Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78 & 2006 ISCBFM All rights reserved 0271-678X/06 $30.00 www.jcbfm.com
Regional temperature changes in the brain during somatosensory stimulation Hubert KF Tru¨bel1,2,5, Laura I Sacolick3 and Fahmeed Hyder1,3,4 1 Department of Diagnostic Radiology, Magnetic Resonance Research Center, Yale University, New Haven, Connecticut, USA; 2Department of Pediatrics, Magnetic Resonance Research Center, Yale University, New Haven, Connecticut, USA; 3Department of Biomedical Engineering, Magnetic Resonance Research Center, Yale University, New Haven, Connecticut, USA; 4Section of Bioimaging Sciences, Magnetic Resonance Research Center, Yale University, New Haven, Connecticut, USA; 5Department of Pediatrics, HELIOS Klinikum Wuppertal, University Witten/Herdecke, Wuppertal, Germany
Time-dependent variations in the brain temperature (Tt) are likely to be caused by fluctuations of cerebral blood flow (CBF) and cerebral metabolic rate of oxidative consumption (CMRO2 ), both of which are seemingly coupled to alterations in neuronal activity. We combined magnetic resonance, optical imaging, temperature sensing, and electrophysiologic methods in a-chloralose anesthetized rats to obtain multimodal measurements during forepaw stimulation. Localized changes in neuronal activity were colocalized with regional increases in Tt (by B0.2%), CBF (by B95%), and CMRO2 (by B73%). The time-to-peak for Tt (42711 secs) was significantly longer than those for CBF and CMRO2 (572 and 1874 secs, respectively) with a 2-min stimulation. Net heat in the region of interest (ROI) was modeled as being dependent on the sum of heats attributed to changes in CMRO2 (Qm) and CBF (Qf) as well as conductive heat loss from the ROI to neighboring regions (Qc) and to the environment (Qe). Although tissue cooling because of Qf and Qc can occur and are enhanced during activation, the net increase in Tt corresponded to a large rise in Qm, whereas effects of Qe can be ignored. The results show that Tt increases slowly (by B0.11C) during physiologic stimulation in a-chloralose anesthetized rats. Because the potential cooling effect of CBF depends on the temperature of blood entering the brain, Tt is mainly affected by CMRO2 during functional challenges. Implications of these findings for functional studies in awake humans and temperature regulation are discussed. Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78. doi:10.1038/sj.jcbfm.9600164; published online 15 June 2005 Keywords: energetics; fMRI; glucose; lactate; MRS; neuroimaging
Introduction Regulating body temperature is one of the oldest and best conserved physiologic functions that developed early in evolution, which subsequently led to homeothermic existence (Nadel, 2003). Specific areas in the brain (e.g., hypothalamus) contain the main temperature sensors (Gu¨ler et al, 2002; Nadel, 2003; Nagashima et al, 2000, Voets et al, 2004) and also serve as the processing unit for thermal regulation of the whole body (Benzinger, 1969; Boulant, 1998). In
Correspondence: F Hyder, Associate Professor, Department of Diagnostic Radiology, Yale University, N143 TAC, 300 Cedar Street, New Haven, CT 06510, USA. E-mail:
[email protected] FH thanks grants from the National Institutes of Health (DC03710; MH-67528) and the National Science Foundation (DBI0095173) for support and Arman Hyder for inspiration. Received 7 March 2005; revised 5 May 2005; accepted 6 May 2005; published online 15 June 2005
certain pathophysiologic states that affect the body (e.g., cardiac arrest with successful resuscitation) or the central nervous system (e.g., traumatic brain injury or stroke), temperature is a main prognostic marker (Ginsberg and Busto, 1998; Thompson et al, 2003a, b; Yager et al, 2004). The focus of therapeutic research for treating these conditions involves maintaining brain temperature: either preserving within the normal range (Thompson et al, 2003a) or cooling below the normal level (Gluckman et al, 2005). Because temperature changes have profound effects on cellular function and integrity, regulation of the brain temperature is an important function of the central nervous system (Baker, 1979). Alterations in cerebral metabolic rate of oxygen consumption ðCMRO2 Þ and cerebral blood flow (CBF), both of which are assumed to be correlated with variations in neuronal activity (Shulman et al, 2004), can cause changes in brain tissue temperature (Tt). The magnitudes of increases in CMRO2 and CBF within an activated region are not proportional (Hyder,
Understanding localized temperature dynamics in brain HKF Tru¨bel et al 69
2004a). In most cases, the CBF increase is slightly greater than actually needed to support the rise in CMRO2 ; (for a recent review, see, Shulman et al, 2001). This mismatch between CMRO2 and CBF leads to focal hyperemia, which is manifested as a rise in hemoglobin saturation on the venous end of the capillary bed (Hyder, 2004a, b) and is the basis of the blood oxygenation level-dependent (BOLD) effect (Ogawa et al, 1993). Because localized changes in CMRO2 and CBF have been detected with sensory stimulation in anesthetized rats (Hyder, 2004b), we hypothesize that localized changes in Tt can also be measured under similar conditions. However, the magnitude of change in Tt—which may be large, small, positive, or negative—is intricately linked to the magnitude of CBF increase versus CMRO2 increase. Although changes in CBF and CMRO2 are well correlated over a wide range of brain activity (Hyder, 2004a, b), no consensus exists about a precise link between changes in Tt, CBF, and CMRO2 during functional brain activation. The motivation, therefore, was to establish a quantitative understanding about regional temperature changes in the brain during stimulation. The goals of the current study were to obtain quantitative and dynamic multimodal measurements (Tt, CBF, CMRO2 , and neuronal activity) during sensory stimulation and then use these measurements to understand regional dynamics in temperature, which might be of interest for further insights into the BOLD effect. This study is the first to use multimodal neurophysiologic acquisitions (by combining magnetic resonance imaging (MRI), optical imaging, temperature sensing, and electrophysiologic methods in an anesthetized rat model) to relate quantitatively dynamic changes in temperature to alterations in metabolism and blood flow. A model was constructed that relied on transfer of different heat components that can transiently affect temperature in a region of interest (ROI). Our results show that a slow stimulation-induced rise in Tt within the ROI is mainly impacted by competing effects of cooling and warming because of changes in CBF and CMRO2 , respectively. In addition, the temperature of arterial blood (or the core) and resting values of oxidative metabolism are vital in regulating Tt under normal conditions.
Qf ðtÞ ¼ CBFðtÞrb Cb ðTa Tv ðtÞÞ Qc ðtÞ ¼ rt Ct V ðTo Ti ðtÞÞð1 expðt=tÞÞ Qe ¼ ðQm;o þ Qf;o Þ
ð1bÞ ð1cÞ ð1dÞ
The first three components reflect stimulationinduced changes in heat (Figure 1), whereas the last component is a fixed factor, which does not directly modulate Tt during brain activation. In equation (1) DHo and DHb are the enthalpies of oxidative phosphorylation and oxygen release from hemoglobin, respectively; rb and rt are the densities of blood and tissue, respectively; Cb and Ct are the specific heat capacities of blood and tissue, respectively; V is the spherical volume of the ROI; Ta and Tv(t) are the arterial and venous temperatures of blood entering and leaving the ROI, respectively, where the former can be assumed to be time-invariant; Ti(t) and To are the brain tissue temperatures inside and outside the ROI, respectively, where the former is equivalent to the measured dynamic temperature, Tt(t), and the latter can be assumed to be time-invariant being equal to the brain tissue temperature inside the ROI for baseline conditions; t is the time-constant for conductive heat loss from the ROI to the surrounding tissue, which is equal to rtVCt/(hA) where h and A are the convection heat transfer coefficient and surface area of the spherical ROI, respectively. In equation (1a) and (1b) Qm and Qf can be adjusted for the tissue mass of the ROI. In equation (1c) and (1d) Qc and Qe both describe conductive heat loss within brain tissue. Qc is the component that varies as a result of the temperature difference between the ROI and its surrounding, whereas Qe
Theory The net heat (Qnet) in the ROI can be modeled to depend on the sum of four components (Pennes, 1948): heats assigned to changes in CMRO2 (Qm) and CBF (Qf) as well as conductive heat loss from the ROI to neighboring regions (Qc) and dissipative heat loss from the ROI to the environment (Qe). Qm ðtÞ ¼ CMRO2 ðtÞðDHo DHb Þ
ð1aÞ
Figure 1 Illustration of the heat transfer model. The region of interest (ROI) and the surrounding tissue, in thermal contact with each other, have different temperatures (i.e., Ti and To, respectively) both of which can be measured. The arterial temperature (Ta) was directly measured from the blood temperature in the aortic arch (Tb) and the venous temperature (Tv) was assumed to be equal to the tissue temperature (Tt). Heat in the ROI can be lost to the surrounding tissue (Qc) or the environment (Qe). The dominant factors for modulating heat in the ROI are tissue metabolism (Qm) and blood flow (Qf). Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78
Understanding localized temperature dynamics in brain HKF Tru¨bel et al 70
results from the temperature difference between brain and ambient air. We chose to separate Qc and Qe because the former is a physiologic effect due to temperature difference between active and inactive tissue and varies over time, whereas the latter is a constant factor caused in part by the experimental procedure. Qc describes heat transfer between active and inactive tissue irrespective of whether the skull is intact or not, and Qe is likely to be insignificant in both cases. The scalar lumped parameter model of equation (1c), which includes convective exchange of the ROI with its surrounding (Shitzer and Kleiner, 1980), does not permit estimation of temperature gradients. The fixed value of Qe in equation (1d) is dependent on the temperature difference between the ambient air and the skull–tissue interface. A maximum estimate of this factor can be obtained from the heats assigned to resting values of CMRO2 (Qm,o) and CBF (Qf,o). Because the net heat for the ROI (Qnet) is related to the specific heat capacity of brain tissue (Ct) and the time derivative of Tt (dTt/dt), the temperature change (DTt) is given by Z 1 ð2Þ DTt ¼ Qnet ðtÞ dt rt Ct where the time integral in equation (2) is dependent on the time interval over which DTt is desired. The blood volume of cerebral microcirculation is small (Craigie, 1945; Boulpaep, 2003) and arterio-venous temperature differences are also intrinsically small under normal conditions (He et al, 2003). Therefore, equation (2) assumes that the heat content of blood would contribute negligibly to Tt, but this condition may need to be relaxed under pathophysiology (Hayashi, 2004). Furthermore DTt in equation (2) is related to changes in different heat components (equation (1)), where inherent impacts from Qm (i.e., CMRO2 ) and Qf (i.e., CBF) may originate from a variety of mechanisms, for example, induced by anesthesia (Shulman et al, 1999) or break in autoregulation (Hernandez et al, 1978).
Materials and methods Animal Preparation All experiments were conducted on Sprague–Dawley rats (male; 200 to 360 g; Charles River, Wilmington, MA, USA). The rats were anesthetized with B1% halothane and 70%/30% N2O/O2 during all other surgical procedures. To prevent heat loss to the environment from the body as well as N2O-induced hypothermia (Quock et al, 1987), the body warmth was maintained constant with a temperaturecontrolled water blanket. A femoral artery was cannulated to withdraw blood samples for blood gas analysis and to monitor blood pressure (BP). A femoral vein was cannulated for infusion of iron oxide contrast agent (AMI-227, Advance Magnetics, Cambridge, MA, USA) for measuring relative changes in cerebral blood volume (CBV) with MRI (Kida et al, 2000). Intraperitoneal catheters were placed in Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78
the abdomen for administration of drugs. Two subcutaneously placed copper needles were inserted into the contralateral forepaw and all snout whiskers were shaved to avoid contaminating somatosensory signals. Stimuli were provided (S88, Grass Instruments, Quincy, MA, USA) in conjunction with an isolation unit (A360, WPI, Sarasota, FL, USA) using electrical impulses of 0.3 ms duration and 2 mA amplitude with 3 Hz frequency. The scalp was retracted and removed to expose the bregma. After all procedures (see below), halothane was discontinued and anesthesia was maintained with intraperitoneal injections of a-chloralose (B40 mg/kg h) and the rats were immobilized with intraperitoneal injections of Dtubocurarine chloride (B0.3 mg/kg h). Ventilation parameters were adjusted to maintain normal physiology.
Preparation for multimodal MRI experiments (n ¼ 6): The skull was cleaned of all tissues and a layer of Saran Wrap was placed over the wound for subsequent placement of the radio-frequency surface coil above the bregma. The rat’s head was secured with a stereotaxic-like holder and tightly fixed by foam cushions around the head (Kida et al, 2001) to minimize motion artifacts and to provide insulation. The rat was placed prone in a cradle, which held the radio-frequency coil and then inserted into the magnet for a functional MRI (fMRI) scan (see below). The ambient (magnet bore) temperature was kept constant. Preparation for dynamic recordings with a multisensor probe (n ¼ 41): The rat was placed on a vibrationfree stereotaxic frame (David Kopf Instruments, Tujunga, CA, USA) inside a Faraday cage. The skull over the somatosensory cortex (4.4 mm lateral, 1.0 mm anterior, with respect to the bregma), guided by fMRI experiments, was thinned to expose the dura (Smith et al, 2002) for B1 1 mm2 square area. After placement of the multisensor probe to layer 4 (see below) cotton swabs were placed above the rat’s head to provide insulation to minimize heat loss through the skull. The room temperature was kept constant. Dynamic multimodal measurements were made during somatosensory stimulation (n ¼ 29). In addition, blood temperature (n ¼ 6) and rate of temperature decay (n ¼ 6) were measured in two separate groups of rats. The blood temperature (Tb) was measured from the aortic arch. An incision was made and the carotid artery was exposed to insert the probe near the branches of the subclavian arteries. The incision was sutured to stabilize the probe. The time-constant (t) was measured using the rate of brain temperature decay extracted from the point of isoelectricity (i.e., euthanasia induced by KCl injection to the heart) to the point when the brain’s temperature equilibrated with the ambient temperature.
Measurements of CBF and CMRO2 with Multimodal MRI All MRI data were collected on a 7.0 T Bruker (Billerica, MA, USA) horizontal-bore spectrometer with a homo-
Understanding localized temperature dynamics in brain HKF Tru¨bel et al 71
geneous transmit and local receive 1H radio-frequency coil. Blood oxygenation level-dependent weighted images were acquired with spin echo fMRI (Hyder et al, 2000a). Quantitative CBF (absolute and fractional changes) data were measured by arterial spin labeling technique (Hyder et al, 2000b). Quantitative changes in CBV were measured by AMI-227 (Advanced Magnetics Inc., Cambridge, MA, USA), which is a plasma-borne MRI contrast agent (Kida et al, 2000). Stimulation-induced changes in BOLD, CBF, and CBV were obtained with sequentially sampled echo-planar imaging (EPI) data (Hyder et al, 1995): 32 32; in-plane resolution of 430 430 mm2; slice thickness of 1,000 mm; repetition delay of 8 secs per slice; and (spin echo) echo time of 40 ms. The BOLD and CBF data were acquired in an interleaved manner, where the CBF data were quantified with slice-selective and non slice-selective inversionrecovery weighted EPI data using three inversion recovery times ranging from 600 to 1,900 ms (Hyder et al, 2000b). All other details of BOLD and CBF imaging are described elsewhere (Hyder et al, 2000a). After completing the BOLD and CBF measurements, the MRI contrast agent (1 mL, 2.2 mg/kg) was infused and another set of BOLD data were acquired, which were used in conjunction with prior BOLD measurements to estimate quantitative changes in CBV (Kida et al, 2000). Changes in CMRO2 were calculated with calibrated fMRI (Hyder et al, 2001) using multimodal MRI measurements of changes in BOLD signal (DS/S), blood flow (DCBF/CBF), and blood volume (DCBV/CBV) from the relationship originally described by Ogawa et al (1993): DCMRO2 DCBF 1 DS DCBV DCBF þ 1þ ð3Þ ¼ CBF a S CBV CBF CMRO2 In equation (3) the value of a(0.470.1) was based on the validation of calibrated fMRI at 7.0 T by comparison of predicted and measured values of DCMRO2 =CMRO2 (Hyder, 2004a) by 13C magnetic resonance spectroscopy (MRS). We assumed that the different experimental runs for BOLD, CBF, and CBV data produced reproducible changes in neuronal activity (per subject).
Dynamic Recordings with a Multisensor Probe To measure electrical activity, blood flow, tissue oxygen partial pressure, and temperature simultaneously, a multisensor probe was custom-designed, which can only be used in an open-skull setting. It combined a high impedance (2 MO) tungsten microelectrode (FHC, Bowdoinham, ME, USA) that was used to measure extracellular activities of a small neuronal ensemble (Hetherington and Swindale, 1999) with a triple-sensor device (250 mm diameter; Oxylite, Oxford Optronix, UK). The tungsten electrode was glued to the side of a needle shaft (30G; Terumo, Tokyo, Japan), which contained the Oxylite triple-sensor device. The triple-sensor device was thus isolated from the electrode. The triple-sensor probe included a copper-constantan thermocouple probe (Miyazawa and Hossmann, 1992), a laser-Doppler probe (830 nm) sensitive to red blood cell flow (Wadhwani and
Rapoport, 1988), and a phosphorescence luminescence probe (485/600 nm) sensitive to tissue oxygenation (Vinogradov and Wilson, 1997). Separate grounds were used for the thermocouple wire and electrodes and the exposed probe tips were physically isolated. The multisensor probe was placed in a stereotaxic holder (Plastic One, Roanoke, VA, USA) and inserted to layer 4. All simultaneously acquired signals were amplified and filtered before digitization (m-1401, CED, Cambridge, UK). Specific acquisition details of the extracellular electrical signals are described elsewhere (Hyder, 2004b). In brief, the raw extracellular electrical data were separated into slow (field potentials) and fast (action potentials) oscillations. The latter signal was used to obtain spiking frequency (n) of a neuronal ensemble by sorting for spike shapes as previously described (Smith et al, 2002). The electrical activity was used to guide placement of the other probes, that is, data acquisition for other sensors was not started if no stimulus-induced changes in electrical activity were detected. By this approach, we avoided including multisensor measurements from nonactive regions. The dynamic changes in temperature, perfusion, and oxygenation were continuously measured with the triple-sensor device, which had a sensitive volume of approximately 0.05 mL around the tip, in an undersampled manner (0.5–2 Hz) for high signal-to-noise ratio (SNR). Since the probe tip behaves as a point source, the sensitive volume of the device was approximated to be spherical. Thus, there was a small volume difference between the sphere sampled by the multisensor probe and the MRI voxel. Because the activated volume spans across several cortical layers (Ueki et al, 1988) and there may be slight disparities between the centers-of-mass of the sphere and the voxel, the partial volume effect will be partly homogenized and therefore cannot be readily unscrambled. This type of partial volume effect differs from the overlap between multiple functional domains (Erinjeri and Woolsey, 2002) where the heterogeneous mixing of activated columns may be separated based on some prior anatomical information. The laser-Doppler probe was used as a relative technique where the signal fluctuations provided a dynamic measure of CBF changes (410%). The luminescence probe provided quantitative values of oxygenation (70.1 mm Hg). We assumed that the dynamics of tissue oxygen partial pressure represented the slowest dynamic measure of CMRO2 changes (Ances et al, 2000). The thermocouple wire provided absolute temperature values (70.011C). Results from different animals were averaged (mean7standard deviation) and time-to-peak was defined as 90% of maximum (or minimum). The fractional change in each measured parameter was calculated as the difference between the averages of the 30 secs period before stimulation and the 30 secs period after the signal reached its peak during stimulation.
Modeling Dynamic Changes in Tt In equation (1a), the values of DHo and DHb were þ 470 kJ/ mol (Siesjo¨, 1978; Yablonskiy et al, 2000) and 28 kJ/mol Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78
Understanding localized temperature dynamics in brain HKF Tru¨bel et al 72
(Ackers et al, 1992; Mills et al, 1976), respectively. Enthalpies are not appreciably impacted by small tissue temperature variations (Woods and Lawrence, 1997), either spatially or temporally. Thus, the dominant parameter affecting Qm was CMRO2 . In equation (1b), the values of rb and Cb were 1.06 g/mL and 3.894 J/g K, respectively (Collins et al, 2004; Holdcroft, 1980; Incropera and DeWitt, 2002). The arterial blood temperature (Ta) can be assumed to be equal to the temperature of blood entering the brain (Tb). A measure of Tb can be used to reflect a time-independent measure of Ta with the assumption that no heat is lost within the main arterial network delivering blood to the brain. Similarly the temperature of venous blood (Tv) can be assumed to be equal to the temperature of brain tissue (Tt). Timedependent variations in Tt may be used to reflect dynamic changes in Tv with the assumption that the brain tissue compartment is in rapid thermal exchange with venous blood. The assumption that blood flowing out of the ROI leaves at the same temperature as the tissue is suitable because of the high water diffusivity coefficient and extensive fluid exchange between blood and tissue at the capillary level. Therefore, the term Ta–Tv(t) in equation (1b) is actually equivalent to Tb–Tt(t), which can be measured. Tb was directly measured from the aortic arch (n ¼ 6) for the current experimental conditions (see above) and generally its value was within 0.021C70.011C of the measured tissue temperature for the baseline condition (Tt,o). In equation (1c), the values of rt and Ct were 1.04 g/mL and 3.664 J/g K, respectively (Collins et al, 2004; Swan, 1974; Woods and Lawrence, 1997), whereas V reflected a sphere with 0.5 mm diameter. For a small ROI with randomly oriented capillaries (diameter o5 mm), an embedded assumption is that temperature within the ROI is homogenous. Because the ROI was approximately the same volume as the sensitive volume of the multisensor probe, the device itself would have no impact on the temperature value and a comparatively simple ‘lumped capacitance method’ can be used to determine time variations of temperature. Thus, time-dependent temperature variations inside the ROI (Ti) are actually the same as the dynamic changes measured in this experiment (Tt). Furthermore, the measured brain tissue temperature under the baseline condition (Tt,o) is equivalent to the brain tissue temperature outside the ROI (To) using the assumption of negligible heterogeneity across adjacent cortical layers, which span a few hundred mm. Therefore, the term To–Ti(t) in equation (1c) is actually equivalent to Tt,o–Tt(t), which can be measured. Because the tissue temperature was altered in the ROI during stimulation, there was heat loss from the ROI to the surrounding tissue using the assumption that the regional temperature variations under baseline conditions are small in comparison to the stimulation-induced changes. The values of A and h were needed to calculate t, where A was the surface area of the sphere and h was 10 J/m2 s K in the calculations (van Leeuwen et al, 2000; Liu, 2001; Shitzer and Kleiner, 1980). Varying h by an order of magnitude had little impact on Qc. The value of t can also be measured if the rate of heat loss is measured when the Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78
rat has been euthanized. The value of t was estimated by measuring the rate of temperature decay from the point of isoelectricity to the point when the brain’s temperature equilibrated with ambient temperature (n ¼ 6). Both calculated and measured values of t were longer than the baseline and stimulation periods. In equation (1d), the heat loss from the brain through the hole or other parts of the skull to the environment (Qe) was assumed to be small and unaltered between stimulation and baseline periods, because the skull opening was closed and the probe itself was thermally isolated. More importantly this heat loss is constant over each experimental run because there is a large temperature gradient between the ambient air and the skull–tissue interface. Thus, the temperature fluctuations inside the brain will not greatly impact the rate of heat loss to the environment over time. Thus Qe was fixed at –(Qm,o þ Qf,o) for all conditions. Baseline values of blood flow (i.e., CBFo ¼ 0.55 70.14 mL/g min) and oxidative metabolism (i.e., CMRO2 ;o ¼ 1:58 0:46 mmol/g/min) were used for modeling, where the former was measured (by MRI) in this study and the latter was based on prior measurements (by MRS) in rat brain (Hyder, 2004b). A baseline value of 0.021 C70.011C was measured for the arterio-venous temperature difference (i.e., Ta–Tv,oTb–Tt,o). These values gave a calculated Qe value of 0.01470.004 J/sec. In addition, Qe was measured (0.01570.007 J/sec) from the euthanasia experiment (see above) because Qm ¼ Qf ¼ 0 under these conditions. The inputs to the model (equation (1); Figure 1) were measured values of CBF(t), CMRO2 ðtÞ, Tb, and Tt,o. To improve the agreement between measured and predicted, values of DTt in equation (2) for a given experimental run were iterated until errors, as determined by least-squares fitting, were minimized (Simulink, Mathworks, Natick, MA, USA).
Results The stimulation-induced increases in spiking frequency of a small neuronal ensemble (n) in the ROI of the contralateral somatosensory cortex were colocalized with increases in BOLD, CBV, CBF, CMRO2 , and Tt (Table 1). This multiparametric approach excluded a closed-skull technique, for example, as used by Brugge et al (1995), to selectively measure brain temperature. To account for convective heat loss through the small opening in the skull, we introduced the term Qe (see Theory), which was also measured (see Materials and methods). In the current study, the stimulus duration was long (2 mins) so that oscillations in the range of 0.1 to 1 Hz for CBF (or CMRO2 ), which can be induced by deep anesthesia or low blood pressure (Jones et al, 1995), would not influence the Tt measurements. Because systemic physiology was controlled and maintained within normal limits (pH E7.32 to 7.38; pCO2 E34 to 38 mm Hg; pO2 4100 mm Hg; BP E87 to 114 mm Hg) throughout all measurements, saturation and hematocrit of arterial
Understanding localized temperature dynamics in brain HKF Tru¨bel et al 73
blood were normal. These procedures allowed reproducibility of stimulus-induced responses per subject without significant changes in BP. The magnitude of changes in CBF and CMRO2 were measured by multimodal MRI, whereas the dynamics were measured by laser-Doppler flowmetry and fluorescence quenching probes, which measure tissue perfusion and oxygenation, respectively. These changes in CBF and CMRO2 are in agreement with prior results of the same model (Hyder et al, 2002). Figure 2 shows that the increases in n, CBF, CMRO2 , and Tt were 81%735% (Po0.06), 95%723% (Po0.01), 73%732% (Po0.02), 0.2% 70.1% (Po0.09), respectively. The time-to-peak for the change in n, CBF, CMRO2 , and Tt were 1176, 572, 1874, and 42711 secs, respectively. Although the time courses of n, CBF, CMRO2 , and Tt all declined after stimulation offset (Figure 2), the
changes in Tt required 41 minute to reach prestimulation baseline levels (Figure 2D). Although the average increase in Tt was small (Figure 3A), the B0.11C rise in the ROI was statistically significant, with maximum and minimum values of þ 0.251C and 0.021C (Figure 3B). The fits of the model to the experimental results are shown in Figure 4. The most dominant components contributing to changes in Qnet were Qm and Qf (Figure 4A), where the tissue warming effect from DCMRO2 (equation (1a)) was higher in magnitude (approximately by 2) than the cooling effect from DCBF (equation (1b)). The time-to-peak for the changes in Qm and Qf were offset by B1 min where the tissue cooling effect (by DCBF) was significantly delayed. The Qm time course showed a rapid decline after stimulation offset (B20 secs), whereas Qf required longer time (B45 secs) to reach prestimula-
Table 1 Summary of data acquired with multi-modal MRI and multi-sensor probe Multi-modal MRI
Multi-sensor probe
DS%
DCBF%
DCBV%
DCMRO2 %
Dn%
DTt%
3.272.4 Po0.05
95723 Po0.01
5.173.3 Po0.06
73732 Po0.02
81735 Po0.06
0.270.1 Po0.09
The table shows stimulation-induced changes in BOLD, CBF, CBV, CMRO2 ; n, and Tt, respectively. The changes in CMRO2 were determined from Equation (3). The MRI data were averaged from an ROI of 1 1 voxel in the contralateral forepaw region which were superimposed over the ROI for the recordings with the multi-sensor probe. The measured baseline values of CBF and n were 0.5470.14 mL g1 min1 and 6.672.4 Hz, respectively. The baseline value of CMRO2 (1.6670.36 mmol g1 min1) was based on prior CBF CMRO2 measurements (Hyder et al., 2001). Data were averaged across all rats (mean7standard deviation; n ¼ 6 for BOLD, CBF, CBV, and CMRO2 ; n ¼ 29 for n and Tt). Student’s t-test with two samples assuming unequal variances (one-tail) was used for comparing mean values.
Figure 2 Multimodal neurophysiologic data. In a-chloralose anesthetized rats during forepaw stimulation changes in (A) spiking frequency of a neuronal ensemble (n) were colocalized with alterations in (B) blood flow (CBF), (C) oxidative metabolism ðCMRO2 Þ; and (D) brain tissue temperature (Tt). The changes in n (A) were measured by electrophysiologic methods. The changes in CBF (B) were measured by MRI and laser-Doppler for the magnitude and dynamics, respectively. The changes in CMRO2 (C) were measured by MRI and phosphorescence luminescence for the magnitude and dynamics, respectively. The changes in Tt (D) were measured by the thermocouple wire. While Tt increased by only B0.2% in the ROI, the other parameters increased considerably more (DCBF%E95%; DCMRO2 % 73%; and Dn%E81%). Data represent mean7standard deviation of all experiments.
Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78
Understanding localized temperature dynamics in brain HKF Tru¨bel et al 74
Figure 3 Stimulation-induced changes of brain tissue temperature (DTt) from all experiments (n ¼ 29). (A) Histogram of DTt values shows that approximately 1/3 of the measurements recorded increments of B0.081C. (B) Distribution of DTt values shows that the spread ranged from 0.021C to þ 0.251C, where 7 measurements (gray) showed statistically insignificant changes in Tt.
tion baseline levels. The least dominant component contributing to changes in Qnet within the ROI was Qc (Figure 4B). Although the time course of conductive heat loss from the ROI (equation (1c)) was similar to that of tissue cooling (by DCBF), the magnitude of change in Qc was much smaller (approximately by 25). The Qc time course showed a slow recovery after stimulation offset (445 secs). The time-to-peak for the change in Qnet was approximately similar to that for Qm (B20 secs), and a slow decline followed after reaching the peak indicating tissue cooling from contributions of Qf and Qc. The Qnet time course showed a drop below the prestimulation baseline levels after stimulation offset because of cooling from the combined effects of Qf and Qc, both requiring B1 minute after stimulation offset to plateau. The Qnet time course was then used (equation (2)) to model the dynamics of DTt (Figure 4D), where the model predictions (black) were in close agreement with measured values (gray). To understand some intricacies of the model, some simulation results are shown in Table 2. Effects of CMRO2 ;o =CBFo and Tb–Tt,o provided some new insights about temperature regulation. The effect of Tb–Tt,o on the magnitude of DTt suggests that the temperature of arterial blood (or the core) impacted the cooling efficiency of the brain. The CMRO2 ;o =CBFo ratio had drastically different effects both on the magnitude and dynamics of DTt. It took longer for the tissue to warm (or cool) as the metabolism rate increased and the magnitude of DTt was higher when the resting metabolism rate was high. While the importance of CBFo and DCMRO2 %=DCBF% ratio are obvious, these simulation results reveal the importance of resting metabolic rate and the core temperature for regulating brain temperature.
Figure 4 Fits of the model (equations (1) and (2); Figure 1) to experimental results. (A) The dominant components contributing to stimulation-induced changes in Qnet within the ROI are heats assigned to changes in CMRO2 (Qm) and CBF (Qf). (B) A less prevailing factor for changes in Qnet is conductive heat loss from the ROI to its surrounding tissue (Qc). (C) The combined effects of Qm(t), Qf(t), and Qc(t) for dynamics of Qnet within the ROI. (D) Comparison of predicted (black) and measured (gray) values of DTt(t). Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78
Understanding localized temperature dynamics in brain HKF Tru¨bel et al 75
Table 2 Sensitivity of the heat transfer model to changes in parameters Parameter DCMRO2 %=DCBF% Tb–Tt,o CBF CMRO2 ;o =CBF o
Value
DTt
Time-to-peak (secs)
0.8 0.7 0.6 +0.021C 01C –0.021C 3.0 mL g1 min1 0.9 ml g1 min1 0.5 ml g1 min1 5.0 mmol mL1 3.0 mmol ml1 1.0 mmol mL1
0.131C 0.111C 0.091C 0.121C 0.111C 0.101C 0.111C 0.111C 0.111C 0.181C 0.111C 0.031C
114 114 114 118 114 112 34 68 114 170 114 44
The theoretical behavior of the heat transfer model is illustrated for a ten-minute stimulation where typical values for parameters (CBFo ¼ 0.9 mL/g/min; CMRO2 ;o =CBFo ¼ 3:0 mmol/mL; Tb–Tt,o ¼ 01C; DCMRO2 %=DCBF% ¼ 0:7) and constants (DHo ¼ +470 kJ/mol; DHb ¼ –28 kJ/mol; rb ¼ 1.06 g/mL; rt ¼ 1.04 g/mL; Cb ¼ 3.894 J/g/K; Ct ¼ 3.664 J/g/K; V ¼ 0.065 mL; tb0 secs; Qe ¼ 0.014 J/s) were assumed. The simulations demonstrate the model behavior when different parameters are changed within a realistic physiologic range of values. Simulations were obtained by rearranging Equation (1) and solving for DTt. The dynamics of changes in CMRO2 ðtÞ and CBF(t) were assessed from prior (or current) experimental results and the values for time-to-peak (i.e., 90% of maximum) were 20 and 5 secs, respectively, with single exponential curves.
Discussion In the current study, the temperature measurements were supported by other multimodal measurements with a specific focus towards understanding their interactions while modulating temperature in the brain (Figure 1). Stimulation-induced changes in CBF (1.070.2) and CMRO2 (0.770.3) or the ratio of DCMRO2 %DCBF% (0.770.3) were much larger than the increase in Tt and reached peak values much faster than Tt (Figures 2–4; Tables 1 and 2). Because temperature varies as a consequence of the time integral of heat transfer to or from the ROI (equation (2)), Tt in an active region increases far more slowly than either CMRO2 or CBF and remains elevated even after CMRO2 or CBF returns to pre-stimulation baseline levels. Quantitative changes in CMRO2 and CBF were measured by multimodal MRI (Hyder, 2004a, b) in a neuroimaging format, where magnitudes of each parameter were analyzed for voxels located in the middle cortical layers, including layer 4. A maximum 10% error was estimated for the range of CBF values in this study (Hyder et al, 2000b). The changes in CMRO2 with calibrated fMRI, as shown by equation (3), which is the original BOLD model (Ogawa et al, 1993), had been tested to an accuracy of about 80% (Kida et al, 2000). Dynamic changes in CMRO2 and CBF were assumed from the optical measurements: a phosphorescence luminescence probe that senses oxygenation by quenching of a dye (Vinogradov and Wilson, 1997) and a laser-Doppler probe, which detects perfusion from moving red blood cells (Wadhwani and Rapoport, 1988), respectively. Both of these methods have relatively high SNR for relative changes in tissue oxygenation and perfusion (Nwaigwe et al, 2000).
The small stimulation-induced changes in brain temperature (B0.11C) were reproducibly measured because the copper–constantan thermocouple wire (Thulin, 1969) has good accuracy (70.011C) in the physiologic range (Boudry, 1976). The current temperature measurements from the brain are in accord with previous functional studies in the human (Gorbach et al, 2003), cat (McElligott and Melzack, 1967; Melzack and Casey, 1967; Serota and Gerard, 1938), and rat (LaManna et al, 1989; Shevelev and Tsicalov, 1997). In contrast temperature changes during seizure activity (Tru¨bel et al, 2004) or hypoxic challenge (LaManna et al, 1989) are much larger. Furthermore, even larger changes in the brain temperature have been reported during alterations in behavior, for example, arousal from sleep (Baker and Hayward, 1967; Baker et al, 1973; Baker, 1979), after feeding (Abrams et al, 1965; Delgado and Hanai, 1966), or provoked by sex or drug (Kiyatkin and Brown, 2003; Kiyatkin and Mitchum, 2003). The current results are, however, not in agreement with a human study, which showed that Tt decreases by B0.21C with visual stimulation (Yablonskiy et al, 2000). Changes in CBF and CMRO2 were not measured in that study and a very low DCMRO2 %=DCBF% ratio was assumed and a water-based 1H MRS method was used to infer temperature changes. A recent study (Collins et al, 2004) simulating local temperature changes in the human brain, where particular attention was given to test a wide range of parameter values (e.g., CBFo, CMRO2 ;o ; Tb, DCMRO2 %=DCBF%), suggest that stimulationinduced changes in Tt for the human brain can range from 0.031C to þ 0.101C for different activated foci. Simulations with the current model suggest that this scenario could be possible in the Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78
Understanding localized temperature dynamics in brain HKF Tru¨bel et al 76
rat if the core temperature was much cooler than the brain temperature. Future experiments can be conducted with larger arterio-venous temperature differences in rat to test this likelihood, which can be achieved by either whole body cooling (Bernard et al, 2002; Flores-Maldonado et al, 2001) or selective brain cooling (Thoresen et al, 2001; Tru¨bel et al, 2004). However, these studies would reveal temperature regulation for the brain under pathophysiology. In this study, the changes in CMRO2 measured (Figure 2C) and modeled (Figure 4A) pertain to glucose oxidation (CMRglc(ox)) because fMRI was calibrated (equation (3)) with 13C MRS measurements of oxidative metabolism of glucose (Kida et al, 2000; Hyder et al, 2001). However, total glucose consumption (CMRglc) includes an additional nonoxidative component (CMRglc(non-ox)) ð4aÞ CMRglc ¼ CMRglcðoxÞ þ CMRglcðnon-oxÞ and the oxygen-to-glucose index (OGI) CMRO2 OGI ¼ CMRglc
ð4bÞ
is a convenient way of determining the degree of oxidative versus nonoxidative glucose breakdown. Because it is well known that OGI ranges from 4.0 (e.g., seizure stimulation) to 5.5 (e.g., awake rest), the measurements of CMRglc and CMRO2 by different methods (Shulman et al, 2001) indicate that some extra glucose is metabolized nonoxidatively. Because the stoichiometry between glucose oxidation (CMRglc(ox)) and oxygen metabolism (CMRO2 ) is 1:6, it can be shown that for the OGI range measured in vivo the stoichiometry between nonoxidative glucose consumption (CMRglc(non-ox)) and oxygen consumption (CMRO2 ) is about 0.091:6 to 0.5:6. In other words, nonoxidative glucose consumption probably contributes as little as 1.5% (i.e., awake rest) or as much as 8.3% (i.e., seizure stimulation) to total glucose consumption. Because ATP yield from nonoxidative glucose consumption is approxiamately 20 times less than that from oxidative glucose consumption (Siesjo¨, 1978), the metabolic heat component contributing to temperature changes is almost exclusively from glucose oxidation for the OGI range measured in vivo. The dynamic measurements of blood flow, oxidative metabolism, and temperature enabled a heat transfer model (Figure 1) to be tested using the bioheat formalism (equations (1) and (2)), originally described by Pennes (1948). Some of the assumptions (Theory; Materials and methods) may be relaxed if temperature was quantitatively mapped throughout the brain (Gorbach et al, 2003; Shevelev et al, 1993). However, being able to measure the multimodal parameters associated with temperature changes in the same regions is important for future studies. A variety of MRS methods can noninvasively map temperature in different regions of the brain (Germain et al, 2001; Li et al, 2001; Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78
de Poorter et al, 1995). Because temperature sensitivity of these methods are low and calibrations are problematic (e.g., dependence on tissue perfusion and tissue types), an alternative 1H MRS approach has been proposed specifically for the brain (Tru¨bel et al, 2003), which is far more sensitive because the method relies on large chemical shift changes of a thermometric sensor, infused through the blood, for monitoring absolute temperature (and pH) in rat brain in vivo. Future temperature mapping studies with this MRS method may be useful for understanding temperature dynamics across different brain regions. It is of interest to note that regional brain temperature changes in an activated area are rather small and mostly positive (Figure 3). Qualitatively this suggests that the magnitude of mismatch between changes in CBF and CMRO2 is small (Table 1; Figures 2B and 2C), which in turn has implications for BOLD. However, a quantitative understanding about regional temperature changes in the brain activation was established from our multimodal measurements of Tt, CBF, and CMRO2 . The small increase in Tt can be attributed to a small mismatch between changes in CBF and CMRO2 . These results also suggest that temperature is being tightly regulated during increased CMRO2 and provides an indirect explanation for the commensurate (albeit higher) increase in CBF during functional activation. However, future studies on brain temperature regulation should measure Tt in a neuroimaging format and relate it with CMRO2 and CBF mapping.
Conclusion Stimulation-induced dynamics of Tt were modeled using measured alterations in blood flow and oxidative metabolism. Three main factors contributed to DTt(t) during brain activation: Qm, Qf, and Qc. The increases in Tt were matched by large increases in Qm, which warmed the ROI. While cooling of the ROI because of Qf and Qc were enhanced during the stimulation period, the net effects of all these components was a small increase in temperature. However, temperature of the arterial blood (or the core) and baseline oxidative metabolism could also impact the efficiency of brain cooling, and therefore changes in Tt.
Acknowledgements The authors thank Peter Brown, Scott McIntyre, and Terry Nixon at the Yale MRRC for maintaining the spectrometer, and Drs Natasja Maandag, Peter Herma´n, and Hal Blumenfeld for helpful discussions. This work will be submitted to fulfill in part the requirements for the degree of Doctor of Philosophy (LIS) at Yale.
Understanding localized temperature dynamics in brain HKF Tru¨bel et al
References Abrams RM, Stolwijk JAJ, Hammel HT, Graichen H (1965) Brain temperature and brain blood flow in unanesthetized rats. Life Sci 4:2399–410 Ackers GK, Doyle ML, Myers D, Daugherty MA (1992) Molecular code for cooperativity in haemoglobin. Science 255:54–63 Ances BM, Wilson DF, Greenberg JH, Detre JA (2000) Dynamic changes in cerebral blood flow, O2 tension, and calculated cerebral metabolic rate of O2 during functional activation using oxygen phosphorescence quenching. J Cereb Blood Flow Metab 21:511–6 Baker MA (1979) A brain-cooling system in mammals. Sci Am 240:114–22 Baker MA, Hayward JN (1967) Autonomic basis for the rise in brain temperature during paradoxical sleep. Science 157:1586–8 Baker MA, McFrye F, Millet VE (1973) Origin of temperature changes evoked in the brain by sensory stimulation. Exp Neurol 38:502–19 Benzinger TH (1969) Heat regulation: homeostasis of central temperature in man. Physiol Rev 49:671–759 Bernard SA, Gray TW, Buist MD, Jones BM, Silvester W, Gutteridge G, Smith K (2002) Treatment of comatose survivors of out-of-hospital cardiac arrest with induced hypothermia. N Engl J Med 346:557–63 Boudry MR (1976) A simple conversion formula for type ‘T’ (copper–constantan) thermocouple readings. J Phys E: Sci Instrum 9:1064–5 Boulant JA (1998) Hypothalamic neurons. Mechanisms of sensitivity to temperature. Ann NY Acad Sci 856: 108–15 Boulpaep EL (2003) The microcirculation. In: Medical physiology (Boron WF, Boulpaep EL, eds), Philadelphia: Saunders, 463–82 Brugge JF, Poon PWF, So ATP, Wu BM, Chan FHY, Lam FK (1995) Thermal images of somatic sensory cortex obtained through the skull of rat and gerbil. Exp Brain Res 106:7–18 Collins CM, Smith MB, Turner R (2004) Model of local temperature changes in brain upon functional activation. J Appl Physiol 97:2051–5 Craigie E (1945) The architecture of the cerebral capillary bed. Biol Rev 20:133–46 de Poorter J, De Wagter C, De Deene Y, Thomsen C, Stahlberg F, Achten E (1995) Noninvasive MRI thermometry with the proton resonance frequency (PRF) method: in vivo results in human muscle. Magn Reson Med 33:74–81 Delgado JM, Hanai T (1966) Intracerebral temperatures in free-moving cats. Am J Physiol 211:755–69 Erinjeri JP, Woolsey TA (2002) Spatial integration of vascular changes with neural activity in mouse cortex. J Cereb Blood Flow Metab 22:353–60 Flores-Maldonado A, Medina-Escobedo CE, RiosRodriguez HM, Fernandez-Dominguez R (2001) Mild perioperative hypothermia and the risk of wound infection. Arch Med Res 32:227–31 Germain D, Chevallier P, Laurent A, Saint-Jalmes H (2001) MR monitoring of tumour thermal therapy. Magma 13:47–59 Ginsberg MD, Busto R (1998) Combating hyperthermia in acute stroke: a significant clinical concern. Stroke 29:529–34 Gluckman PD, Wyatt JS, Azzopardi D, Ballard R, Edwards AD, Ferriero DM, Polin RA, Robertson CM, Thoresen
M, Whitelaw A, Gunn AJ (2005) Selective head cooling with mild systemic hypothermia after neonatal encephalopathy: multi-centre randomised trial. Lancet 365: 663–70 Gorbach AM, Heiss J, Kufta C, Sato S, Fedio P, Kammerer WA, Solomon J, Oldfield EH (2003) Intraoperative infrared functional imaging of human brain. Ann Neurol 54:297–309 Gu¨ler AD, LeeH, Iida T, Shimizu I, Tominaga M, Caterina M (2002) Heat-evoked activation of the ion channel, TRPV4. J Neuroscience 22:6408–14 Hayashi N (2004) A new concept of brain hypothermia treatment and pitfalls in intensivecare unit hypothermia management. In: Hypothermia for acute brain damage (Hayashi N, Bullock R, Dietrich DW, Maekawa T, Tamura A, eds), Heidelberg: Springer-Verlag, 49–75 He Q, Zhu L, Weinbaum S (2003) Effect of blood flow on thermal equilibration and venous rewarming. Ann Biomed Eng 31:659–66 Hernandez MJ, Brennan RW, Bowman GS (1978) Cerebral blood flow autoregulation in the rat. Stroke 9:150–5 Hetherington PA, Swindale NV (1999) Receptive field and orientation scatter studied by tetrode recordings in cat area 17. Vis Neurosci 16:637–52 Holdcroft A (1980) Body temperature control. London, UK: Baillier Tindall Hyder F (2004a) Neuroimaging with calibrated fMRI. Stroke 35(Suppl 1):2635–41 Hyder F (2004b) Deriving changes in CMRO2 from calibrated fMRI. In: Brain Energetics and Neuronal Activity: Applications to fMRI and Medicine (Shulman RG, Rothman DL, eds), London, UK: Wiley, 147–71; Chapter 9 Hyder F, Kennan RP, Kida I, Mason GF, Behar KL, Rothman DL (2000b) Dependence of oxygen delivery on blood flow in rat brain: a 7 Tesla nuclear magnetic resonance study. J Cereb Blood Flow Metab 20:485–98 Hyder F, Kida I, Behar KL, Kennan RP, Maciejewski PK, Rothman DL (2001) Quantitative functional imaging of the brain: towards mapping neuronal activity by BOLD fMRI. NMR Biomed 14:413–31 Hyder F, Renken R, Kennan RP, Rothman DL (2000a) Quantitative multi-modal functional MRI with blood oxygenation level dependent exponential decays adjusted for flow attenuated inversion recoveries (BOLDED AFFAIR). Magn Reson Imag 18:227–35 Hyder F, Rothman DL, Blamire AM (1995) Image reconstruction of sequentially sampled echo-planar data. Magn Reson Imag 13:97–103 Hyder F, Rothman DL, Shulman RG (2002) Total neuroenergetics support localized brain activity: implications for the interpretation of fMRI. Proc Natl Acad Sci USA 99:10771–6 Incropera FP, DeWitt DP (2002) Fundamentals of heat and mass transfer. Hoboken, NJ: Wiley Jones SC, Williams JL, Shea M, Easley KA, Wei D (1995) Cortical cerebral blood flow cycling: anesthesia and arterial blood pressure. Am J Physiol 268:H569–75 Kida I, Hyder F, Behar KL (2001) Inhibition of voltagedependent sodium channels suppresses the functional MRI response to forepaw somatosensory activation in the rodent. J Cereb Blood Flow Metab 21:585–91 Kida I, Kennan RP, Rothman DL, Behar KL, Hyder F (2000) High-resolution CMRO2 mapping in rat cortex: a multi-parametric approach to calibration of BOLD image contrast at 7 Tesla. J Cereb Blood Flow Metab 20:847–60
77
Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78
Understanding localized temperature dynamics in brain HKF Tru¨bel et al 78
Kiyatkin EA, Brown PL (2003) Fluctuations in neural activity during cocaine self-administration: clues provided by brain thermorecordings. Neuroscience 116: 525–38 Kiyatkin EA, Mitchum RD, Jr (2003) Fluctuations in brain temperature during sexual interaction in male rats: an approach for evaluating neural activity underlying motivated behavior. Neuroscience 119:1169–83 LaManna JC, McCracken KA, Patil M, Prohaska OJ (1989) Stimulus-activated changes in brain tissue temperature in the anesthetized rat. Metabol Brain Dis 4:225–37 Li L, Shapiro EM, Leigh JS (2001) Significant precision improvement for temperature mapping. Magn Reson Med 46:678–82 Liu J (2001) Uncertainty analysis for temperature prediction of biological bodies subject to randomly spatial heating. J Biomech 34:1637–42 McElligott JG, Melzack R (1967) Localized thermal changes evoked in the brain by visual and auditory stimulation. Exp Neurol 17:293–312 Melzack R, Casey KL (1967) Localized temperature changes evoked in the brain by somatic stimulation. Exp Neurol 17:276–92 Mills FC, Johnson ML, Ackers GK (1976) Oxygenationlinked subunit interactions in human hemoglobin: experimental studies on the concentration dependence of oxygenation curves. Biochemistry 15:5350–62 Miyazawa T, Hossmann KA (1992) Methodological requirements for accurate measurements of brain and body temperature during global forebrain ischemia of rat. J Cereb Blood Flow Metab 12:817–22 Nadel E (2003) Regulation of body temperature. In: Medical physiology (Boron WF, Boulpaep EL, eds), Philadelphia: Saunders, 1231–41 Nagashima K, Nakai S, Tanaka M, Kanosue K (2000) Neuronal circuitries involved in thermoregulation. Auton Neurosci 85:18–25 Nwaigwe CI, Roche MA, Grinberg O, Dunn JF (2000) Effect of hyperventilation on brain tissue oxygenation and cerebrovenous pO2 in rats. Brain Res 868:150–6 Ogawa S, Menon RS, Tank DW, Kim SG, Merkle H, Ellermann JM, Ugurbil K (1993) Functional brain mapping by blood oxygenation level-dependent contrast magnetic resonance imaging. Biophys J 64:803–12 Pennes HH (1948) Analysis of tissue and arterial blood temperatures in the resting human forearm. J Appl Physiol 1:93–122 Quock RM, Panek RW, Kouchich FJ, Rosenthal MA (1987) Nitrous oxide-induced hypothermia in the rat. Life Sci 41:683–90 Serota HM, Gerard RW (1938) Localized thermal changes in the cats brain. J Neurophysiol 1:115–24 Shevelev IA, Tsicalov EN (1997) Fast thermal waves spreading over the cerebral cortex. Neuroscience 76: 531–40 Shevelev IA, Tsicalov EN, Gorbach AM, Budko KP, Sharaev GA (1993) Thermoimaging of the brain. J Neurosci Methods 46:49–57 Shitzer A, Kleiner MK (1980) On the relationship between blood perfusion, metabolism and temperature in biological tissue heat balance. J Biomech Eng 102:162–9 Shulman RG, Hyder F, Rothman DL (2001) Lactate efflux and the neuroenergetic basis of brain function. NMR Biomed 14:389–96
Journal of Cerebral Blood Flow & Metabolism (2006) 26, 68–78
Shulman RG, Rothman DL, Behar KL, Hyder F (2004) Energetic basis of brain activity: implications for neuroimaging. Trends Neurosci 27:489–95 Shulman RG, Rothman DL, Hyder F (1999) Stimulated changes in localized cerebral energy consumption under anesthesia. Proc Natl Acad Sci USA 96:3245–50 Siesjo¨ BK (1978) Brain energy metabolism. New York: Wiley & Sons, 5 Smith AJ, Blumenfeld H, Behar KL, Rothman DL, Shulman RG, Hyder F (2002) Cerebral energetics and spiking frequency: the neurophysiological basis of fMRI. Proc Natl Acad Sci USA 99:10765–70 Swan H (1974) Thermoregulation and bioenergetics. New York, NY: Elsevier Thompson HJ, Pinto-Martin J, Bullock MR (2003a) Neurogenic fever after traumatic brain injury: an epidemiological study. J Neurol Neurosurg Psychiatry 74:614–9 Thompson HJ, Tkacs NC, Saatman KE, Raghupathi R, McIntosh TK (2003b) Hyperthermia following traumatic brain injury: a critical evaluation. Neurobiol Dis 12:163–73 Thoresen M, Simmonds M, Satas S, Tooley J, Silver IA (2001) Effective selective head cooling during posthypoxic hypothermia in newborn piglets. Pediatr Res 49:594–9 Thulin A (1969) Composite thermocouple-resistance sensors with linear output. J Phys E: Sci Instrum 2:1069–72 Tru¨bel H, Herman P, Kampmann C, Huth C, Maciejewski PK, Novotny E, Hyder F (2004) A novel approach for selective brain cooling: implication for hypercapnia and seizure activity. Int Care Med 30:1829–33 Tru¨bel HKF, Maciejewski PK, Farber JA, Hyder F (2003) Brain temperature measured by 1H NMR in conjunction with a lanthanide complex. J Appl Physiol 94:1641–9 Ueki M, Linn F, Hossman KA (1988) Functional activation of cerebral blood flow and metabolism and after global ischemia of rat brain. J Cereb Blood Flow Metab 8: 486–94 van Leeuwen GM, Hand JW, Lagendijk JJ, Azzopardi DV, Edwards AD (2000) Numerical modeling of temperature distributions within the neonatal head. Pediatr Res 48:351–6 Vinogradov SA, Wilson DF (1997) Extended porphyrins. New IR phosphors for oxygen measurements. Adv Exp Med Biol 411:597–603 Voets TH, Droogmans G, Wissenbach U, Janssens A, Flockerzi V, Nilius B (2004) The principle of temperature-dependent gating in cold- and heat sensitive TRP channels. Nature 430:748–54 Wadhwani KC, Rapoport RI (1988) Blood flow in the central and peripheral nervous system. In: Laser¨ berg PA, ed), Boston, MA: Doppler blood flowmetry (O Kluwer Academic, 265–88 Woods RL, Lawrence KL (1997) Modeling and Simulation of Dynamic Systems. New Jersey, NJ: Prentice Hall Inc Yablonskiy DA, Ackerman JJH, Raichle ME (2000) Coupling between changes in human brain temperature and oxidative metabolism during prolonged visual stimulation. Proc Natl Acad Sci USA 97:7603–8 Yager JY, Armstrong EA, Jaharus C, Saucier DM, Wirrell EC (2004) Preventing hyperthermia decreases brain damage following neonatal hypoxic–ischemic seizures. Brain Res 1011:48–57
