Determination of whole-brain oxygen extraction fractions by fast measurement of blood T2 in the jugular vein

Abstract

The oxygen extraction fraction (OEF) of the brain reports on the balance between oxygen delivery and consumption and can be used to assess deviations in physiological homeostasis. This is relevant clinically as well as for calibrating BOLD fMRI responses. OEF is reflected in the arteriovenous difference in oxygen saturation fraction (Y_v - Y_a), which can be determined from venous T₂ values when arterial oxygenation is known. A pulse sequence is presented that allows rapid measurement (< 1 min) of blood T₂s in the internal jugular vein. The technique combines slice-saturation and blood inflow to attain high signal to noise ratio in blood and minimal contamination from tissue. The sequence is sensitized to T₂ using a non-selective Carr-Purcell-Meiboom-Gill (CPMG) T₂ preparation directly after slice saturation. Fast scanning (TR of about 2 s) is possible by using a non-selective saturation directly after acquisition to rapidly achieve steady-state longitudinal magnetization. The venous T₂ (for 10 ms CPMG inter-echo time) for normal volunteers was 62.4 ± 6.1 ms (n = 20). A calibration curve relating T₂ to blood oxygenation was established using a blood perfusion phantom. Using this calibration, a whole-brain OEF of 0.37 ± 0.04 was determined (n = 20), in excellent agreement with literature values.

Introduction

The oxygen extraction fraction (OEF) of the brain reports on the balance between oxygen delivery and consumption (1). Slight changes in OEF may reflect physiological perturbation, and a method to rapidly and noninvasively assess this parameter should be useful for clinical assessment of brain homeostasis. In addition, the BOLD functional MRI effect reflects focal changes in OEF during activation (2–4), and changes in baseline OEF will affect the size of the BOLD effect. Measurement of whole-brain OEF would therefore be useful to calibrate the BOLD effect for baseline blood oxygenation in a manner similar to Lu et al. (5). The OEF is defined as (1,6):

$$\mathit{OEF}=\frac{\mathit{oxygen}\mathit{consumption}}{\mathit{oxygen}\mathit{delivery}}=\frac{{\mathit{CMRO}}_2}{C_a·\mathit{CBF}}=\frac{{\mathit{CMRO}}_2}{[{Hb}_{\mathit{tot}}]·Y_a·\mathit{CBF}}$$ [1]

in which CMRO₂ is the cerebral metabolic rate for oxygen and C_a the oxygen content, the product of total hemoglobin concentration [Hb_tot] in mM and the arterial oxygen saturation fraction (Y_a). CBF is the cerebral blood flow.

The venous deoxygenation fraction (1−Y_v) is directly related to OEF via the modified Kety equation accounting for hypoxia (3)

$$1-Y_v=1-Y_a+\mathit{OEF}·Y_a$$ [2]

The blood deoxygenation fraction (1−Y) in any type of vessel is directly related to 1/T₂ in that vessel (7–12) and calibration curves relating these two parameters can be used to determine Y-values from absolute T₂ (13–17). Determination of the venous relaxation times T_2,v, would allow determination of the oxygenation fractions of blood in these vessels, and, if Y_a is known (or assumed) and a calibration curve available, the subsequent calculation of OEF.

Measurement of blood T₂ in vivo is challenging because of the rapid flow of blood in larger vessels and the small size of microvascular blood vessels. To avoid the wash-in/wash-out effect encountered in conventional spin-echo sequences for large blood vessels, magnetization-prepared T₂ weightings (T₂ prep) are typically employed before image acquisition (18). Flow-insensitive T₂ prep uses only non-selective (NS) RF pulses and consists of a train of refocusing pulses bracketed by an excitation pulse and a flip-back pulse with a total duration TE_prep (18). This was first used to measure coronary venous blood T₂ a decade ago (14). Recently, this approach was extended to determine T₂ in the sagittal sinus (19). In that study, the complication of partial volume effects with tissue around the sagittal sinus was reduced by using the difference between the control and the spin labeled (on venous side) scans (19). This spin-tagging technique took about 5 minutes due to the long TR (8 s) needed to wait for equilibrium and the use of a control/label pair for subtraction.

Jugular venous blood primarily drains the cerebral hemispheres (usually 97%, minimally 93% of the intracranial circulation system) (20) and is a favorable sample source to estimate whole-brain OEF. In this work, we present a pulse sequence that allows rapid measurement (48 s) of jugular blood T₂s by exploiting the fact that the internal jugular vein (IJV) has faster flow and a bigger cross-sectional area than the sagittal sinus. In order to determine venous oxygenation (Y_v) and thus OEF from the measured T₂ value, a calibration curve relating T₂ and Y at different hematocrit (Hct) and oxygenation levels (Y) was measured in vitro using a blood circulation system.

Methods

Pulse Sequence for In Vivo Blood T₂ Measurement

All experiments were performed on a 3T Philips Achieva scanner (Philips Medical Systems, Best, The Netherlands). In vivo experiments used the body coil for transmission (maximum RF amplitude 575 Hz) and an 8-channel head coil for reception. The pulse sequence diagram for measuring blood T₂ in the jugular vein is shown in Figure 1a. It includes 4 blocks within each TR: slice-selective (SS) saturation, non-selective (NS) T₂ prep, SS excitation/acquisition and NS saturation. The NS T₂ prep block starts with a hard pulse excitation (90°_x, 0.44 ms), followed by a series of composite refocusing pulses (90°_x180°_y90°_x, pulse width (pw) = 1.75 ms) with an MLEV phase pattern (21), and then a flip-back pulse (270°_x360°_−x, 3.05 ms), which was made composite to compensate for RF field (B₁) inhomogeneity (18).

Figure 1.

(a) Pulse sequence diagram for measuring blood T₂ in the internal jugular vein (IJV). There are 4 blocks within each TR: slice-selective (SS) saturation, non-selective (NS) T₂ prep, SS excitation/acquisition, and NS saturation. T_sat = 200 ms, τ_CPMG = 10 ms, TE_prep = [20, 40, 80, 160] ms, and TE = 15 ms, T_delay = 2 s. (b–c) Localization of IJV: (b) Anatomical scout image with sagittal view. Angiographic scout images with coronal view (c) and with sagittal view (d) are also shown. The orange line indicates the center of the imaging slice (red) perpendicular to the neck vessels in both coronal view and sagittal view. The green box is the localized shimming box that aligns with the imaging slice and is centered at the side of the largest IJV of the subjects.

In order to generate a preparation sequence with multiple echo times, TE_prep = [20, 40, 80, 160] ms, and constant inter-echo spacing of τ_CPMG = 10 ms, the number of refocusing pulses, k, was chosen to be 2, 4, 8, and 16, respectively (19). The necessity to use composite pulses to accomplish reliable refocusing in vivo and the limit of the RF amplitude of the body coil within a human scanner resulted in a RF pulse fraction of almost 18% of the refocusing interval. Using the theory in (22), it can be calculated that this will lead to almost 9% overestimation (Eq. [2] of (22): ΔT₂/T₂ ≈ (1/2) (pw/τ_CPMG) = (1/2)·(1.75/10) ≈ 8.8%) of the measured T₂, due to the slower T₁ decay of the temporarily stored longitudinal magnetizations during the composite pulses. The corrected echo times without T₁ influence can be calculated based on Eq. [3] of (22): TE_prep,corr = TE - k· (pw/2) · (1−T_2,v/T_1,v) = [18.3, 36.6, 73.2, 146.5] ms, assuming T_2,v = 60 ms and T_1,v = 1800 ms (23). These TE_prep,corr values are insensitive to different relaxation time ratios of blood (T₁/T₂ between 10 to 40). At the end of the T₂ prep block, the longitudinal magnetizations of both static tissue and flowing blood are weighted by their T₂ factor, exp(−TE_prep,corr/T₂). The residual transverse magnetizations are dephased by a spoiling gradient after the last tip-up pulse (18).

After SS excitation (THK = 5 mm), linear-order flow compensation (FC) gradients, and data acquisition with following spoiling gradients, a NS adiabatic saturation pulse (BIR-4 (24), of length 10 ms, and bandwidth 400 Hz) immediately resets all remaining longitudinal magnetization to zero. As a consequence, after a fixed time delay, T_delay, the longitudinal magnetization of the blood right before the next T₂ prep block is proportional only to 1−exp(−T_delay/T₁). By using the same predelay time (T_delay) between global saturation and the next scan, the same steady-state will be achieved before each T₂ prep block, independent of the length of the preparation and the previous T₂ weighting history (14,18).

When using the sequence above, tissue can obscure the blood signals, because tissue magnetization also experiences T₂ weighting and its T₂ is on the order of 60 ~ 90 ms at 3T, comparable to venous blood T₂ at τ_CPMG = 10 ms. Given that blood velocity (v) of healthy volunteers is in the range of 5~40 cm/s in the IJV (25), a SS saturation pulse applied on the same image slice is performed at T_sat = 200 ms before the SS excitation. This T_sat is short enough to saturate the static tissue signal and yet long enough to replenish the flowing spins in the imaging slice (T_sat > THK/v). Even with the reported slowest velocity v = 5 cm/s in the IJV, T_sat = 200 ms is still twice the time for complete replacement of blood in an imaging slice of thickness (THK) 5 mm.

In Vivo Data acquisition

A total of 20 healthy volunteers (age: 39 ± 13 yrs; range: 24–63 yrs; 10 females and 10 males) were enrolled for parameter optimization and determination of blood T₂, all of whom provided informed written consent in accordance with local IRB guidelines.

As part of scan planning, phase-contrast MRA (PCA) survey images were acquired in both coronal and sagittal planes to visualize the location and orientation of IJVs and internal carotid arteries (ICAs). PCA survey scans used a 50 mm slab, TR/TE = 20/5.8 ms, FOV = 250 × 250 mm², and a scan matrix of 256 × 128 (acquisition time: 20 s for 2 averages). Phase-contrast velocity measurements for IJVs were performed on a 5 mm axial slice, positioned perpendicular to the targeted IJV about 15 mm below the superior bulb (Figure 1b, c, d). The encoding velocity was set to be 40 cm/s. Acquisition parameters: TR/TE = 15/9 ms, 5 dynamics, FOV = 128 × 128 mm², acquisition matrix = 128 × 128 (reconstruction 256 × 256). The total acquisition duration for velocity measurement was about 30 sec.

The blood T₂ measurements were performed at the same location as the blood velocity measurement. Acquisition parameters were: FOV = 200 × 150 mm², acquisition matrix = 192 × 132, with left-right being the phase encoding direction. Data were zero-filled to 0.78 × 0.78 mm² in-plane resolution. In order to minimize image distortion and signal loss, k-space was acquired with a 6-segment gradient echo EPI (N_shots = 6), with SENSE acceleration factor = 2, resulting in a minimum TE of 15 ms. T_delay = 2 s was used for most of the measurements, which led to a total T₂ measurement time ≈ 2 s (T_delay) × 6 (N_shots) × 4 (N_TEprep) = 48 s.

A SS pre-saturation block was added to minimize the stationary tissue signals but not perturb the T₂ weighting of the flowing blood. To illustrate its effect, data were acquired for five subjects with and without the SS saturation block. Both images and fitted T₂s from blood vessels were compared.

Since the NS T₂ prep and NS saturation block both aim for blood covered within the body transmit coil (brain + neck) before it flows through the IJV and because the RF pulses have limited bandwidth, it is important to properly shim on the whole volume within the transmit coil, in addition to the local imaging slice. Therefore, as suggested from our blood T₁ study (23), only 1^st order shims were adjusted locally following a global shim in order not to affect the shimming around the neck area to an extent that the field there is outside the bandwidth. The effect of NS saturation block was therefore examined for different areas of the brain (top, middle, and bottom) by using the same BIR-4 pulse right before SS excitation/acquisition, while keeping the shimming conditions used in blood T₂ measurements.

Effects of different T_delay were evaluated on five volunteers, with the expectation that longer T_delay should only increase the signal-to-noise ratio at the cost of longer scan duration, but not change the measured blood T₂s. T_delay values of 2, 3, 4, 5, and 6 s were used, corresponding to a total acquisition duration of about 1, 1.5, 2, 2.5, and 3 min, respectively, while all other imaging parameters were kept the same. Some subjects (three of them) have left and right IJVs similar in orientation and sizes so they were also chosen to inspect the correspondence of measured blood T₂s from the two vessels. The same T₂ results were confirmed from the left and right IJVs of the same subject (data not shown).

Subjects in all studies were instructed to stay awake during the scans, which was checked frequently, to avoid potential baseline oxygenation changes during sleep.

In Vitro Experiments

Calibration curves relating transverse relaxation times and oxygenation (Y = 0.35 ~ 1.0) were measured for bovine blood over a range of physiologically relevant hematocrit (Hct) values (0.36, 0.40, 0.46, 0.50). Blood was circulated in a gas exchange perfusion system at controlled physiological temperature (37°C) within the 3T magnet. Details of the blood preparation and the perfusion system setup were described previously (26). Once a stable Y value is achieved within a mixing chamber, bovine blood is pumped into a circulation loop, and maintained at a speed about 1 cm/s through a 1 cm-diameter, 10 cm-long, vertically positioned sample tube for the relaxation time measurements. Blood was sampled right before and after the MR measurements and oxygenation determined using a blood gas analyzer (Radiometer, ABL700) to ensure less than 3% difference of Y. In addition, blood pO2 was monitored during the MR measurement with a needle-encased sensor (Oxford Optronix) placed close to the tube.

In vitro blood studies were conducted with the same 3T magnet as the in vivo scans, but a quadrature head coil was used for both transmission (maximum RF amplitude 851 Hz) and reception. In vitro blood T₂ measurements utilized a sequence similar to the one described in Figure 1a, but without the SS saturation block before the T₂ prep, because this SS saturation approach is only suited for fast flowing blood and there is no need to suppress static tissues in the phantom. Within the T₂ prep block, hard pulses were used for 90° and 180° pulses with durations 0.3 ms and 0.6 ms respectively and the CPMG phase pattern was applied, similar to other work in the field (27,28). Imperfect refocusing pulses will lead to undesired signal loss. So a water phantom (with T₂ ≈ 2 s) was first scanned to confirm proper 180o pulses (data not shown), which were easy to attain over the small sample volume. Using τ_CPMG = 10 ms, T₂ was measured using six equally-spaced TE_preps ranging from 20 to 120 ms. After SS excitation (THK = 10 mm), a single-shot gradient-echo EPI was employed for acquisition (FOV = 64 × 64 mm², 32 × 32 acquisition matrix, TE = 11 ms). T_delay = 3 s was fixed between the NS saturation after acquisition and the following T₂ prep block for all of the in vitro blood T₂ measurements. The total measurement time was 3 s (T_delay) × 1 (N_shots) × 6 (N_TEprep) ≈ 18 s for each specific Y and Hct value.

Data Analysis

Matlab 7.0 (MathWorks, Inc., Natick, MA, USA) was used for data processing. Since the surrounding tissue signals were suppressed, regions of interest (ROIs) at IJVs could be easily drawn manually on the images acquired at TE_prep = 20 ms. Blood T₂ for each pixel was fitted using a linear-least-square algorithm for the monoexponential signal decay curve. For in vivo data, the shortened TE_prep,corr for composite refocusing pulses was used; while the original TE_prep for hard 180° pulses was used in vitro. For individual measurements, the standard error of the estimated T₂s fitted from ROI-based mean values are reported. For a group of measurements, the mean and standard deviation (std) of the averaged T₂s over all subjects were calculated.

Extensive literature exists on studying the dependence of blood T₂ on Y, Hct and τ_CPMG (3,9,12,16,27–30). To explain the exact mechanisms underlying the measured effects, both an exchange model between plasma and erythrocyte (30) and a diffusion model with weak magnetic inhomogeneities (29) have been postulated. Our in vitro blood T₂ values obtained at 3T are still being analyzed to further elucidate this topic. However, exact knowledge of the mechanistic relationship between 1/T₂ and (1−Y) is not needed to measure a calibration curve. Therefore, the curves between T₂ and Y were fitted similar to the literature, namely as an empirical quadratic equation, which approximates the experimental data very well:

$$1/{\text{T}}_2=\text{A}+\text{C}{(1-\text{Y})}^2$$ [3]

The fitted calibration curve at Hct = 0.40 was used to convert the blood T₂ measured in vivo to corresponding oxygen saturation fraction levels Y. Subsequently, OEF was estimated from:

$$\text{OEF}=({\text{Y}}_{\text{a}}-{\text{Y}}_{\text{v}})/{\text{Y}}_{\text{a}}$$ [4]

with Y_a assumed to be 0.98 (5). Because typical physiological Hct values range between 0.36 and 0.50, errors for estimating Y and OEF by using a calibration curve at a fixed Hct = 0.40 were also assessed.

Results

In Vitro Experiments

In Figure 2a, the transverse relaxation rates (1/T₂) of bovine blood at τ_CPMG = 10 ms are shown as a function of oxygen saturation fraction for four different hematocrits. The data clearly conform to linearity with respect to the square of deoxygenation level, (1−Y)², as described by Eq. [3]. The fitted parameters A and C for each condition as well as the calculated T₂ for a venous oxygenation level of 0.61 are summarized in Table 1.

Figure 2.

Dependencies of the transverse relaxation rate of blood at 3T on deoxygenation (a) and oxygen saturation fraction (b). (a) Results for all four Hct values show a linear dependence (Eq. [3]) on the square of deoxygenation level, (1−Y)² for τ_CPMG = 10 ms (R² > 0.99). (b) The corresponding T₂ vs. Y curves. The average of the Hct = 0.40 and Hct = 0.46 curves in (a) and (b) were used for the calibration curve for in vivo OEF estimation. The 50 to 80 ms T₂ range is zoomed out to show the effect of Hct, which causes only about 4 ~ 6% error as calculated in the Results session. All fit parameters are summarized in Table 1.

Table 1.

Quadratic fitting of Eq. [3] of in vitro blood (3T, 37°C) at τ_CPMG = 10 ms with four Hct values within physiology range, as well as the calculated T₂ for a typical venous oxygenation level of 0.61. The standard errors of A and C are shown as the uncertainty of the fit, which are then taken into account in the uncertainty estimation of the venous T₂ with Y=0.61.

Hct	A	C	T2 (ms) (Yv=0.61)
0.36	6.16±0.24	58.4±1.1	66.5±1.3
0.40	7.18±0.33	60.1±2.7	61.3±2.0
0.46	7.18±0.07	59.0±0.4	61.9±0.4
0.50	8.46±0.43	63.2±2.2	55.3±1.7

For the purpose of estimating Y from our in vivo blood T₂ measurements, the average constants of fitted results at Hct = 0.40 and 0.46 (A = 7.18, C = 59.6) were used. It can be seen from both Figure 2b and Table 1 that at a τ_CPMG of 10 ms, Hct = 0.40 and Hct = 0.46 have almost identical relationship between T₂ and Y over a broad range of possible venous Y values. The dependency of T₂ on Hct reduces with decreasing oxygenation, which is helpful for the accuracy of our estimates. For the range of T₂ values between 54 and 72 ms measured in the jugular vein in our study (reported below), Y estimated using the calibration curve at assumed Hct range between 0.40 and 0.46 will cause an overestimation of only about 4.4% for true Hct = 0.36 and an underestimation of about 6.0% for true Hct = 0.50. These will be the limits of the estimation errors of Y for the venous blood since most physiological Hct values are within this range.

In Vivo Experiments

In order to assess whether all blood is refreshed after the SS saturation, we first measured blood velocities in most subjects (n = 13). Average blood velocities in the IJVs were 15±6 cm/s, with a range of 5~21 cm/s (n = 13). These are all fast enough (limit v > 5 cm/s) for the proposed SS saturation not to affect the blood T₂ measurements. The effect of the SS pre-saturation block is demonstrated in Figure 3, where an image collected without saturation (Figure 3a,b) is compared to one with saturation (Figure 3d,e). Tissue signals were reduced significantly when using SS saturation, allowing easier selection of vascular ROIs. However, no significant difference was observed between the T₂ curves without (Figure 3c) and with SS saturation (Figure 3f), providing validation for this approach.

Figure 3.

Effect of the slice-selective pre-saturation block on image contrast and measured T₂ values. The first row shows an image (TE_prep = 20 ms) collected without saturation (a), with the zoomed-in view in the yellow box (b) and corresponding T₂ curves (c) from vascular ROIs (blue: IJV; red: ICA). In the second row, SS saturation was enabled (d–f). By minimizing static tissue magnetization using SS saturation before T₂ prep, identifying of ROIs around the jugular veins are substantially facilitated (d, e). This additional pulse also does not affect the T₂ fitting for the flowing blood (f). The estimated T₂s fitted from the averaged ROI signals are reported with the standard error of the nonlinear fit.

In order to test the effectiveness of the post-acquisition NS saturation, we performed T₂ measurements as a function of T_delay. The results in Figure 4 confirm that there is negligible effect of T_delay, based on which we used T_delay = 2 s, allowing data acquisition in a period of 48 s. One disadvantage of using a short predelay is a reduction in SNR, but based on the measured standard deviation, this is not yet an issue when using T_delay = 2 s.

Figure 4.

Average T₂s fitted for all subjects as a function of T_delay, showing excellent agreement for all delays. Error bars reflect the std across subjects. It can be seen that longer T_delay leads to higher SNR and less variation among subjects, with the cost of longer acquisition time.

The individual blood T₂ values from the IJV for all twenty volunteers are listed in Table 2. No significant differences in T₂ values (two -tailed Student t-test: P = 0.48) were observed between female subjects (T_{2, IJV} = 63.4 ± 7.0 ms, n = 10) and male subjects (T_{2, IJV} = 61.4 ± 5.3 ms, n = 10). Since T₂ of fully oxygenated blood depends strongly on Hct (Figure 2), it was found not reliable to estimate Y_a from T_{2, ICA}. We therefore assumed the reasonable value of Y_a = 0.98 for healthy subjects to calculate OEF from Eq. [4]. The averaged blood T₂ among the 20 healthy volunteers was T_{2, IJV} = 62.4 ± 6.1 ms, corresponding to Y_v = 0.61 ±0.03 and OEF = 0.37±0.04. When Y_a = 0.96 or 1.0 are taken instead, OEF values of 0.36 ±0.04 and 0.39 ±0.03 are found, respectively.

Table 2.

Measured T_{2, IJV} of 20 subjects and their individual Y_v (using Eq. [3], A = 7.18, C = 59.6) and corresponding OEF estimation (Eq. [4], assuming Y_a = 0.98). The standard errors of the fitted T_{2, IJV}, and their error propagations into Y_v and OEF, are shown as the uncertainty of the results for each individual.

subject	T2, IJV (ms)	Yv	OEF
female:1	68±6	0.64±0.06	0.34±0.06
2	54±2	0.56±0.03	0.42±0.03
3	67±2	0.64±0.02	0.35±0.03
4	60±3	0.60±0.04	0.39±0.04
5	72±1	0.66±0.02	0.32±0.02
6	54±1	0.56±0.02	0.42±0.02
7	70±2	0.65±0.02	0.33±0.02
8	67±2	0.64±0.02	0.35±0.03
9	67±2	0.64±0.02	0.35±0.03
10	55±2	0.57±0.03	0.42±0.03
mean ± std:	63.4 ± 7.0	0.62 ± 0.04	0.37 ± 0.04
male:1	61±3	0.61±0.04	0.38±0.04
2	68±4	0.64±0.04	0.34±0.04
3	57±3	0.58±0.04	0.41±0.04
4	54±4	0.56±0.05	0.42±0.06
5	68±4	0.64±0.04	0.34±0.04
6	55±1	0.57±0.02	0.42±0.02
7	64±2	0.62±0.03	0.36±0.03
8	66±1	0.63±0.02	0.35±0.02
9	64±2	0.62±0.03	0.36±0.03
10	57±1	0.58±0.02	0.41±0.02
mean ± std:	61.4 ± 5.3	0.61 ± 0.03	0.38 ± 0.03
female and male data combined
mean ± std:	62.4 ± 6.1	0.61 ± 0.03	0.37 ± 0.04

Discussion

A new approach for rapidly measuring blood T₂ in the major neck vessels and consequent determination of whole-brain OEF using a blood T₂ calibration as a function of oxygen saturation fraction is presented. The MRI measurement combines a T₂ prep sequence for blood T₂ measurement (14,18,19) and a pre-preparation slice-selective saturation pulse for reducing surrounding tissue signals. It was shown that the latter does not interfere with the T₂-weighting of the blood flowing into the slice. A non-selective saturation pulse right after image acquisition resets the longitudinal magnetization of the blood independent of the previous T₂ weighting pulses (14), which made the method more efficient than just waiting for its full longitudinal recovery (19).

In Vitro Calibration Experiments

T₂ values of blood are often converted to the blood oxygen saturation fractions (Y) using in vitro T₂ - Y calibration curves, which can be obtained either directly from human blood samples of the individuals studied (13,14) or indirectly from fresh animal blood (16,19) with comparable erythrocyte shape and physiological properties. Since the transverse relaxation time depends on the magnetic field and the field dependence has not been well characterized, previous work done at 1.5T (13,16,28) cannot be used reliably for T₂ calibration at 3T, and we therefore measured a calibration curve for this paper. The new calibration data are of comparable order of magnitude to those of a recent characterization of in vitro human blood T₂ at 3T as a function of Y and τ_CPMG that were fitted with a diffusion model (27). Based on results from Table 1 and Eq. [4] in that paper, we calculated a calibration curve at τ_CPMG = 10 ms, giving A = 6.91 and C = 44.4. Compared to our results (A = 7.18 and C = 59.6), these smaller calibration variables correspond with higher T₂s at the same Y values, and would lead to an underestimation of oxygenation levels by about 10~15% for the range of T₂ values between 54 and 72 ms. In order to better judge this small difference, it is important to realize that these calibration curves were acquired under different conditions. First, as can be seen in the Results/in vitro blood section, lower Hct would cause higher T₂. Our calibration curve is an average of Hct at 0.40 and 0.46, while the other study averaged over a range of Hct from 0.35 to 0.53 (27) for 10 subjects. Since the individual data were not presented, we cannot judge the origin of potential differences Second, our experiment was conducted at physiologic temperature as opposed to room temperature in (27). However, lower temperature should increase relaxation slightly, which can not explain the higher T₂ values obtained by Chen et al. Third, bovine blood instead of human blood was used in this study, Based on literature evidence (31), the physiologic and physical properties of erythrocytes in higher mammals are very similar, and no significant difference in transverse relaxation is expected More importantly, we used hard refocusing pulses in the T₂ prep, while Chen et al. used composite 180° pulses, which would overestimate T₂ by 9% as demonstrated in Methods/Pulse Sequence section. The first and last issues, in our assessment, resulted in the difference between the two studies.

In Vivo Experiments

The signal decay as a function of echo time (TE_prep,corr in this work) should be governed by the blood water protons experiencing relaxation during transverse evolution of magnetization. The mechanism of such relaxation is quite complex, and includes chemical exchange of water molecules between plasma and erythrocytes and/or diffusion through intra/extracellular magnetic gradients (3,7–9,12,29,32). As a consequence, the relaxation time measured depends on the inter-echo spacing τ_CPMG. Here we report blood T₂ at τ_CPMG = 10 ms, to be comparable to a recent study in the sagittal sinus (19). Similar to this earlier paper, the length of the refocusing pulses was not negligible. Since this lengthens the measured relaxation time due to partial T₁ decay during the composite pulses (22), TE_prep,corr is used for in vivo T₂ fitting, which, in our case, reduced the effective τ_CPMG to 9.15 ms. We used an in vitro calibration curve at τ_CPMG = 10 ms. This could lead to some slight underestimation of T₂ (less than 3%, as estimated based on the measured relaxation time at 10 ms), and we have neglected this small difference.

The in vivo blood T₂ measurements may have another source of irreversible dephasing of the signal from flowing through the inhomogeneous magnetic field. During half of the 10 ms interecho time, blood flows about 0.75 mm in IJVs (in average, v = 15 cm/s, from this study) and up to 4 mm in ICAs (v < 1 m/s, (33)). The shimming employed in this study (global shim followed by local 1^st order linear shim) should minimize this external magnetic field inhomogeneity effect for IJVs, and less optimized for ICAs. As shown from the results (Figure 3, Table 1), the T₂ values measured in jugular veins were more robust than in carotids. The same effect was also reported in an earlier work (13). Careful shimming is a necessity to avoid irreversible dephasing of blood signals due to fast flowing between long interecho times.

At 3T and higher fields, B₁ inhomogeneity reduces pulse efficiency and complicates the measurement of relaxation rates through incorrect refocusing. To counteract these challenges, we used composite refocusing pulses with MLEV phase pattern for T₂ prep (18,21) and a BIR-4 pulse (24) for NS saturation. Reproducible T₂ measurements could be obtained in the IJV (Figure 3, Table 1). In principle, in addition to the stronger Y-dependence for T₂ in arteries, the less stable ICA results can also be due to the more nonuniform B₁ experienced by blood. For instance, when using 160 ms T₂ prep, the first pulse may be experienced by blood at the edge of the body coil range, while the last one may be 15 cm higher, where B₁ could be different. However, this is expected to be a small effect when using body coils for excitation. If a head-only transmit coil, instead of the body coil, were to be used in this work, the effect on jugular T₂ measurement remains to be seen, but this problem is expected to be much worse for carotid T₂ measurement. To address this problem, another set of composite pulses were used in (14) and adiabatic pulses were suggested in (34). However, longer pulses length and higher power deposition are associated with these alternative pulses.

OEF Estimation

When assuming Y_a = 0.98 for our normal subjects, and combining the blood T₂ measurements in the jugular vein with our calibration curve obtained from in vitro blood T₂ measurements, a whole-brain OEF value of 0.37 ± 0.04 (range: 0.32–0.42, n = 20) was found (Table 2). This is in excellent agreement with literature values (Table 3) obtained from catheterization (35), PET (36–38) and MRI (15,16,19,39–41). Compared to other modalities, MRI has the advantage of being noninvasive, but different MR techniques have different acquisition schemes and results may vary with methodology or the location of the measurement. For instance, when small local draining veins are the desired ROIs (16), high resolution images need to be acquired to avoid partial voluming with tissue or CSF and a larger number of averages is necessary to increase SNR. Even though most MRI approaches provide similar OEF results, the durations of the measurements vary. For example, in a recently published study (42) determining venous oxygenations (27) the oximetry measurement took 4.5 min. Another recently proposed promising method (40) acquired T₂′ decay curves that were fitted with a sophisticated model for OEF and other parameters including deoxygenated blood volume, 1/T₂ of tissue and CSF volume fraction. However, this technique, which was validated for OEF measurement on a rat model (43), required a rather long acquisition time (8.5 min). Alternatively, MR susceptometry (41) was demonstrated to be useful for whole-brain OEF measurement at jugular vein. This technique records phase images to calculate susceptibility differences between blood in jugular vein and surrounding tissues. This method is also very feasible, but requires a somewhat longer acquisition time to fully characterize the background gradients from the air-tissue interface (4.3 min). In contrast, our method is based on a simple T₂ model and takes only 0.8 minute for OEF measurement, making it a very suitable complementary technique for a variety of applications where a rapid whole-brain evaluation of OEF is useful.

Table 3.

Comparison of determined OEF with existing literature values for normal subjects.

Modality	Reference	Measurement location	OEF: mean±std (range, n)	Measurement time (minutes)
Catheterisation	Chieregato et al., 2003 (35)	Jugular bulb	0.42 (0.29~0.54, n=12)
PET	Leender et al., 1990 (36)	Gray Matter	0.42 (0.27~0.50, n=34)
PET	Raichle et al., 2001 (37)	Gray/White Matter	0.40±0.09 (n=19)
PET	Fox et al., 1988 (38)	Visual Cortex	0.361
MRI 1.5T T2	Oja et al., 1999 (15)	draining veins near Visual Cortex	0.45 (0.37~0.54, n=7)2	11.0
MRI 1.5T T2	Golay et al., 2001 (16)	draining veins near Visual Cortex	0.38±0.04 (0.30~0.46, n=8)	6.0
MRI 3.0T T2	Lu et al., 2007 (19)	Sagittal Sinus	0.34±0.06 (n=24)	4.3
MRI 3.0T T2	Chen et al., 2010 (42)	IJV	0.40±0.083 (n=10)	4.5
MRI 1.5T T2′	An et al., 2000 (39)	Gray/White Matter	0.40±0.02 (0.37~0.44, n=8)	8.5
MRI 3.0T T2′	He et al., 2007 (40)	Gray/White Matter	0.38±0.05 (0.33~0.48, n=9)	8.5
MRI 3.0T susceptibility	Fernandez-Seara et al., 2006 (41)	IJV	0.33±0.08 (n=5)	4.3
MRI 3.0T T2	this study	IJV	0.37±0.04 (0.32~0.42, n=20)	0.8

¹Calculated using CMRO₂ = 1.5 μmol/g/min, Hct = 0.41, and CBF = 0.5 ml/g/min.

²Recalculated from Oja’s T_2v values using the newly determined calibration constants by Golay et al. at 1.5T (16).

³Calculated using Y_v = 0.58 and Y_a = 0.98 (42).

There are some confounding factors to keep in mind when using T₂ to determine blood oxygenation. As shown in Figure 2, Table 1, and other literature (9,12,32), blood T₂ depends on Hct, which may affect the OEF determination. The range of Hct values for women versus men generally is: female: 0.36 ~ 0.44; male: 0.41 ~ 0.50 (44). As calculated in the Results Section and visualized in Figure 2b, use of the calibration curve for Hct values between 0.40 and 0.46 would lead to 6.7% underestimation of OEF for subjects with Hct = 0.36 and 12.4% overestimation for Hct = 0.50. So even though the measured jugular T₂ values showed no significant variation between female and male groups, Y_v and OEF calibrated with a true Hct, if different, could potentially show some small difference. For instance for an average female group with lower Hct, a lower Y_v and higher OEF would be found. Therefore, caution should be exercised when employing a universal T₂ - Y calibration curve, especially for subjects with hematological diseases. Similarly, for physiological conditions such as hypoxia, the arterial blood oxygen saturation, Y_a, can not be assumed to be 98%. In this case, one could use the pulse oximeter to measure Y_a from subject’s fingertip to determine Y_a so that OEF can still be determined non-invasively through Eq. [4].

Potential Applications

In recent work (5), Lu et al. have shown that the baseline venous oxygenation of individuals is inversely correlated with BOLD response and may be a substantial factor causing intersubject variations in fMRI signals. Fast measurement of this physiological parameter on each subject before or after the functional scans, would allow it to be used as a linear regressor for normalization of BOLD signals among subjects. This will be especially important for patient groups with underlying aberrant physiology.

Demonstrated by Xu et al. in a different work (17), whole-brain OEF measurements can also be used for estimation of whole-brain CMRO₂, by using the relationship in Eq. [1]. Through phase-contrast MRA, whole-brain blood flow can be quantified by assessing the velocities across ROIs on both the ICAs and vertebral arteries. Since the amount of the blood supplying the cerebral brain should equal to the amount of the draining blood, IJVs could be alternative vessels for evaluating whole-brain blood flow, with the advantage of much less pulsation effects and relatively bigger cross sections. With additional high-resolution T₁ weighted images, the intracranial tissue volume can be used to normalize the total CBF to yield the unit volume CBF, which is the CBF that needs to be used in Eq. [1] when looking at whole-brain metabolism.

Conclusions

A technique was developed to measure T₂ of venous blood in the jugular vein in less than one minute, with minimum tissue contamination. In addition, a calibration curve was measured to allow determination of venous oxygenation and OEF based on the measured T₂s. This approach is expected to be useful for clinical application, e.g. to establish ischemia in patients, as well as for fast calibration of baseline OEF for BOLD experiments.