.262.1 to 45.661.4 mm Hg). Dynamic end-tidal forcing achieved the desired target PETCO2 for each subject while maintaining PETO2 constant. As a result of CO2-induced increases in ventilation, the FIC showed a greater cyclic fluctuation in PETO2 . These were associated with spatially widespread fluctuations in BOLD signal that were eliminated largely by the control of PETO2 during DEF. The DEF system can provide flexible, convenient, and physiologically well-controlled respiratory challenges in the MRI environment for mapping dynamic responses of the cerebrovasculature. Journal of Cerebral Blood Flow & Metabolism advance online publication, 4 April 2007; doi:10.1038/sj.jcbfm.9600465 Keywords: BOLD; dynamic end-tidal forcing; FMRI; hypercapnia
Introduction Respiratory challenges have been performed in humans simultaneously with magnetic resonance imaging (MRI) to measure cerebrovascular reserve (van der Zande et al, 2005), and they are now increasingly popular for investigating the complex contrast mechanisms of the blood oxygenation leveldependent (BOLD) functional MRI (FMRI) signal.
Correspondence: Dr RG Wise, CUBRIC, School of Psychology, Cardiff University, Park Place, Cardiff, CF10 3AT, UK. E-mail:
[email protected] This study was Supported by the Wellcome Trust (RGW, Advanced Training Fellowship grant code 067037/Z/02/Z), UK Medical Research Council (RGW, KTSP and PJ), and Dunhill Medical Trust (PJ). Received 3 September 2006; revised 22 December 2006; accepted 4 January 2007
Although MRI is not the only technique for examining cerebrovascular responses to respiratory challenges (others include transcranial Doppler ultrasound (TCD) and positron emission tomography (PET)), MRI does provide good spatial resolution with a temporal resolution that is well matched to the temporal response of the vasculature. Hypercapnia, an elevated partial pressure of arterial carbon dioxide (PaCO2 ), is a convenient means of providing a cerebrovascular challenge and is therefore the most commonly applied challenge in FMRI. Carbon dioxide (CO2) in arterial blood is a potent vasodilator producing a global increase in cerebral blood flow (CBF). Hypercapnia increases BOLD signal (Posse et al, 2001; Rostrup et al, 2000; Vesely et al, 2001). The CBF increase induced by hypercapnia results in increased venous oxyhemoglobin concentration and decreased venous deoxyhemoglobin concentration, producing a longer T2*.
Dynamic end-tidal forcing for FMRI RG Wise et al 2
Respiratory challenges that seek to elevate (hyperoxia) or decrease (hypoxia) arterial oxygen partial pressure (PaO2 ) have also been applied in conjunction with MRI. In studies of cerebral physiology, these have sought to investigate cerebrovascular responses, such as changes in blood flow, arising from modified oxygen levels (Bulte et al, 2007; Floyd et al, 2003; Kolbitsch et al, 2002). Functional magnetic resonance imaging studies have used modified oxygenation levels as a means of understanding and modeling BOLD contrast (Rostrup et al, 1995; Tuunanen et al, 2006). To date, respiratory challenges presented in the MRI scanner have been simple and have not directly attempted to control the most important physiological parameters, namely arterial blood gas concentrations. Instead, the most common method of inducing a change of blood gas concentrations has been to deliver a fixed fraction of CO2 or oxygen that is modified compared with air (FIO2 E20.7% and FICO2 E0%), for example, to FICO2 E5% (see below; Hoge et al, 1999b). We will refer to this as a fixed-inspired challenge (FIC). Also common is the breath-hold. Aimed largely at causing hypercapnia, it inevitably reduces PaO2 , potentially confounding the interpretation of nominally hypercapnia-FMRI studies based on it. Although the FIC is convenient to deliver, the arterial CO2 and oxygen partial pressures reached along with the rate of change of these partial pressures after any FIC transition are variable and unpredictable because they are determined by the individual’s ventilatory response to CO2. The ventilatory response will vary between subjects. Although it is known to depend on body weight, sex, environmental, and other factors such as drug treatment (Dahan and Teppema, 2003) it is to a large degree unpredictable. Changes in arterial CO2 and oxygen partial pressures are reflected closely in the end-tidal expired partial pressures of these gases (PETCO2 and PETO2 , respectively) in healthy subjects, making endtidal measurements a convenient surrogate for arterial partial pressures (Robbins et al, 1990). An additional concern with the FIC approach is the significant interaction between changes in PETCO2 and PETO2 . In studies of hypercapnia, when the inspired fraction of carbon dioxide (FICO2 ) is increased, ventilation increases, causing a rise in PETO2 . In studies of hypoxia, if the inspired fraction of oxygen (FIO2 ) is decreased sufficiently, the resulting increase in ventilation reduces the PETCO2 . Hence a pure change of only one of the arterial gases is achieved rarely using the FIC approach. Dynamic end-tidal forcing (DEF) is a technique that largely overcomes these difficulties. The general technique was developed originally using fastswitching solenoid valves to control gas flow (Howson et al, 1987; Robbins et al, 1982a, b; Swanson, 1972). It has since been developed for use with mass flow controllers of gas flow and adopted for rigorous investigation of ventilatory and cerebrovascular responses to changes in PaCO2 Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
and PaO2 (Ainslie and Poulin 2004; Ide et al, 2003; Poulin et al, 1996, 1998; Poulin and Robbins, 1998). The system was further developed for use in the field at high altitude (Gamboa et al, 2003). It is a modified successor to this field system that we have developed for use in the MRI scanner. The DEF technique uses computer feedback control of inspired PCO2 and PO2 (PICO2 and PIO2 ) to rapidly (breath by breath) and independently achieve specified target levels of PETCO2 and PETO2 . Dynamic end-tidal forcing has not until this study been applied to FMRI, partly because of the technical challenges involved. The principal challenge is to analyze end-tidal gases within the period of a single breath (B4 secs) when the gas analyzers typically need to be sited several meters from the MRI scanner, necessitating a gas sampling tube of several meters in length with its associated gas transport delay. However, the potential rewards of a DEF system for use with FMRI are significant. Computer control of PETCO2 and PETO2 would provide accurate, targetcontrolled, repeatable respiratory challenges that could be automatically synchronized with FMRI data acquisition. Dynamic end-tidal forcing is able to force fast changes in arterial gas partial pressures by transiently delivering greater excursions in inspired gas partial pressures than can be easily accomplished with traditional challenges. Confounds of ‘mixed’ CO2 and oxygen challenges could be avoided, opening the way for better-controlled quantitative studies of BOLD and perfusion image contrast. Dynamic end-tidal forcing would also allow more advanced respiratory experiments aimed at mapping the central networks involved in the control of breathing (McKay et al, 2003). This study aims to implement and hence show the feasibility of DEF for FMRI and to compare it to the traditional FIC. We have chosen to do this using a simple hypercapnic challenge, this being the most common respiratory challenge used in FMRI. We compare the behavior of respiratory parameters during the two techniques to show the differing degrees of control afforded by them. We evaluate the BOLD signal responsiveness to changes in PETCO2 and the effects of concomitant PETO2 changes, using both the DEF and FIC to illustrate the potential improvements in control of the BOLD signal offered by DEF.
Materials and methods Implementation of Dynamic End-Tidal Forcing for Magnetic Resonance Imaging Figure 1 gives a schematic representation of our experimental setup for DEF. A laptop personal computer controlled the system using BreatheDP software (v1.0, Department of Physiology, Anatomy and Genetics, Oxford, UK). Having specified the desired end-tidal partial pressures of O2 and CO2, the controlling computer calculates the required inspired partial pressures to achieve those targets throughout the course of the respiratory experiment, while also monitoring the end-tidal expired levels actually
Dynamic end-tidal forcing for FMRI RG Wise et al 3
comfort (HC150 Respiratory Heated Humidifier, Fisher & Paykell, Maidenhead, UK). The CO2 supply was not humidified because of its comparatively low flow rate. Gas delivery was controlled by four gas-specific mass flow controllers (Model MFC 1559A, MKS Instruments, Wilmington, MA, USA) powered by two two-channel power supplies (Model PR4000, MKS Instruments, Wilmington, MA, USA). BreatheDP software was built on LabVIEW (National Instruments Corp., Austin, TX, USA). An analogue-to-digital (A to D) PCMCIA card (NI DAQCard-1200, National Instruments Corp., Austin, TX, USA) was used for acquisition of gas partial pressures from the gas analyzers and for digitization of the inspired and expired volumes from the volume flow transducer. This card also provided two analogue signal outputs to the mass (gas) flow controller power supplies, the other two channels being supplied by a second PCMCIA card (NI DAQCard-AO-2DC, National Instruments Corp., Austin, TX, USA). Air was supplied by a compressor (Model 150/ 500 fitted with a breathing air filter, Bambi Air Compressors Ltd, Birmingham, UK). Other gases were supplied from high-pressure cylinders or hospital piped gases. For the feedback system to control end-tidal gases, it was necessary to minimize delays in gas mixture delivery and sampling. This was achieved by mixing the four gases in a home-built mixing chamber that rested on the chest of the volunteer. The gas sampling line (4 1.8 m, 1.2 mm internal diameter, monitoring line, Intersurgical Ltd, Wokingham, UK) from the tight-fitting facemask (Hans Rudolf, Kansas City, MO, USA) was kept as short as possible but needed to be long enough to pass out of the
achieved. The gas mixtures were delivered through a fast gas mixing system in which CO2, O2, N2, and air were mixed as closely as possible to the volunteer to minimize delays in delivery. The controlling computer sampled respiratory variables every 10 ms. The direction and timing of respiratory flow (inspiration or expiration) were measured using an MR-compatible turbine volume flow transducer (VMM400, Interface Associates, Laguna Niguel, CA, USA). Tidal gases were constantly sampled from the volunteer’s facemask, and the partial pressures of CO2 and O2 were measured using rapidly responding gas analyzers (Models CD-3A and S-3A; AEI Technologies, Pittsburgh, PA, USA). The BreatheDP control software measured the end-tidal partial pressures of CO2 and O2 (PETCO2 and PETO2 , respectively) from these continuous measurements. It compared these with the desired PETCO2 and PETO2 and then modified, on a breath-by-breath basis, the inspired PCO2 and PO2 (PICO2 and PIO2 ). The feedback control was performed using an integral-proportional algorithm of the deviation of the desired end-tidal partial pressures from those measured (Robbins et al, 1982a, b). The proportional control uses the difference between the target and the actual end-tidal gas values for each breath to modify the breathed mixture and is designed to reduce the effects of breath-to-breath variation, whereas the integral component uses the sum of these differences to reduce drift and steady-state error. During DEF, the total gas delivery rate to the breathing system was maintained at 70 liters per minute (lpm) to avoid rebreathing of expired gases and to minimize the delay in supplying updated gas mixtures to the subject. The O2, N2, and air supplies were humidified for volunteer
O2
(backup)
3T scanner room CO2
CO2
MFC O2
O2
Gas mixing chamber
MFC High pressure gas cylinders
N2
N2
MFC Air
Air
MFC
Compressor
MFC Power Supply
Pulse oximeter
Baffles
Humidifiers
VMM turbine Filter
O2 analyser (backup)
Face mask
MFC Power Supply
Safety button VMM
Laptop
Exhaust
D to A Indicates direction of gas flow "BreatheDP"
Gas sampling line
D to A A to D O2 cell
O2 analyser
R2 flow control (pump)
CO2 cell
Buzzer (default on)
99%
CO2 analyser Rapid gas analysers
Figure 1 The dynamic end-tidal forcing system. The system, controlled by a laptop PC running BreatheDP (v1.0), was used to deliver the hypercapnia challenges to target values of PETCO2 and PETO2 . D to A, digital-to-analogue converter; A to D, analogue-todigital converter; MFC, mass (gas) flow controller; VMM, ventilation measurement module. For delivery of the fixed inspired challenge during the same session, flow of CO2, O2, and N2 were switched off and a mixture of 5% CO2 (balance air) was supplied, alternated with air. Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
Dynamic end-tidal forcing for FMRI RG Wise et al 4
3T magnet enclosure to the gas analyzers situated outside the scan room next to the penetration panel. The time (delay and dispersion) to register 95% of a step change in CO2 partial pressure through the sampling line was measured to be 1.6 sec. Once factored into the evaluation of end-tidal gases by the control software, this delay proved short enough to allow feedback control of end-tidal gases. This short delay was achieved using a suction pump (R2 Flow Control, AEI Technologies, Pittsburgh, PA, USA) placed at the end of the sampling line. Four safety features were incorporated into the end-tidal forcing system (Figure 1) because of the potential for delivering severely hypoxic gas mixtures. First, the volunteer’s heart rate and pulse oxygenation were monitored continuously using a finger pulse oximeter (In Vivo Research, Orlando, FL, USA). Second, the volunteer in the scanner was instructed to actively depress a switch mounted on a handle. Should the volunteer let go of the handle, a buzzer would sound in the control room. Third, an additional (backup) oxygen analyzer (Portable Oxygen Analyzer, Servomex Group Ltd, Crowborough, UK) sampled and provided a continuous display of oxygen partial pressure in the supply to the facemask. This sounded an alarm to the experimenter if the oxygen fraction should decrease below a specified level (15% in this study). Finally, a separate oxygen supply was attached to the mixing chamber in case of a need to provide emergency enrichment of oxygen, independent of the computer-controlled system. The subject’s end-tidal gases were closely monitored by the experimenters.
The Fixed Inspired Challenge For comparison with DEF, each volunteer underwent a fixed inspired challenge (FIC). To provide the FIC, the gas supply system shown in Figure 1 was modified by introducing a supply of premixed 5% CO2 in a balance of air (19.7% O2) (BOC (Special Gases) Ltd, UK). To cycle between periods of hypercapnia and normocapnia, the supply of gas to the volunteer was manually switched between air and 5% CO2 (balance air), while the flow rate was maintained at 30 lpm. This is adequate for preventing rebreathing of exhaled gases and is consistent with the flow rates used in previous FMRI studies of hypercapnia.
Respiratory Challenges Four healthy male volunteers (mean age = 37, range = 32 to 40) were studied under approval of the Oxfordshire Clinical Research Ethics Committee, having given written informed consent. The application of a facemask and introduction of the volunteer into the MRI scanner temporarily modifies breathing (Maxwell et al, 1985). Therefore, 10 mins of quiet air-breathing was allowed for acclimatization after introducing the volunteer into the scanner. This period was used for prescribing image locations. The FIC was performed first in all volunteers to allow the end-tidal partial pressures from that challenge to be used as targets for the subsequent DEF challenge. The FIC run consisted of a period of 3 mins of normocapnia (normal airbreathing, FICO2 E0, FIO2 E20.7%) followed by four cycles of hypercapnia (2 mins, FICO2 E5%, FIO2 E19.7%) and Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
normocapnia (2 mins, air breathing), giving an experimental duration of 19 mins. The initial 3 mins period of scanning the FIC was used to measure resting PETCO2 and PETO2 required to set the targets for the DEF that followed shortly after FIC, while the volunteer remained in the scanner. The DEF respiratory protocol was analogous to that for the FIC. After an initial 3 mins period of normocapnia, there were four cycles of 2 mins hypercapnia and 2 mins normocapnia. Target PETCO2 and PETO2 levels were specified in the DEF setup. The target PETO2 was set to the resting value for each volunteer. The ‘normocapnic’ PETCO2 target was set to be 1 mm Hg above the resting value for each volunteer (Table 1). This was to allow for the fact that DEF can in general only increase PETCO2 from basal levels. The hypercapnic PETCO2 target was set to the same value achieved at the end of hypercapnic periods of the FIC (average across cycles) for each volunteer (Table 1). Although these choices of target PETCO2 for DEF provide a smaller dynamic range, they were chosen to ensure that an easily tolerable maximum PETCO2 level was reached for each volunteer.
Functional Magnetic Resonance Imaging Identical functional imaging acquisitions were performed for FIC and DEF with the same slice prescription. Imaging was performed at 3 T using a Siemens Trio system. Gradient-echo echo-planar imaging was performed with T2*-weighted contrast: TE = 32 ms. One thousand one hundred and forty volumes were acquired with TR = 1 secs, flip angle 701, and an image matrix of 64 64 voxels, giving in-plane voxel dimensions of 3 3 mm. Each volume comprised 16 contiguous 6-mm-thick axial slices. For each subject a T1weighted structural scan was acquired (1 mm in plane voxel size, 3 mm slices). This was used to segment gray matter in subsequent analyses.
Data Analysis Respiratory parameters (gas partial pressures and minute ventilation) were averaged within subject across each of the four periods of normocapnia and hypercapnia and then further averaged across subjects (Table 1). Note that for calculation of parameters in Table 1, the temporal blocks across which averaging was performed were delayed by 20 secs with respect to the nominal onset of each period. This approximate delay was chosen by inspection of the respiratory data (Figure 2) as the period after which the respiratory parameters had begun to show a clear response to the increase or decrease in CO2. Experimentally, the delay of the responses from the nominal onset of each block occurs because of finite times taken for gas switching, mixing, inhalation, and finally physiological responses. As a measure of the oxygen control, the standard deviation of PETO2 across time was calculated for each subject (stPET2O , Table 1). Statistical 2 comparison of respiratory parameters between FIC and DEF was performed using the paired Student’s t-test. Blood oxygen level-dependent contrast images were analyzed using tools from the FMRIB Software Library (FSL, http://www.fmrib.ox.ac.uk/fsl). The four-dimensional time series was motion corrected using MCFLIRT (Jenkin-
Dynamic end-tidal forcing for FMRI RG Wise et al 5
Table 1 Respiratory parameters and gray matter BOLD signal variation during periods of normocapnia and hypercapnia.
DSBOLD(GM) (%) stPETO (mm Hg) 2 stBOLD;O2 (%) Gray matter BOLD signal reactivity to CO2 (%/mm Hg)
Target PETCO2 (mm Hg) PETCO2 (mm Hg)d PICO2 (mm Hg)d Target PETO2 (mm Hg) PETO2 (mm Hg)d PIO2 (mm Hg)d Minute ventilation (l/min)d
Fixed inspired challenge
Dynamic end-tidal forcing
1.9770.31 3.571.1 0.1370.09 0.2770.04
1.27a70.26 1.7a70.6 0.0570.03 0.24b70.05
Normocapnia
Hypercapnia
N/A 38.272.1 2.470.5 N/A 110.374.2 144.570.7 12.571.6
N/A 45.671.4 32.572.8 N/A 112.871.2 133.370.9 17.772.1
Normocapniac
Hypercapnia
40.671.9 46.871.3 40.9e71.8 46.4f71.6 17.8e72.5 34.0f72.1 108.073.9 107.8a73.5 108.2a74.7 132.5a73.4 124.7a74.2 17.3e72.6 22.4b74.4
Group mean (n = 4) and s.d. are given. DEF, dynamic end-tidal forcing; FIC, fixed inspired challenge; DSBOLD(GM), % change in cortical gray matter (GM) BOLD signal in response to the hypercapnic challenge compared with normocapnia; stBOLD;O2 , mean % BOLD signal variation (temporal s.d.) attributable to PETO2 variation, derived from the fit of PETO2 to the cortical gray matter BOLD signal; stPETO , (mm Hg) mean of temporal s.d. of PETO2 . 2 a Indicates a significantly smaller value for DEF than FIC (P < 0.05, one-tailed, paired t-test). b Indicates a significantly different value for DEF and FIC (P < 0.05, two-tailed, paired t-test). c DEF necessitates a mildly elevated ‘baseline’ arterial level of CO2 to hold the PETCO2 constant. The ‘normocapnia’ PETCO2 target was set 1 mm Hg above the measured resting level. d Mean across 2 mins period. e Indicates a significantly larger value for DEF than during the corresponding periods in FIC (P < 0.05, one-tailed, paired t-test). f Indicates no significant difference in the value for DEF and FIC (P > 0.05, two-tailed, paired t-test).
son et al, 2002), and nonlinear high-pass temporal filtering was performed (Gaussian-weighted least squares straight line fitting, with a high-pass filter cut-off at 360 secs). After coregistration of the FMRI volumes and the T1-weighted individual structural scan (FMRIB Linear Image Registration Tool, FLIRT) (Jenkinson and Smith, 2001), regions of gray and white matter were automatically identified from the T1-weighted scans (FMRIB Automated Segmentation Tool) (Zhang et al, 2001). The spatial-mean BOLD time course was extracted from gray matter voxels and then averaged across hypercapnia–normocapnia cycles to produce a mean BOLD response curve for display purposes and to estimate the fractional change in BOLD signal induced by the respiratory challenges (DSBOLD(GM) (%)). PETCO2 was the intended driver of changes in the BOLD signal. We estimated for each subject the time delay between changes in PETCO2 and the subsequent BOLD signal response by cross-correlation of these signals. For the FIC, the group mean (7standard deviation) lag of the BOLD signal after change in PETCO2 was 1373 secs, and for DEF this was 127 5 secs (no significant difference in a paired t-test, P > 0.05). Pooling across both techniques, the group mean lag of BOLD signal after change in PETCO2 was 1274 secs. Therefore, this was used in subsequent analysis as the representative time of the BOLD signal to respond to changes in blood gas partial pressure. We then examined the degree of BOLD signal change explicable by the induced changes in PETCO2 and PETO2 . A linear regression was performed (Matlab 7, The MathWorks Inc., Natick, MA, USA) for the mean gray matter BOLD time course with the PETCO2 waveform, to yield gray matter estimates of BOLD signal responsiveness to CO2. Included in this regression were also the PETO2 (as changes in venous oxygenation are the source of BOLD signal) and minute ventilation (volume breathed per minute) wave-
forms. In this model, minute ventilation was orthogonalized with respect to PETCO2 to remove the component of the ventilation signal that correlates with PETCO2 (Figure 2) and hence to ensure that where they correlate, BOLD signal changes would be assigned to changes in PETCO2 . The BOLD–PETO2 regression coefficient was used to identify for each subject the temporal standard deviation of the BOLD signal accounted for by fluctuations in PETO2 (group mean stBOLD;O2 (%), Table 1). An analogous regression analysis of BOLD signal was performed on a voxel-by-voxel basis using FEAT (FMRIB Expert Analysis Tool, http://www.fmrib.ox.ac.uk/fsl version 5.1). The following pre-processing, in addition to the motion correction and temporal filtering described above, was applied to each subject’s time-series of FMRI volumes: spatial smoothing using a Gaussian kernel of full-width-half-maximum 5 mm and subtraction of the mean of each voxel time-course from that time-course. The FMRI signal was then linearly modeled on a voxel-byvoxel basis using FILM (FMRIB’s Improved Linear Model) with local autocorrelation correction (Woolrich et al, 2001). PETCO2 incorporating the empirically determined 12 secs BOLD time lag was the main regressor of interest. Further regressors were incorporated in the linear model including PETO2 and minute ventilation (orthogonalized with respect to the PETCO2 as described above). To account for breathing and other motion-related artifacts, inspired and expired volume and 6 motion parameters (rotations and translations each about 3 axes) were also included in the linear model. The images of the PETCO2 regression coefficient (or parameter estimate) were transformed into the standard space of the Montreal Neurological Institute (MNI) using FLIRT, and a mean value across subjects was calculated to yield maps of the mean percentage BOLD Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
Dynamic end-tidal forcing for FMRI RG Wise et al 6 Fixed Inspired Challenge Grey Matter BOLD Signal
Dynamic End–Tidal Forcing
(Arb. units)
1.02
1.01
1.01
1
1
48
48
46
46
44
44
42
42
40
40
38
38
40
40
30
30
20
20
10
10
0
0
120
120
115
115
110
110
105
105
100
100
95
95
145
145
140
140
135
135
130
130
125
125
120
120
115
115
25
25
20
20
15
15
10
10
End–Tidal CO2 PETCO (mmHg )
2
Inspired CO2 PICO (mmHg)
2
End–Tidal O2 PETO (mmHg)
2
Inspired O2 PIO (mmHg)
2
1.02
(l/min)
Minute Ventilation
5
0
50
100 150 Time (s)
200
5
0
50
100 150 Time (s)
200
Figure 2 Functional magnetic resonance imaging BOLD and respiratory recordings. The left column presents data for the fixed inspired challenge (FIC) of 5% CO2, whereas the right column shows the result of dynamic end-tidal forcing (DEF) from one representative volunteer. Mean quantities across cycles of hypercapnia within the imaging session are shown, and error bars represent the s.d. across those cycles. The 120 secs duration of the hypercapnic challenge is indicated by the gray bar. The first row shows BOLD signal averaged across gray matter, whereas the second row displays PETCO2 , the dotted line indicating the resting PETCO2 before the challenge began. The elevated inspired CO2 (third row, PICO2 ) causes the raised PETCO2 . End-tidal oxygen remained constant during DEF but showed a repeatable variation during FIC (fourth row). Again, the dotted line indicates the resting PETO2 before the challenge began. Inspired oxygen PIO2 and minute ventilation (V, volume breathed per minute) are shown in rows 5 and 6, respectively. signal change per unit (mm Hg) change in PETCO2 , that is BOLD reactivity to CO2. The significance of the model fit to each voxel time series was calculated, yielding statistical parametric maps of contrasts for PETCO2 , and Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
residual BOLD signal changes explained by PETO2 and changes in minute ventilation. Z statistic images were thresholded using clusters determined by Z > 2.3 and a (corrected) cluster significance threshold of P = 0.05.
Dynamic end-tidal forcing for FMRI RG Wise et al 7
Results Physiological Challenge
The administration of the FIC and DEF was tolerated well by all four volunteers. Good control of end-tidal CO2 and oxygen was provided by the DEF system despite the adaptations necessary for using the system in the MRI environment. Dynamic end-tidal forcing produced a consistent hypercapnic challenge while holding end-tidal oxygen constant. Mean gray matter BOLD signal and respiratory parameters are shown for one representative volunteer in Figure 2. Table 1 also summarizes, across the group of four volunteers, changes in gray matter BOLD signal and respiratory parameters. PETCO2 during the hypercapnic challenges of FIC was not significantly different from that during DEF. This confirms that the chosen PETCO2 targets were successfully met by DEF. During the normocapnic portions of each challenge, the PICO2 and PETCO2 were significantly higher for DEF than for the FIC since the technique necessitates a mildly elevated ‘baseline’ level of CO2 to hold PETCO2 constant by modifying PICO2 on a breath-by-breath basis (see also Figure 2, second row). Residual hyperventilation in the FIC increases the PETCO2 difference in this period. The raised baseline level of PETCO2 in DEF also induced a general increase in minute-ventilation (Table 1). In general, PETCO2 rose faster at the start of the hypercapnic challenge during DEF than FIC. PETCO2 shows a squarer waveform for DEF than for FIC. Dynamic end-tidal forcing aims to modify PICO2 over a short timescale to elicit a particular PETCO2 , whereas for FIC, PETCO2 is governed by PICO2 and minute ventilation. In seeking a step increase in PETCO2 followed by a plateau, DEF initially chooses a sharp rise in PICO2 (Figure 2, row 3). The undershoot of PETCO2 was evident immediately after the end of the FIC hypercapnic challenge (140 to 200 secs), also reflected in an undershoot of the gray matter BOLD signal (Figure 2, rows 1 and 2). The PETCO2 undershoot was caused by ‘over breathing’ at the end of the period of elevated PICO2 . Although PICO2 returns to normal in the FIC, minute-ventilation (V) remains elevated for approximately 60 secs, causing a temporary reduction in PaCO2 and hence PETCO2 . By comparison, during the return to normocapnia, DEF avoids the PETCO2 undershoot by more gradually reducing the PICO2 . DEF successfully maintains a constant PETO2 , (Figure 2, row 4) by decreasing PIO2 when minute ventilation rises in response to hypercapnia. Importantly, PICO2 remains lower than normal during the period of hyperventilation after the end of the hypercapnic challenge. This prevents the steep rise in PETO2 seen after 120 secs for the FIC, which only falls again once ventilation returns to normal. The maintenance of normoxia during DEF is also reflected in the lower values of PICO2 and PETO2 during normocapnia and hypercapnia compared with the FIC in which PETO2 rises (Figure 2 row 4
and Table 1). The sharp decrease in PIO2 for the FIC results from the fractional displacement of oxygen in the 5% CO2 (balance air) mixture (19.7% O2). Nevertheless, this was not adequate to maintain PETO2 constant during the FIC while the minute ventilation varies in response to changing PETCO2 : CO2 and O2 Related Changes in Blood Oxygenation Level-Dependent Signal
The fractional BOLD signal increase (DSBOLD(GM)) during hypercapnia, compared with baseline preceding the challenge, was significantly smaller during DEF than FIC (Table 1). This is likely to result largely from the smaller degree of change of PETCO2 in the normocapnia–hypercapnia cycles during DEF because of the slightly elevated baseline PETCO2 . Furthermore, the undershoot of PETCO2 after hypercapnia provides a greater amplitude of PETCO2 fluctuation over the cycle. Dynamic end-tidal forcing significantly reduced the temporal variation in PETO2 compared with FIC during the normocapnia–hypercapnia cycles (Table 1). Figure 3 illustrates the group mean (percentage) BOLD signal reactivity normalized to the induced changes in PETCO2 . The spatial distribution of the BOLD signal reactivity is similar between the FIC and DEF techniques. This similarity is to be expected as both techniques provide a potent vasodilatory challenge through the same mechanism. The BOLD signal reactivity appears greatest in areas of high blood vessel density, for example sagittal sinus, where it is beyond 0.5% per mm Hg CO2. It also appears higher in gray matter where cerebral blood volume is greater than in white matter (Rostrup et al, 2000). The comparison of the global gray matter BOLD signal reactivity from the linear regression indicates a slightly smaller reactivity when measured using DEF compared with FIC (Table 1), potentially arising from PETO2 changes during the FIC (discussed below). Figure 4A illustrates the mean gray matter BOLD signal for the same subject as shown in Figure 2 and the fitted contributions to this signal from PETCO2 ; PETO2 , and minute ventilation. In the linear regression, the PETCO2 provides a good fit to the BOLD signal across most of gray and white matter (Figure 4B). This is consistent with the global vasodilatory effect of CO2. The inclusion of PETO2 provides a significantly improved fit (the contribution being shown by the yellow line in Figure 4A) to the gray matter signal for FIC but not for DEF. This suggests a significant degree of PETO2 -related signal change in FIC that is not accounted for by the changes in PETCO2 . This is confirmed by the extensive regions of significant BOLD–PETO2 regression for FIC in Figure 4B. The degree of mean gray matter BOLD signal variation (temporal standard deviation stBOLD;O2 ) explicable by the PETO2 regressor is greater for FIC than DEF with a trend toward significance (one-tailed t-test, P = 0.07). The variation of PETO2 helps to explain the higher BOLD signal than predicted by PETCO2 during the later phases of the hypercapnic period and the early phase Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
Dynamic end-tidal forcing for FMRI RG Wise et al 8
of the recovery (Figure 4A). This increase in BOLD signal probably results from increased venous blood oxygenation when PETO2 is elevated. The additional BOLD signal changes accounted for by the inclusion of minute ventilation (and instantanteous breathed volumes, data not shown) are small and largely confined to image intensity boundaries. These changes in signal are likely to arise from breathing-related motion and B0 shifts of the image.
Discussion We have shown the feasibility of DEF as a means of delivering respiratory challenges for FMRI. By adapting a previously described DEF system (Robbins et al, 1982a, b) through minimizing expired gas sampling delays and using an MR-compatible turbine flow transducer, we achieved automated control of the partial pressure of end-tidal carbon dioxide (PETCO2 ) and oxygen (PETO2 ). In our example, a hypercapnic challenge aiming at a step change in PETCO2 , DEF successfully achieved target PETCO2 values with a sharp temporal profile. This allowed the BOLD signal responsiveness to CO2 to be mapped. The simple hypercapnic challenge was chosen as the first test of feasibility of DEF for FMRI because of the widespread use of hypercapnia in FMRI. The principal benefit of the DEF over the FIC is that the target PETCO2 and PETO2 and their temporal profiles are independent and are under the control of the experimenter. This improved control feeds through to the temporal profile of the BOLD response, eliminating the undershoot seen at the end of the FIC as the elevated ventilation drives the PETCO2 transiently below normal. We have shown the significance of controlling PETO2 in such a simple hypercapnia FMRI study. In addition to the intended CO2 related BOLD signal change, in the FIC there is a measurable BOLD signal change owing to additional unintended arterial oxygenation-related fluctuations. This effect was largely eliminated by the control of PETO2 by DEF. Respiratory Challenges for Functional Magnetic Resonance Imaging
The need for flexible control of end-tidal gas partial pressures in MRI has been shown by a series of studies Fixed Insired Challenge
%BOLD change/mmHg
0
0.5
of human cerebrovascular physiology. Modulation of carbon dioxide levels has been most common because of the potent vasodilatory effects of CO2. Some early studies measuring BOLD signal changes as a marker of cerebrovascular reactivity employed a simple breathhold to elevate PaCO2 (Kastrup et al, 1999; Li et al, 2000; Liu et al, 2002). Although a breath-hold is convenient, it is something of an ill-defined challenge. Partial pressure of arterial carbon dioxide rises with time, PaO2 falls with time, and the duration is strictly limited. The majority of studies of hypercapnia have used a fixed and elevated FICO2 (similar to the FIC experiment used here), typically by mixing 5% CO2 with the remaining balance as air (Pattinson et al, 2006). The resulting rise in PETCO2 depends on the volunteer’s ventilatory response to CO2. The FIC approach has been used to map cerebrovascular reactivity using PET and FMRI (Lythgoe et al, 1999; Rostrup et al, 2000; van der Zande et al, 2005) and to measure the relationship between CBF and cerebral blood volume using PET (Rostrup et al, 2005). Hypercapnia has also been used as a convenient method of increasing baseline CBF to investigate the latter’s effect on the measured stimulus-induced BOLD response. Such studies have produced mixed results with reports of decreased (Cohen et al, 2002), unchanged (Corfield et al, 2001), and increased (Posse et al, 2001) stimulus-induced BOLD responses. Hypercapnia, delivered by FIC, has been used as a means of calibrating the BOLD response by providing an increased CBF without a change in metabolic consumption of oxygen (Hoge et al, 1999b). The only report of a breathing system that does not use a fixed inspired fraction of CO2 is that of Vesely et al (2001). The system was used in a study to measure cerebrovascular reactivity. It employed a rebreathing reservoir and required volunteers to match their breathed volume to the supply of fresh gas. Although this achieved a rapid transition to hypercapnia, it did rely on volunteers monitoring the volume of the fresh gas reservoir, a task that could interfere with other tasks during an FMRI scan and suppress the natural ventilatory responses. It should be noted that for studies of hypocapnia, voluntary hyperventilation is required. A reduced PETCO2 cannot be achieved by decreasing FICO2 because this is normally close to zero in the basal state. Thus, DEF can be used in Dynamic End–Tidal Forcing
%BOLD change/mmHg
0
0.5
Figure 3 BOLD signal reactivity per unit change in end-tidal carbon dioxide. Group mean results are shown after registration to the standard space of the Montreal Neurological Institute. The central group-mean anatomical image identifies the axial slices of fractional BOLD signal change per mm Hg change in PETCO2 , shown on the left for the fixed-inspired challenge (FIC) and on the right for dynamic end-tidal forcing (DEF). Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
Dynamic end-tidal forcing for FMRI RG Wise et al 9 Dynamic End–Tidal Forcing
Fixed Inspired Challenge 2.5
2.5
2
2
∆ BOLD signal (%)
∆ BOLD signal (%)
A
1.5 1 0.5 0
0
50
100 Time (s)
150
1.5 1 0.5 0
0
200
50
100 Time (s)
150
200
Mean gray matter (GM) signal
PETO fitted component
PETCO2 fitted component
Minute ventilation fitted component
2
PETCO2, PETO2 and minute ventilation fitted to GM signal
B
Fixed Inspired Challenge
Dynamic End–Tidal Forcing End–Tidal CO2 PETCO
2
Z score 32
2.3
End–Tidal O2 PETO2 Z score 13
2.3 Minute ventilation Z score 2.3
7
Figure 4 BOLD signal variation explicable by variation of end-tidal carbon dioxide, end-tidal oxygen levels, and minute ventilation (one representative volunteer). Results from the fixed inspired challenge are shown on the left, and those from dynamic end-tidal forcing are shown on the right. Data are from the same volunteer as represented in Figure 2. (A) The fractional change in mean gray matter BOLD signal (black line) is modeled by the PETCO2 waveform (blue), the PETO2 waveform (yellow), and the minute ventilation waveform (green). The total fit is given by the red line. The inclusion of the PETO2 waveform into the linear regression results in an improved fit for FIC but not DEF. (B) Images show the Z scores indicating a significant linear regression (Z > 2.3, cluster corrected P < 0.05) of the PETCO2 ; PETO2 ; and minute ventilation with BOLD signal. Only BOLD signal changes for FIC and not for DEF correlated significantly with the PETO2 variation.
conjunction with hyperventilation to maintain a normal PETO2 and a steady, but below normal PETCO2 . In FMRI, the ability to alter arterial blood oxygenation has been exploited less but should now be harnessed in humans because of the potential it offers for quantitative examination of BOLD contrast. A previous study using a hypoxic and hyperoxic FIC showed that a reduction in arterial oxygen saturation decreases T2*, whereas an increase in arterial oxygen saturation increases T2* (Rostrup et al, 1995). A recent study of BOLD signal and CBF measurement by arterial spin labeling has showed a spatially heterogeneously reduced degree of task-related signal change in regions engaged in a motor task during mild hypoxia (FIO2 = 12%) (Tuunanen and Kauppinen, 2006). The visual stimulus-induced BOLD response appears to be reduced by hypoxic hypoxia probably as a result of arterial deoxygenation and increased cerebral perfu-
sion. Mild hypoxic hypoxia has also been used to show a spatially heterogeneous dissociation between BOLD FMRI and CBF in the human visual cortex using quantitative FMRI (Tuunanen et al, 2006). Somewhat more extreme respiratory challenges can be administered to animals to quantify the effects of changing blood gases on BOLD, CBF, and CMRO2. However, such studies highlight the interaction between changes in PaCO2 and PaO2 that would ideally be separated to provide a purer physiological challenge. Dynamic endtidal forcing can achieve this separation. Independent PETCO2 and PETO2 Control
To rigorously investigate cerebrovascular physiology using techniques other than MRI, such as TCD, the ability to independently control PETCO2 , and PETO2 by DEF has proved invaluable (Ide et al, 2003; Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
Dynamic end-tidal forcing for FMRI RG Wise et al 10
Poulin et al, 1996, 1998). With the spatial and temporal resolution afforded by FMRI techniques, DEF in the MRI scanner would allow detailed mapping of the dynamic blood flow response with rigorous temporal physiological control. This study illustrates the usefulness of DEF. In an ideal hypercapnic challenge PETO2 remains constant while PETCO2 is elevated. An increase in PaCO2 causes vasodilation and increased global CBF, without a change in cerebral oxygen usage. The resulting decrease in venous de-oxyhemoglobin within the image voxels increases T2* and elevates the BOLD signal during gradient–echo EPI. Hypercapnia also causes minute ventilation (breathed volume per minute) to rise appreciably as shown in Figure 2. The increased ventilation results in a rise of PaO2 and hence PETO2 (Figure 2, left column). During the delivery of CO2 this is somewhat mitigated by the slightly reduced FIO2 in the standard commercial mixture (from 20.7% to 19.7%). However, the persistently elevated minute ventilation after the return to normal FICO2 (B0%) and FIO2 (20.7%) causes another rise of PETO2 . In the example shown in Figure 2, PETO2 is elevated by approximately 15 mm Hg. The effect of this change in oxygen tension during hypercapnia has not been investigated previously in FMRI studies. This increase in PETO2 represents a small increase in the oxygen carried in arterial blood, both that dissolved in the plasma (because of the low solubility of oxygen) and the large majority that is carried by hemoglobin. An increase in PaO2 from around 100 to 115 mm Hg would be expected to increase arterial oxygen saturation by approximately 1% (Severinghaus 1979). On the assumption that the amount of oxygen extracted as the blood passes through the capillaries is unchanged between FIC and DEF, the increase in venous oxygenation would parallel the small increase in arterial oxygenation for the FIC. A 1% increase in venous oxygenation is expected to translate to an approximate 0.3% increase in BOLD signal when applying the BOLD signal model of Hoge et al (1999a). This is consistent with our measured oxygen-related BOLD signal increase in the FIC, which is largely removed during control of PETO2 by the DEF. Changing oxygenation levels represent a source of systematic error in conventional FMRI hypercapnia studies, where they are not taken into account. In addition to the larger range of PETCO2 presented in the FIC, the PETO2 -related BOLD signal changes are likely to contribute to the larger signal changes seen in the FIC compared with the DEF. The PETO2 -related BOLD changes may also contribute to the slightly larger apparent BOLD signal reactivity to CO2 measured during the FIC compared with the DEF. A further example of the usefulness of DEF for MRI investigation studies lies in examining the effects of hyperoxia on CBF, which has recently been performed using FMRI (Bulte et al, 2007). When breathing 100% oxygen, CBF falls; however, this is accompanied by a Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
fall in PaCO2 , and it is unclear whether this was wholly or partly responsible for the fall in CBF or whether there is an independent vasoconstrictive effect of elevated PaO2 . In an MRI study in which elevated inspired partial pressures of CO2 were administered in concert with elevated oxygen levels, the independent PaO2 vasoconstrictive effect was inferred from a regression analysis (Floyd et al, 2003). With the availability of DEF in the MRI environment to fix PETCO2 while elevating PETO2 , such a study becomes more directly and easily performed. Similarly, in studies of mild hypoxia, DEF would allow the effect of hypoxia-induced ventilation increases in lowering PETCO2 to be removed. Advantages and Disadvantages of DEF for Functional Magnetic Resonance Imaging
To fix PETCO2 during the ‘normocapnic’ condition, it is necessary to set the target value slightly higher (in this study, 1 mm Hg higher) than the mean PETCO2 recorded at rest, with the resulting anticipated effect of flattening out the normal resting fluctuations in PETCO2 (Harris et al, 2006; Van den Aardweg and Karemaker 2002; Wise et al, 2004). Although this keeps PETCO2 within a physiologically reasonable range, this consistent elevating effect on the PETCO2 during the baseline condition should be considered when designing and interpreting respiratory challenges. This method is adopted because reducing PETCO2 below resting values would require a voluntary increase in ventilation (hyperventilation), while a small increase in PETCO2 is easily achieved by slightly increasing the inspired partial pressure of CO2 without the volunteer noticing. This necessary rise in DEF target PETCO2 during normocapnia is apparent in Table 1 (40.9 mm Hg) given that the mean resting PETCO2 before hypercapnic challenges began was 39.6 mm Hg. This value is higher than the mean value during the normocapnic periods of the FIC (38.2 mm Hg). The decreased PETCO2 during normocapnic periods relative to before the FIC, results from the hypercapnia-induced increase in ventilation that persists for approximately 1 min after PICO2 returns to normal. In this study, the FIC always preceded the DEF challenge. This was because we chose to use the values of PETCO2 and PETO2 reached in FIC as the fixed targets for the DEF. At the higher PETCO2 levels, this provides a fairer comparison between the techniques as it limits the increase in ventilation in response to CO2. Choosing instead the PETCO2 target for DEF to maintain the dynamic range of PETCO2 seen for the FIC would have necessitated a mean target value of around 48 mm Hg because of the undershoot of PETCO2 seen in the FIC. This would generally shift the PETCO2 and BOLD signal values for DEF to a higher range. Our present approach of basing the DEF targets on the FIC provided a demonstration that DEF could achieve the requested targets. Using this method, we were confident that the DEF approach would offer a
Dynamic end-tidal forcing for FMRI RG Wise et al 11
tolerable hypercapnic challenge for volunteers within one scan session. One disadvantage of our approach is that a smaller dynamic range of PETCO2 and therefore BOLD signal change was investigated with the DEF than the FIC, potentially biasing comparison of BOLD signal change from DEF and FIC. This difference in the range of PETCO2 studied with DEF and FIC could contribute to the smaller apparent BOLD signal reactivity for DEF than FIC if the BOLD–PETCO2 relationship is nonlinear. The DEF technique does involve considerably more expense and setup time than a typical FIC. It is quite possible therefore that for some studies the FIC may remain preferable, for example, studies in which a tight-fitting face mask could discourage volunteers or those that do not benefit greatly from control of PETCO2 and PETO2 . Although we achieved adequate control of end-tidal gas partial pressures, the performance of the DEF system is likely to be limited by the gas sampling delay of 1.6 secs imposed by the need to site the gas analyzers several meters from the MRI magnet. Compared with earlier DEF systems developed for use in a standard (non-MRI) laboratory environment (Robbins et al, 1982a, b), the response capability of the MRcompatible system is likely to be reduced. In the experiments presented here, it was necessary to extend the gas sampling delays assumed in the gas control software to ensure that end-tidal values were correctly estimated. Because of the extended sampling delay, we also anticipate that for volunteers whose breathing rates increase greatly during hypercapnia, it may be necessary in the future to reduce the gas partial pressure feedback gain to avoid oscillations in end-tidal levels. Despite the complexities of implementing DEF in the MRI environment, its development provides new opportunities for quantitative FMRI investigations. It has the capacity to fix end-tidal gas partial pressures, not only within an individual subject’s scan session but also between sessions and subjects. The changes in PETCO2 and PETO2 demanded by the experimenter, since they are target driven, become largely independent of the individual subject’s ventilatory response to hypercapnia, hypoxia, and hyperoxia. The computer control of end-tidal gas partial pressures facilitates synchronization of the respiratory challenge with MRI data acquisition. Previously, many FIC approaches have relied on manual switching of inspired gas mixtures. Dynamic end-tidal forcing enables more complex respiratory challenges to be implemented in the MRI environment than previously attempted, such as multiple levels of PETCO2 within a single scan session and rapidly or slowly ramping end-tidal gas partial pressures. Figure 2 provides a hint that DEF may be useful for quantitative modeling and mapping of the temporal response parameters of CBF. The slope of the PETCO2 rise was steeper than its BOLD signal counterpart in DEF, in contrast to the FIC approach. This may reveal a limitation in the speed of response of the cerebral vasculature worthy of further investigation with rapid changes in arterial gas partial pressures (Poulin et al, 1996).
A potential new application for DEF in FMRI lies in controlling end-tidal gases, especially PETCO2, as a means of reducing PaCO2 -related fluctuations in CBF. It has recently been shown that the natural low-frequency fluctuations in PaCO2 arising from slow changes in breathing give rise to low-frequency noise (0 to 0.05 Hz) in the BOLD signal (Wise et al, 2004). Reducing the fluctuations in PETCO2 by DEF has also been shown to reduce the variability in CBF (Harris et al, 2006), which gives rise to these BOLD signal changes. Dynamic endtidal forcing therefore has the potential to reduce lowfrequency breathing-related physiological noise in FMRI. Although the confounds of using a face mask may outweigh the benefits of this approach for many FMRI studies, it is likely to be beneficial in studies where breathing patterns may correlate with the stimulation presented to the subject. Such studies may otherwise have CO2-related BOLD signal changes that are falsely identified as task-related neuronal activity. Dynamic end-tidal forcing may also have similar applications in stabilizing end-tidal gases during a drug challenge, especially when respiration is stimulated, as has previously been observed in an FMRI study of cocaine (Breiter et al, 1997). If the CBF– PETCO2 relationship is already characterized for a volunteer, DEF could also be harnessed to compensate by modifying PETCO2 for potential pharmacological modulation of CBF and the confound that this presents in interpreting stimulus-induced BOLD signal changes. In summary, we have showed the feasibility of DEF for FMRI for a hypercapnic challenge in a small cohort of volunteers. The control of end-tidal gases afforded to the experimenter can now be applied in quantitative studies of FMRI BOLD contrast as well as studies aimed at probing cerebrovascular physiology across the full range of respiratory challenges in human volunteers.
References Ainslie PN, Poulin MJ (2004) Ventilatory, cerebrovascular, and cardiovascular interactions in acute hypoxia: regulation by carbon dioxide. J Appl Physiol 97:149–59 Breiter HC, Gollub RL, Weisskoff RM, Kennedy DN, Makris N, Berke JD, Goodman JM, Kantor HL, Gastfriend DR, Riorden JP, Mathew RT, Rosen BR, Hyman SE (1997) Acute effects of cocaine on human brain activity and emotion. Neuron 19:591–611 Bulte DP, Chiarelli PA, Wise RG, Jezzard P (2007) Cerebral perfusion response to hyperoxia. J Cereb Blood Flow Metab 27:69–75 Cohen ER, Ugurbil K, Kim SG (2002) Effect of basal conditions on the magnitude and dynamics of the blood oxygenation level-dependent fMRI response. J Cereb Blood Flow Metab 22:1042–53 Corfield DR, Murphy K, Josephs O, Adams L, Turner R (2001) Does hypercapnia-induced cerebral vasodilation modulate the hemodynamic response to neural activation? Neuroimage 13:1207–11 Dahan A, Teppema LJ (2003) Influence of anaesthesia and analgesia on the control of breathing. Br J Anaesth 91:40–9 Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
Dynamic end-tidal forcing for FMRI RG Wise et al 12
Floyd TF, Clark JM, Gelfand R, Detre JA, Ratcliffe S, Guvakov D, Lambertsen CJ, Eckenhoff RG (2003) Independent cerebral vasoconstrictive effects of hyperoxia and accompanying arterial hypocapnia at 1 ATA. J Appl Physiol 95:2453–61 Gamboa A, Leon-Velarde F, Rivera-Ch M, Palacios JA, Pragnell TR, O’Connor DF, Robbins PA (2003) Selected contribution: acute and sustained ventilatory responses to hypoxia in high-altitude natives living at sea level. J Appl Physiol 94:1255–62; 1253–54 Harris AD, Ide K, Poulin MJ, Frayne R (2006) Control of end-tidal PCO2 reduces middle cerebral artery blood velocity variability: implicat. Neuroimage 29:1272–7 Hoge RD, Atkinson J, Gill B, Crelier GR, Marrett S, Pike GB (1999a) Investigation of BOLD signal dependence on cerebral blood flow and oxygen consumption: the deoxyhemoglobin dilution model. Magn Reson Med 42:849–63 Hoge RD, Atkinson J, Gill B, Crelier GR, Marrett S, Pike GB (1999b) Linear coupling between cerebral blood flow and oxygen consumption in activated human cortex. Proc Natl Acad Sci USA 96:9403–8 Howson MG, Khamnei S, McIntyre ME, O’Connor DF, Robbins PA (1987) A rapid computer controlled binary gas mixing system for studies in respiratory control. J Physiol (Lond) 403:103P Ide K, Eliasziw M, Poulin MJ (2003) The relationship between middle cerebral artery blood velocity and endtidal PCO2 in the hypocapnic-hypercapnic range in humans. J Appl Physiol 95:129–37 Jenkinson M, Smith SM (2001) A global optimisation method for robust affine registration of brain images. Med Image Anal 5:143–56 Jenkinson M, Bannister P, Brady M, Smith S (2002) Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage 17:825–41 Kastrup A, Li TQ, Glover GH, Moseley ME (1999) Cerebral blood flow-related signal changes during breath-holding. Am J Neuroradiol 20:1233–8 Kolbitsch C, Lorenz IH, Hormann C, Hinteregger M, Lockinger A, Moser PL, Kremser C, Schocke M, Felber S, Pfeiffer KP, Benzer A (2002) The influence of hyperoxia on regional cerebral blood flow (rCBF), regional cerebral blood volume (rCBV) and cerebral blood flow velocity in the middle cerebral artery (CBFVMCA) in human volunteers. Magn Reson Imaging 20:535–41 Li TQ, Kastrup A, Moseley ME, Glover GH (2000) Changes in baseline cerebral blood flow in humans do not influence regional cerebral blood flow response to photic stimulation. J Magn Reson Imaging 12:757–62 Liu HL, Huang JC, Wu CT, Hsu YY (2002) Detectability of blood oxygenation level-dependent signal changes during short breath hold duration. Magn Reson Imaging 20:643–8 Lythgoe DJ, Williams SC, Cullinane M, Markus HS (1999) Mapping of cerebrovascular reactivity using BOLD magnetic resonance imaging. Magn Reson Imaging 17:495–502 Maxwell DL, Cover D, Hughes JM (1985) Effect of respiratory apparatus on timing and depth of breathing in man. Respir Physiol 61:255–64 McKay LC, Evans KC, Frackowiak RS, Corfield DR (2003) Neural correlates of voluntary breathing in humans. J Appl Physiol 95:1170–8 Pattinson KT, Rogers R, Mayhew SD, Tracey I, Wise RG (2007) Pharmacological FMRI: measuring opioid effects on the BOLD response to hypercapnia. J Cereb Blood Flow Metab 27:414–23 Posse S, Kemna LJ, Elghahwagi B, Wiese S, Kiselev VG (2001) Effect of graded hypo- and hypercapnia on fMRI
Journal of Cerebral Blood Flow & Metabolism (2007), 1–12
contrast in visual cortex: quantification of T2* changes by multiecho EPI. Magn Reson Med 46:264–71 Poulin MJ, Liang PJ, Robbins PA (1996) Dynamics of the cerebral blood flow response to step changes in end-tidal PCO2 and PO2 in humans. J Appl Physiol 81:1084–95 Poulin MJ, Liang PJ, Robbins PA (1998) Fast and slow components of cerebral blood flow response to step decreases in end-tidal PCO2 . J Appl Physiol 85:388–97 Poulin MJ, Robbins PA (1998) Influence of cerebral blood flow on the ventilatory response to hypoxia in humans. Exp Physiol 83:95–106 Robbins PA, Conway J, Cunningham DA, Khamnei S, Paterson DJ (1990) A comparison of indirect methods for continuous estimation of arterial PCO2 in men. J Appl Physiol 68:1727–31 Robbins PA, Swanson GD, Howson MG (1982a) A prediction–correction scheme for forcing alveolar gases along certain time courses. J Appl Physiol 52:1353–7 Robbins PA, Swanson GD, Micco AJ, Schubert WP (1982b) A fast gas-mixing system for breath-to-breath respiratory control studies. J Appl Physiol 52:1358–62 Rostrup E, Knudsen GM, Law I, Holm S, Larsson HB, Paulson OB (2005) The relationship between cerebral blood flow and volume in humans. Neuroimage 24:1–11 Rostrup E, Larsson HB, Toft PB, Garde K, Henriksen O (1995) Signal changes in gradient echo images of human brain induced by hypo- and hyperoxia. NMR Biomed 8:41–7 Rostrup E, Law I, Blinkenberg M, Larsson HB, Born AP, Holm S, Paulson OB (2000) Regional differences in the CBF and BOLD responses to hypercapnia: a combined PET and fMRI study. Neuroimage 11:87–97 Severinghaus JW (1979) Simple, accurate equations for human blood O2 dissociation computations. J Appl Physiol 46:599–602 Swanson GD (1972) Dynamic forcing in the study of the human respiratory system. Stanford, CA: Stanford University Tuunanen PI, Kauppinen RA (2006) Effects of oxygen saturation on BOLD and arterial spin labelling perfusion fMRI signals studied in a motor activation task. Neuroimage 30:102–9 Tuunanen PI, Murray IJ, Parry NR, Kauppinen RA (2006) Heterogeneous oxygen extraction in the visual cortex during activation in mild hypoxic hypoxia revealed by quantitative functional magnetic resonance imaging. J Cereb Blood Flow Metab 26:263–73 Van den Aardweg JG, Karemaker JM (2002) Influence of chemoreflexes on respiratory variability in healthy subjects. Am J Respir Crit Care Med 165:1041–7 Van der Zande FH, Hofman PA, Backes WH (2005) Mapping hypercapnia-induced cerebrovascular reactivity using BOLD MRI. Neuroradiology 47:114–20 Vesely A, Sasano H, Volgyesi G, Somogyi R, Tesler J, Fedorko L, Grynspan J, Crawley A, Fisher JA, Mikulis D (2001) MRI mapping of cerebrovascular reactivity using square wave changes in. Magn Reson Med 45:1011–3 Wise RG, Ide K, Poulin MJ, Tracey I (2004) Resting fluctuations in arterial carbon dioxide induce significant low frequency variations in BOLD signal. Neuroimage 21:1652–64 Woolrich MW, Ripley BD, Brady M, Smith SM (2001) Temporal autocorrelation in univariate linear modeling of FMRI data. Neuroimage 14:1370–86 Zhang Y, Brady M, Smith S (2001) Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE Trans Med Imag 20:45–57
