Comparison of MRI Methods for Measuring Whole-Brain Venous Oxygen Saturation

Abstract

Purpose

In this work, we compare susceptometry-based oximetry (SBO) and two T₂-based methods for estimating resting baseline SvO₂ in the superior sagittal sinus (SSS).

Methods

SBO is a field-mapping technique whereas in T₂-based methods the intravascular blood signal is isolated either with velocity-encoded projections [projection-based T₂ (PT₂)] or a tag-control scheme [T₂-relaxation under spin tagging (TRUST)] after T₂-preparation. The measurements were performed on twelve healthy subjects (mean age=33±6 years) at 3 Tesla field strength. The reliability, precision, and reproducibility were examined for the three techniques.

Results

The mean (± standard deviation) SvO₂ quantified by SBO, PT₂, and TRUST were found to be 65.9±3.3, 65.6±3.5, and 63.2±4.1%. The standard deviation (SD) for 10 consecutive measurements in the quantified SvO₂ was less than 2.7%, 4.7%, and 5.0% for SBO, PT₂, and TRUST across all subjects. In testing reproducibility across different days, the resulting SDs were 2.6, 3.5, and 2.0% for SBO, PT₂, and TRUST.

Conclusion

The results indicate that all three SvO₂ quantification techniques to be reliable with good agreement between PT₂ and SBO while TRUST yielded slightly lower values compared with the other two techniques.

INTRODUCTION

Measuring global cerebral venous oxygenation (SvO₂) in the sagittal sinus or jugular vein provides direct assessment of the brain’s ability to extract and metabolize oxygen. Clinical estimation of brain oxygen saturation has traditionally been performed by jugular vein catheterization (1–3), which has risks, including venous infection and thrombosis (4). Near-infrared spectroscopy (NIRS) is a noninvasive technique that is being used both clinically and in the research setting to quantify regional oxygenation in brain tissue (5). However, NIRS quantifies superficial regional instead of global oxygenation and cannot differentiate the signal of the arterial and venous blood (6). Some noninvasive MRI approaches have emerged recently for quantifying SvO₂ by means of measurement of blood T₂, such as the TRUST method (7,8) and projection-based T₂ quantification (PT₂) (9), building on earlier work by Wright et al (10).

T₂-based methods quantify the transverse relaxation of water protons in whole blood caused by spin sampling of two frequency-shifted compartments, intra- and inter-erythrocyte. Such a frequency shift is created due to local field inhomogeneity in the immediate vicinity of paramagnetic deoxyhemoglobin. Thus, a greater fraction of deoxyhemoglobin leads to enhanced signal attenuation by T₂ relaxation. However, converting the measured T₂ to SvO₂ requires an ex vivo calibration curve (11,12).

An alternative approach makes use of the induced magnetic field of intravascular blood such as in susceptometry-based oximetry (SBO) (13,14). In SBO, the vessel is modeled as a long paramagnetic cylinder (14) immersed in an external uniform magnetic field, and SvO₂ is computed based on the induced field shift in intravascular blood relative to the surrounding tissue, which serves as a calibration-free reference.

Previous studies demonstrated the utility of SBO for quantifying SvO₂ in the internal jugular (14), superior sagittal sinus (SSS) (15), and femoral veins (16). In agreement with the theoretical model, both computer simulations (17) and experimental results (15) confirmed that the method is robust to deviations from vessel circularity [such as the more triangular cross section of the SSS] for vessels with tilt angles less than 30° relative to the static magnetic field.

Detection of changes in SvO₂ in response to various physiologic stimuli has been demonstrated for both types of methods (9,18–20). For example, PT₂, SBO, and TRUST were used to detect the effect of hypercapnia on SSS SvO₂ (9,18,21). SBO was also used to detect changes in SSS SvO₂ in response to apneic challenge (19). In addition, TRUST was used to examine and quantify changes in SSS SvO₂ in response to hypoxia and hyperoxia (20).

Although data in the literature suggest agreement among the three techniques in terms of the measured SvO₂ values, these methods have never been compared under controlled conditions in the same subjects. In order for these methods to be used clinically, their reliability and mutual agreement needs to be evaluated. The purpose of this study was to perform a systematic comparison of three methods for quantifying resting baseline SvO₂ in the SSS with pulse sequences and analysis algorithms replicated to closely match those in the parent literature.

METHODS

Theory of MRI-Based SvO₂ Quantification

T₂-Based Methods

The dependence of blood T₂ on SvO₂ results from the paramagnetism of deoxygenated hemoglobin, which creates field gradients in the vicinity of red blood cells. The enhanced transverse relaxation rate R₂ can be described in terms of the Luz-Meiboom chemical exchange model (22), which attributes this effect to irreversible dephasing of spins undergoing exchange between two compartments of different water proton resonance frequencies. The relaxation rate (R₂) in response to the Carr-Purcell-Meiboom-Gill (CPMG) pulse train can be written as (10):

$$R_2=R_{2,0}+\mathit{Hct}(1-\mathit{Hct}){(\mathrm{\alpha }{\mathrm{\omega }}_o(1-{\mathit{SvO}}_2))}^2\times \mathrm{\tau }\left[1-\frac{2\mathrm{\tau }}{{\mathrm{\tau }}_{cp}}tanh\left(\frac{{\mathrm{\tau }}_{cp}}{2\mathrm{\tau }}\right)\right]$$ [1]

where Hct is the hematocrit (volume fraction of erythrocytes in whole blood), τ is the exchange time of protons between the two chemically shifted sites, ω_o is the proton Larmor frequency, α is a dimensionless quantity related to the geometry of erythrocyte and the magnetic susceptibility of deoxyhemoglobin, τ_cp is the interecho spacing in the T₂-preparation CPMG train, and R_2,0 is the relaxation rate of fully oxygenated blood (i.e., arterial blood not in the presence of hypoxia).

Both T₂-based methods examined here use the same T₂-preparation by means of CPMG pulse trains of variable duration to confer T₂-weighting, except that the blood water signal is isolated in different ways. In TRUST, the signal is mapped and partial volume effects minimized by performing pairwise subtraction between the control and label images (7) analogous to arterial spin labeling (ASL):

$$\mathrm{\Delta }S=S_{\mathit{blood}}-S_o\text{exp}\left(\mathit{eTE}\left(\frac{1}{T_1}-\frac{1}{T_2}\right)\right)$$ [2]

where eTE is the duration of T₂-preparation portion of the sequence. T₁ is taken to be 1624 ms in keeping with previous work (7,23).

In the PT₂ method (9), the blood signal for each T₂-preparation is isolated by taking the complex difference (CD) between a pair of projections with equal and opposite velocity-encoding. Here the CD can be defined as (9):

$$\mid CD\mid =2M\left|sin\left(\frac{\mathrm{\pi }V}{2\mathit{VENG}}\right)\right|exp\left(-\frac{\mathit{eTE}}{T_2}\right)$$ [3]

where M depends on pulse sequence parameters (TE, TR, flip angle, and venous blood water spin density) (24), and V represents the blood flow velocity averaged over the SSS lumen. VENC should be equal to the blood flow velocity averaged over the cardiac cycle to minimize the complex difference |CD| dependence on blood flow velocity (9).

Susceptometry-Based Oximetry (SBO)

Susceptometry-based oximetry relies on measurement of the magnetic susceptibility difference Δχ=Δχ_do Hct (1-SvO₂) between the intravascular blood and surrounding tissue (13,14), where Δχ_do=4π (0.27) ppm (25) is the susceptibility difference (in SI units) between fully deoxygenated and fully oxygenated blood. Thus, Δχ_do is a universal constant that is independent of sequence parameters and the subject’s hematocrit level. If the vessel is modeled as a long cylinder of length much greater than the diameter, the induced magnetic field (ΔB) relative to the surrounding tissue can be approximated as (14):

$$\begin{array}{l}\mathrm{\Delta }B\approx \frac{1}{6}\mathrm{\Delta }\mathrm{\chi }{B}_o(3{\text{cos}}^2\mathrm{\theta }-1)\\ =\frac{1}{6}\mathrm{\Delta }{\mathrm{\chi }}_{\text{do}}\mathit{Hct}(1-{\mathit{SvO}}_2)B_o(3{\text{cos}}^2\mathrm{\theta }-1)\end{array}$$ [4]

where θ is the vessel tilt angle relative to the main field (B_o).

Pulse Sequences for SvO₂ Quantification

TRUST

The pulse sequence used was analogous to that described by Xu et al (26). It consists of a nonselective T₂-preparation CPMG pulse train (10,27) (τ_cp=10 ms) composed of 180° composite pulses to impart T₂-weighting. The duration of the CPMG train is defined in terms of an effective echo time (eTE). For each eTE, a single-shot echo planar imaging (EPI) readout was used to scan k-space twice, both with and without tagging of venous blood superior to the imaging slice. Scan parameters are: imaging slice thickness=10 mm, labeling slab thickness=50 mm, gap between imaging slice and labeling slab=25 mm, field of view (FOV)=230 × 230 mm², matrix size=64 × 64 with 5/8 partial Fourier sampling, repetition time (TR)=3000 ms, EPI echo time=7.47 ms, and inversion time (TI) between blood tagging and imaging =1200 ms, yielding a temporal resolution of 24 s. T₂ was computed by fitting the amplitudes obtained at eTEs=0, 40, 80, and 160 ms to an exponential decay and T₂ was converted to SvO₂ based on the calibration curve given in (12).

PT₂

The pulse sequence used has been described by Jain et al (9). The sequence includes a nonselective T₂-preparation CPMG train (τ_cp=10 ms) similar to that used in TRUST. However, only the two central k-space lines with opposite velocity encodings (±VENC) were acquired for each eTE. In distinction to the TRUST sequence, eTE values were corrected for analysis to account for the periods the magnetization is longitudinal as a result of the composite 180° pulses (28). Hence, in lieu of 20, 40, 80, and 160 ms, values of 18.3, 36.6, 73.2, and 146.5 ms were used to compute T₂, matching the values reported for the derivation of the calibration curve (11). Scan parameters are: imaging slice thickness=5 mm, FOV=176 × 176 mm², matrix size=176 × 1, TR=1875 ms, and TE=10.2 ms, yielding a temporal resolution of 15 seconds. Note that the faster temporal resolution relative to TRUST is achieved by use of complex difference processing to isolate the blood signal, eliminating the need to wait over a second (TI=1200 ms) for blood to flow from the tagging slab to the imaging slice. The shorter projection readout TR (compared with that of the EPI readout used in TRUST) also improves temporal resolution slightly. VENC was set to 20 cm/s, close to the average blood flow velocity in the SSS to minimize sensitivity of T₂ to SSS venous blood variation (9).

SBO

The pulse sequence was designed to simultaneously quantify SvO₂ in the SSS and average blood flow velocity in internal carotid and vertebral arteries (15). It is composed of four interleaves that are repeated to collect two flow-compensated gradient echoes differing in TE (interleaves 1 and 3) at the level of the SSS, and a pair of gradient echoes (interleaves 2 and 4) at the same TE, but differing in first gradient moment (0 and ≠0), at the level of the neck. Scan parameters are: imaging slice thickness=5 mm, field of view (FOV)=208 × 208 mm², matrix size=208 × 208, flip angle=25°, TR=35 ms (effective TR at each slice=70 ms), VENC=60 cm/s, and ΔTE=7.04 ms for interleaves 1 and 3, yielding a temporal resolution of 30 s.

MR Imaging Experiments

All imaging was performed at 3 Tesla (T) on a Siemens Tim Trio system with a 12-channel head and neck coil combination. Twelve healthy human subjects (mean age=33±6 years, 3 females and 9 males) participated in the study, which was approved by the Institutional Review Board (IRB) of the University of Pennsylvania. Resting baseline measurements were performed with each of the three techniques in each subject at the level of the SSS. Each examination included 10 successive measurements for each technique, for a total scan time of 11 min. All pulse sequences were programmed using SequenceTree (29). The execution order of the three sequences was permuted among subjects to minimize bias. After completion of the MRI exams, the Hct level was measured for each subject by means of capillary blood sample obtained from the fingertip (Hb 201+, Hemocue, Angelholm, Sweden).

Intrasubject reproducibility measurements were conducted in four healthy males (mean age 35±8 years). Each subject was scanned in three separate sessions, at least one day apart but all within a 2-week period, following the same MRI protocol.

Image Processing and Analysis

TRUST

Difference images were obtained at eTEs of 0, 40, 80, and 160 ms and a region of interest encompassing the SSS with four voxels containing the largest difference signal manually selected from the eTE=0 ms image. The data were then fitted to Eq. [2] to calculate T₂ using unadjusted eTEs (i.e., without accounting for the duration of the 180° composite pulses in the T₂-preparation CPMG train). This is necessary as the calibration equation used for converting T₂ values to SvO₂ (12):

$$R_2=\frac{1}{T_2}=A+B(1-{\mathit{SvO}}_2)+C{(1-{\mathit{SvO}}_2)}^2$$ [5]

was derived using T₂ values calculated with unadjusted eTE values. In Eq. [5], A, B, and C are constants (listed in Table 1) depending on the Hct for each subject and the τ_cp (10 ms in this study).

Table 1.

A, B, and C Parameters of TRUST Calibration Equation (i.e., Eq. [5])

Subject #	A (s−1)	B(s−1)	C(s−1)
1	7.094	0.438	60.96
2	7.332	0.489	61.45
3	7.289	0.479	61.37
4	5.971	0.297	58.17
5	7.067	0.433	60.90
6	7.616	0.594	61.85
7	7.382	0.503	61.55
8	7.616	0.594	61.85
9	7.321	0.487	61.43
10	5.379	0.249	56.58
11	6.982	0.419	60.71
12	5.971	0.297	58.17

PT₂

T₂-weighted projection images were obtained at eTEs of 20, 40, 80, and 160 ms. A region of interest including the four voxels containing the largest difference signal was drawn manually from the eTE=20 ms image. CD images were then fitted to Eq. [3] using eTEs adjusted for the time of the 180° composite pulses (18.3, 36.6, 73.2, and 146.5 ms) (28) to quantify T₂. T₂ values were converted to SvO₂ using an ex vivo calibration curve derived with the adjusted eTE values (11):

$$R_2=\frac{1}{T_2}=\frac{1}{T_{2,\mathrm{o}}}+K{(1-{\mathit{SvO}}_2)}^2$$ [6]

The constant K is a function of α, ω_o, Hct and (τ_cp/2τ) in Eq. [1]. This calibration curve was determined for four Hct levels within the physiologic range. SvO₂ values were interpolated from the calibration curve to match each subject’s Hct.

SBO

A phase difference image was computed from echoes 1 and 3 as $\mathrm{\Delta }{\mathrm{\varphi }}_{\mathit{map}}=arg(Z_1{Z}_3^{\ast })=\mathrm{\gamma }\mathrm{\Delta }B\mathrm{\Delta }TE$, where Z₁ and Z₃ are the complex pixel values of the two echoes and the asterisk indicates the complex conjugate. To reduce the effects of static background field inhomogeneity in the constructed phase image resulting from air–tissue interfaces, a retrospective correction method was implemented by fitting the static field inhomogeneity to a second-order polynomial (30). Subsequently, the phase difference (Δϕ) between intravascular blood and surrounding tissue was used to quantify SvO₂ as:

$${\mathit{SvO}}_2=\left[1-\frac{2\mid \mathrm{\Delta }\mathrm{\varphi }\mid }{\mathrm{\gamma }{\mathrm{\chi }}_{do}\mathrm{\Delta }{\mathit{TEB}}_o({\text{cos}}^2\mathrm{\theta }-1/3)\mathit{Hct}}\right]$$ [7]

Hct was determined from a capillary blood sample for each subject. The tilt angle (θ) of the vessel with respect to the main magnetic field was measured from the scout images.

The superior sagittal sinus ROI was selected based on thresholding of complex difference images, which robustly isolates the vessel signal. A region of interest containing approximately 100 voxels of white and gray matter was drawn in the vicinity of the SSS to minimize effects of small variations in magnetic field (see Figure 1).

FIG. 1.

Magnitude (a) and phase (b) images at the level of SSS. The region of interest of the background tissue is denoted by the black box.

RESULTS

Figure 1 shows an example of (a) magnitude and (b) phase difference images obtained by SBO at the level of the SSS. Figure 2 illustrates the steps involved in calculating T₂-values by means of the PT₂ method. The localizer and projection images, the latter displayed versus eTE, are presented in Figures 2a and b, and Figure 2c shows the signal plotted versus eTE. Similarly, the steps involved to calculate T₂ by means of TRUST are presented in Figure 3. An example of cropped TRUST difference images as a function of eTE is shown in Figure 3a, and Figure 3b shows the difference signals as a function of eTE.

FIG. 2.

Axial magnitude image of the SSS (a), T2-weighted projections at eTE=18.3, 36.6, 73.2, and 146.5 ms (b), and signal versus eTE (c).

FIG. 3.

TRUST difference images at eTE=0, 40, 80, and 160 ms at the level of SSS (a) and signal versus eTE (b).

SvO₂ values quantified with the three techniques (SBO, PT₂, and TRUST) for the twelve study subjects are listed in Table 2. Average SvO₂ values were 65.9±3.3, 65.6±3.5, and 63.2±4.1% for SBO, PT₂, and TRUST, respectively. Two-way analysis of variance treating methods as a fixed and subjects as a random effect showed the difference between the methods to be significant (P<0.005). Thus, a one sample paired t-test was performed post hoc on each of the three pairs of data. PT₂ and SBO did not differ (P=0.7), however, TRUST-derived SvO₂ values were found to be different from those quantified with SBO (P<0.01) and PT₂ (P<0.05). Figure 4 shows correlations of the quantified SvO₂ between the different methods: PT₂ versus SBO (Fig. 4a), TRUST versus PT₂ (Fig. 4b), and TRUST versus SBO (Fig. 4c). Table 3 displays mean and SD of SvO₂ measured three times on separate days for 4 of the 12 subjects with each of the three techniques. Average SvO₂ values were 66.7±2.6, 66.1±3.5, and 64.2±2.0% for SBO, PT₂, and TRUST.

Table 2.

SvO₂ (%, Mean±SD) for 12 Subjects Obtained with SBO, PT₂, and TRUST

SvO2 (%)
Subject #	SBO	PT2	TRUST
1	72.2±2.7	71.0±4.7	71.0±2.4
2	64.5±0.5	64.0±2.4	60.9±3.1
3	60.8±1.5	60.0±3.0	60.5±5.0
4	65.0±2.4	64.5±3.4	59.6±3.9
5	64.4±0.6	62.3±1.8	61.8±3.5
6	66.4±1.7	68.8±2.7	61.9±1.9
7	66.7±1.2	71.3±4.3	67.8±3.8
8	68.5±1.6	68.1±1.7	67.8±2.5
9	69.3±1.5	63.6±2.1	66.1±1.7
10	60.7±1.4	62.6±2.7	60.0±1.4
11	65.1±1.7	66.0±2.0	63.3±2.0
12	67.2±1.7	65.0±4.0	57.3±1.7
Mean±SD	65.9±3.3	65.6±3.5	63.2±4.1

FIG. 4.

Correlations comparing measured SvO₂ (%) between PT₂ versus SBO (a), TRUST versus PT₂ (b), and TRUST versus SBO (c) for all 12 subjects. Error bars represent intrascan standard deviations. The dashed lines denote the 95% confidence interval for the linear fit.

Table 3.

Summary of Intersession SvO₂ (%, Mean±SD) in Four Subjects_a

SvO2 (%)
Subject #	SBO	PT2	TRUST
1	64.7±1.5	65.0±1.0	62.7±2.1
2	64.3±1.5	63.7±1.5	62.3±0.6
3	68.3±2.1	71.3±4.0	65.3±5.8
4	69.5±0.7	64.5±0.7	66.5±0.7
Mean±SD	66.7±2.6	66.1±3.5	64.2±2.0

^aMeans are from three repeated measures on separate days for SBO, PT₂, and TRUST.

DISCUSSION AND CONCLUSIONS

Previous studies have demonstrated the robustness of SBO (15,19), PT₂ (9), and TRUST for SvO₂ quantification (7,20,21). However, these techniques have not been directly compared in a controlled manner in the same cohort of subjects. In this study, we performed a systematic comparison of these three techniques in healthy subjects at resting baseline.

Mean SSS SvO₂ for SBO, PT₂, and TRUST are in good overall agreement with those reported previously (7,9,15). For all three techniques examined in this work, vessel ROI selection was performed in a semi-automated manner, therefore greatly reducing rater dependence. For both TRUST and PT₂, we followed the same selection criteria that were used and described previously (7), selecting the four voxels containing the largest difference signals in the eTE=0 ms image (TRUST) or eTE=20 ms image (PT₂) to estimate the signal decay constant from Eqs. [2] and [3]. For SBO, a region of interest (ROI) that includes approximately 100 voxels of white and gray matter was drawn in the vicinity of the superior sagittal sinus as indicated in Figure 1. In addition, the superior sagittal sinus ROI was selected based on thresholding of complex difference images, which robustly isolates the vessel signal. This semi-automated approach makes the analysis relatively free from rater dependence.

Each measurement was repeated 10 times consecutively in each session resulting in SD in the quantified SvO₂ of less than 2.7%, 4.7%, and 5.0% for SBO, PT₂, and TRUST across all subjects (Table 2). Some of these intrasubject SvO₂ variations are likely physiologic. To test the reproducibility of the methods, the measurements were repeated in triplicate on different days in four of the subjects (Table 3) yielding SDs of 2.6, 3.5, and 2.0% for SBO, PT₂, and TRUST, respectively, suggesting excellent serial reproducibility. Because the reported values for each method in the intersession results were averaged over 10 successive measurements on each day, the data suggest short-term, i.e., intrasession, variability to be larger than longer-term, i.e., intersession, variability. This suggests that the variability due to random measurement noise (which is largely removed by averaging over the 10 repeats and thus does not contribute to the intersession variability) is greater than any true physiologic drift across different days. This indicates that resting state SvO₂ is a stable and reproducible physiologic parameter, and supports the use of averaging over multiple consecutive measurements to improve SvO₂ estimation precision.

The SvO₂ values quantified from the three techniques were highly correlated with each other (R₂=0.51, P<0.01), (R₂=0.48, P<0.05), and (R₂=0.50, P<0.05) (Figs. 4a–c). Furthermore, for each of the three correlation plots, the 95% confidence intervals for the slope of the regression line includes the line of identity. This suggests that the bias between the various methods does not have a statistically significant SvO₂ dependence over the range of SvO₂ values measured. A small discrepancy in mean SvO₂ measured with the two T₂-based methods was observed. TRUST yielded an average SvO₂ value lower by 2.4% relative to PT₂ which was statistically significant (P=0.019). One potential source of this discrepancy may be the longer echo time used in the EPI readout of our TRUST sequence relative to more recent implementations by Xu et al (26). In that work, longer echo time was shown to lead to an overestimation of R₂, and thus underestimation of SvO₂, compared with a shorter (3.6 ms) echo time achieved with parallel imaging (26). The authors of that study found a 2 s⁻¹ overestimation of R₂ for a 7.0 ms versus 3.6 ms TE, corresponding to approximately 3% underestimation of SvO₂. This could explain our observed lower SvO₂ in TRUST relative to PT₂. Another potential source of the discrepancy could be differences in the readout (EPI in TRUST versus projection in PT₂) or blood isolation scheme (tag/control subtraction in TRUST versus phase contrast complex difference in PT₂) used in the two techniques. Both these factors are expected to influence signal signal to noise ratio, especially at the longer eTEs, which could create a bias in the calculated R₂. This potential effect could be further investigated by directly comparing R₂ quantified from sequences with different readouts (EPI versus projection) or blood isolation schemes (tag/control versus complex difference), but which are otherwise identical.

Another difference between the two T₂-based methods is that corrected eTEs were used for the PT₂ calibration curve used in (9), in contrast to uncorrected ones in the TRUST work by Lu and Ge (7). While substitution of signal values from one calibration curve to another results in a less than 1% difference in quantified SvO₂ values, use of the correct eTE values (T₁ corrected for PT₂, uncorrected for TRUST) is more important, as calculating SvO₂ with corrected eTE values in TRUST results in a 3% reduction in quantified SvO₂, which would have suggested a 6% underestimation of SvO₂ in TRUST relative to PT₂, erroneously suggesting a much larger discrepancy between the two T₂-based methods in our study than was actually observed. This demonstrates the fact that T₂ calibration curves are specific to a particular method and not interchangeable. The need for a sequence specific calibration curve highlights one disadvantage of T₂-based SvO₂ quantification methods in comparison to SBO, as discussed further below.

Each of the three methods has distinct advantages and disadvantages for quantification of SvO₂. SBO requires a suitable adjacent reference tissue and correction for the vessel tilt angle (θ) with respect to the external magnetic field (B_o) (14). It is also sensitive to the induced magnetic field inhomogeneity at the air–tissue interface or between adjacent tissues types (30,31). Therefore, using SBO to quantify oxygenation level in the jugular vein is challenging due to the presence of oropharynx and its proximity to the trachea, which causes severe static field inhomogeneity. In addition, vessels are modeled as long cylinders. While this approximation has been shown to cause only small errors (17), it limits the method’s applicability to certain vessels such as jugular vein, femoral vein, and superior sagittal sinus (SSS).

Unlike SBO, T₂-based methods are not limited by vessel orientation, the need for a suitable reference tissue, or the vessel’s proximity to sources of background field gradient such as the trachea. For example, PT₂ and TRUST have been successfully applied to SvO₂ quantification along the internal jugular vein (8,9). However, one major limitation of the T₂ methods is that a calibration curve must be separately derived for each field strength, pulse sequence, and Hct, whereas these parameters are explicitly included in the SBO model. Factors such as the freshness of the blood, different species of the blood, and experimental temperatures influence the measured T_2. For example, long-term storage of blood sample causes the formation of paramagnetic methemoglobin, which cannot bind oxygen. Additionally, variation in the erythrocyte size for blood samples from different species (e.g. human, rodent, or bovine) causes alterations in the erythrocyte permeability (32), which influences the quantified SvO₂.

Though based on the same principle, the PT₂ and TRUST methods have specific advantages and limitations. Because the complex difference signal of flowing blood is modulated by flow velocity, PT₂ can only be applied in situations where the blood flow is relatively nonpulsatile and not changing over time. Therefore, it is not suitable to detect changes in SSS oxygenation in response to dynamic stimuli such as apnea during which flow changes occur over seconds (19). Another potential limitation of this method is that the vessel is spatially resolved in one dimension, requiring appropriate selection for the FOV to avoid vessel overlap. In contrast, TRUST uses a 2D EPI readout and therefore is not susceptible to vessel overlap as in PT₂.

SBO has the added benefit of allowing for simultaneous quantification of total cerebral blood flow with relatively high temporal resolution for quantifying CMRO₂. For example, in some of the authors’ recent work, CMRO₂ was measured at 30 s temporal resolution in response to hypercapnia (15) and three seconds temporal resolution in response to apnea (19). High temporal resolution SBO has been applied outside the brain as well. A SBO sequence with 1.25 second temporal resolution SvO₂ quantification and simultaneous 120 ms temporal resolution projection-based flow quantification was implemented to evaluate the vascular response of femoral/popliteal vessels during a cuff-induced ischemia paradigm (16). Thus, SBO is well suited to study the temporal variations in SvO₂ and CMRO₂ in response to a variety of physiologic challenges in different organ systems.

In conclusion, we performed a systematic comparison of susceptometry-based oximetry (SBO) and two T₂-based methods (projection based T₂ (PT₂) and TRUST) for quantification SvO₂ in the superior sagittal sinus (SSS) at resting baseline. The results indicate good agreement between the SBO and PT₂ with average SvO₂ values 65.9±3.3 and 65.6±3.5, respectively, while TRUST showed a mean SvO₂ of 63.2±4.1% which is lower by less than 3% compared with SBO and PT₂. Choice of an optimal method should be based on application specific considerations, such as the availability of an appropriate calibration curve in T₂ methods, the target vessel of interest, and the desired temporal resolution.