Synopsis

Keywords: Oxygenation, Oxygenation

TRUST and T2-TRIR can be used to measure venous blood oxygenation (Y_v) and oxygen extraction fraction (OEF). T2-TRIR also offers simultaneous blood T1 (T1_b) measurements for hematocrit calculation in lieu of invasive blood samples. Here, TRUST and T2-TRIR were compared across a broad Y_v range using hypoxic and hypercapnic challenges in healthy volunteers, while T2-TRIR-derived hematocrit estimates were compared with venipuncture results. T2-TRIR- and TRUST-derived Y_v and OEF values exhibited a small constant bias (-5.4±3.5% and 6.0±4.0% respectively, both p<0.01) across all stimuli. Hematocrit by venipuncture and from T1_b were unbiased but had broad confidence intervals (1.7±6.3%, p=0.20).

Introduction

Venous blood oxygenation (Y_v) can be derived from venous blood T2 (T2_b). Y_v measurements enable the estimation of oxygen-extraction fraction (OEF) and, combined with cerebral blood flow, the cerebral metabolic rate of oxygen (CMRO₂).¹ These biomarkers reflect cerebral oxygen metabolism and are thought to be indicative of tissue viability. T2-Prepared-Blood-Relaxation-Imaging-with-Inversion-Recovery (T2-TRIR) was introduced as an alternative to T2-Relaxation-Under-Spin-Tagging (TRUST) to simultaneously measure T2_b and T1_b.² T1_b enables hematocrit (Hct) estimation in lieu of invasive blood samples. We previously demonstrated that TRUST and T2-TRIR were equally reproducible, but T2-TRIR-derived T2_b values were significantly higher compared to TRUST.³ Here, we further investigated this bias over a broader O₂ saturation range using hypoxic and hypercapnic gas challenges in healthy volunteers and validated T2-TRIR-derived Hct estimates versus venipuncture.

Methods

Twelve healthy volunteers (4F:8M, age 36±10 years) were scanned on a 3T Philips Achieva system. The scan protocol included five respiratory phases: baseline (room air), mild hypoxia, severe hypoxia, hypercapnia, and recovery (room air) which were performed using a computer-controlled gas blender (RespirAct, Thornhill Research, Toronto, Canada). During mild and severe hypoxia, end-tidal pO₂ (EtO₂) was targeted at 55 mmHg and 40 mmHg respectively. During hypercapnia, baseline end-tidal pCO₂ (EtCO₂) was increased by 10 mmHg. Fingertip pulse oximetry (S_pO₂) and cerebral tissue oxygenation index (TOI) were recorded during the whole measurement. During each phase, TRUST and T2-TRIR scans at the superior sagittal sinus and a phase contrast (PC) scan above the bifurcation of the carotid arteries were acquired. TRUST and T2-TRIR scans were acquired as reported previously using effective echo times of 0, 40, 80, and 160ms.³ Complete blood count was measured by venipuncture.

For TRUST data, label and control images were subtracted and the difference signal was fitted to obtain T2_b.⁴ For T2-TRIR data, a dedicated voxel-selection strategy was used, as described previously.³ T1_b and T2_b were simultaneously fitted using the following model:

$$S=M_0\times\left[(1-(e^{-\frac{eTE}{T2_b}}\times IE))\times e^{-\frac{TI}{T1_b}}\right]$$
From the TRUST- and T2-TRIR-derived T2_b values, Y_v was calculated using an empirically derived model⁵:

$$R_2=77.5\times Hct\times(1-Yv)^{2}+27.8\times(1-Yv)^{2}+6.95\times Hct+2.34$$
For Y_v calculations, first, Hct from the venipuncture (Hct_VP) was used for TRUST and T2-TRIR. Additionally, Hct was estimated from T2-TRIR-derived T1_b (Hct_TRIR) using a model by Li et al (2016)⁶ and used to calculate Y_v and OEF from T2-TRIR-derived T2_b. OEF was calculated as follows:

$$OEF=\frac{Y_a-Y_v}{Y_a}\times 100\%$$
In which Y_a is the arterial oxygenation measured by the pulse oximeter.

Hct_VP and Hct_TRIR were compared but only for baseline and recovery scans because the model for Hct_TRIR requires estimation of Y_v.⁶ At baseline and recovery, Y_v was estimated at 65%.⁶ For the other stimuli, Y_v could not be estimated and using Y_v from the TRUST or T2-TRIR measurements would indirectly use Hct_VP.

Flow-weighted mean velocity was calculated from the PC scans. Bland-Altman analysis and repeated measurements ANOVA or Friedman test in case of non-normality were performed. Post-hoc comparisons with baseline values included Bonferroni multiple comparison correction.

Results

Mean EtO₂ and EtCO₂ are shown in Figure 1. S_pO₂, TOI, and blood flow were significantly affected by the respiratory challenges (Figure 2A-C). For one volunteer, S_pO₂ reduction was smaller presumably due to gas leakage caused by facial hair.

During mild hypoxia, T2_b, Y_v, and OEF were lower compared to baseline but this was only significant for Y_v measured by TRUST (p<0.01; Figure 3). During severe hypoxia, T2_b, Y_v, and OEF further decreased whereas during hypercapnia, T2_b and Y_v increased and OEF decreased compared to baseline. T2_b, Y_v, and OEF values were similar at baseline and recovery.

A significant bias was observed between TRUST- and T2-TRIR-derived T2_b, Y_v, and OEF values (all p<0.01; Figure 4). For Y_v and OEF this bias was constant across the range of measured values, indicated by the slopes of 1.0 (confidence interval: 0.92-1.10 and 0.89-1.13).

T1_b was significantly lower during severe hypoxia and hypercapnia compared to baseline (p=0.02 and p<0.01; Figure 5A). While no significant bias was found between Hct_VP and Hct_TRIR (1.7±6.3%, p=0.20; Figure 5B), the observed correlation between these Hct values was only weak (r²=0.17, p=0.04; Figure 5C). Using Hct_TRIR to calculate T2-TRIR-derived Y_v and OEF, the bias between the two techniques to estimate Y_v and OEF became more variable (Figure 5D+E).

Discussion

TRUST and T2-TRIR can be used interchangeably for Y_v and OEF estimation as long as the fixed bias is corrected. The observed variance of the differences between the two techniques is comparable to the previously reported repeated measures variance of either one.³ However, when replacing venipunctures with T2-TRIR-derived Hct, the uncertainty between the techniques becomes larger, especially for OEF. The source of uncertainty in Hct_TRIR estimates remains unknown. The reduction of T1_b with hypoxia was expected because of the shorter T1 of deoxygenated hemoglobin.⁶ Surprisingly, T1_b reduced further during hypercapnia which could be caused by a pH effect⁷ or flow-related artefacts.

Conclusion

In conclusion, T2-TRIR- and TRUST-derived Y_v and OEF values can be compared because of their constant bias across a wide range of values. Using T2-TRIR, Y_v and OEF can be estimated without a venipuncture but the uncertainty grows noticeably. Additionally, Hct can be estimated from T2-TRIR-derived T1_b values but further validation with blood samples is still needed.

Acknowledgements

This work was supported by the National Heart, Lung, and Blood Institute (1R01-HL136484-01A1).