T2 Relaxation of Human Blood  at 3T Revisited:  In Vivo and In Vitro Meaurement using TRUST MRI
Adam Bush1, Jon Detterich2, Thomas Coates3, Herbert Meiselman4, and John Wood1

1Biomedical Engineering/ Cardiology, University of Southern California/ Children's Hospital Los Angeles, Los Angeles, CA, United States, 2Cardiology, Children's Hospital Los Angeles, Los Angeles, CA, United States, 3Hematology, Children's Hospital Los Angeles, Los Angeles, CA, United States, 4Physiology and Biophysics, University of Southern California, Keck School of Medicine, Los Angeles, CA, United States

Synopsis

Precise knowledge of the T2 of blood (T2b) is required for spin-echo based blood oxygenation determination methods such as TRUST (T2 Relaxation Under Spin Tagging). In this study we measure the T2b in vivo and vitro using TRUST MRI. After correcting for physiologic variable we found that our model of T2b is statistically significantly different from the models used by other groups. We conclude those model lead to errors in derived parameters including oxygen saturation, oxygen extraction fraction and cerebral metabolic rate.

Introduction

Precise knowledge of the $$$T_{2}$$$ of blood ($$$T_{2b}$$$) is required for spin-echo based blood oxygenation determination methods such as TRUST (T2 Relaxation Under Spin Tagging). Several groups have analyzed the $$$T_{2b}$$$ but previous studies have been limited by non-physiologic measurement conditions including the use of nonhuman blood [3] and the lack of calibration for lower HCT commonly found in congenital anemia syndromes. In this study we characterized the $$$T_{2b}$$$ dependence on hematocrit (HCT) and oxygen saturation in human blood at 3T over the physiologic range of hematocrit and blood oxygenation. For validation, we also measured sagittal sinus saturation in 35 normal controls and 12 patients with chronic anemia.

Methods

This study was performed under an IRB approved protocol with informed consent/assent. Imaging was performed on a Philips 3T Achieva scanner with 8 channel head coil. T2b calibration was performed using the blood of three subjects (1F, 33+15.1 years). Blood samples were centrifuged and separated into packed red cells and plasma. Suspensions were recombined to create blood samples with HCT ranging from 13%-52%. A custom, 37o C temperature controlled blood imaging reservoir was used for in vitro imaging measurements. Blood was deoxygenated external to the scanner in a 37o C temperature controlled glove box using 5% CO2 /95% N2 gas mixture that maintained the pH at approximately 7.4. Upon desaturation, blood was transferred to the imaging reservoir and agitated every 2-3 minutes to minimize sedimentation. Saturation was measured using a hemoximeter (OSM3 Radiometer). Following localization, TRUST was measured using a pulse sequence designed by Lu and colleagues [4]. Arterial spin labeling was disabled. For any given saturation, 1/T2b was well described by the relationship: $$\frac{1}{T_{2b}}={R_{2}}=\frac{1}{T_{2o}}+K(1-Y)^{2}$$ where T2o is the T2 of oxygenated fully saturated blood, Y is the oxygen saturation and K is a constant dependent on several parameters including inter-echo spacing (τ) and hematocrit. The joint dependency of 1/T2b, hematocrit and oxygen saturation was well described by a hyperplane: $$\frac{1}{T_{2b}}=A*HCT+B*HCT*(1-Y)^{2}+C*(1-Y)^{2}+D$$ where A,B,C,D are empirical fitting coefficients. Two dimensional least squares was used to find A,B,C and D. For clinical correlation of the newly derived calibration curve, TRUST imaging was performed in 35 healthy controls (CTL) and 13 patients with chronic anemia (ACTL). The T2prep duration was 0, 40, 80 and160 with τ =10ms and standard parameters [4]. T2 value was derived after mono-exponential fitting of the difference of control and tagged images. For clinical correlation of the newly derived calibration curve, TRUST imaging was performed in 35 healthy controls (CTL) and 13 patients with chronic anemia (ACTL). The T2prep duration was 0, 40, 80 and160 with τ =10ms and standard parameters [4]. T2 value was derived after mono-exponential fitting of the difference of control and tagged images.

Results

One hundred and twenty six blood measurements of HCT, oxygen saturations and 1/T2b were used for model fitting. 1/T2b rose linearly with the square of oxygen extraction (Figure 1A). The slope of this relationship steepened with increasing hematocrit. Figure 1B demonstrates an approximately linear relationship between K and HCT, K=.0566 *HCT+ 2.87, R2= .907. Model parameters where found to be A=8.13, B=121, C=5.46 and D=2.46, R2= 0.977. Our model predicted a mean saturation of 0.592 and 0.634 in the CTL vs ACTL whereas the Lu bovine model [3] predicted mean saturation of 0.634 and 0.629 respectively. Bland Altman analysis between saturation predicted by our model and Lu model is shown in Figure 1D. The agreement is highly nonlinear with hematocrit. The Lu calibration saturation estimates an average of 5% higher absolute saturation. However, there is linear decrease in the bias for HCTs less than 35%. Figure 2 compares the sagittal sinus saturation estimates calculated using both techniques in CTRL and ACTL.

Discussion

We report the first description of changes in 1/T2b of human blood across the entire physiological range of hematocrits and oxygen saturations. By controlling for physiological variables such as temperature, spontaneous blood sedimentation and pH, we are confident our data represents the most accurate surrogate physiological account of human T2b blood to date. Our model well characterized the relationships between 1/T2b , hematocrit and oxygen saturation . These data suggest that human T2b is longer than that of bovine blood (3), leading to larger predicted AVO2 consistent with invasive measurements. Importantly, our data allows accurate TRUST oximetry in patients having hematocrits less than 35%, commonly observed in patients with chronic anemias. Care must be taken in the measurement and quantification of SvO2 for even deceivingly small errors is saturation will lead to much larger errors in oxygen extraction fraction and cerebral metabolic rate.

Acknowledgements

Special thanks to Jon Chia at Philips Healthcare for technical support and pulse sequence implementation on our Philips 3T Achieva scanner.

This work was supported by the CTSI Translational Science, CTSI Neuropsychology core, and National Heart Lung and Blood Institute 1U01HL117718-01.

References

1. Luz Z,Meiboom S. Nuclear Magnetic Resonance Study of the Protolysis of Trimethylammonium ion in Aquesous Solution: Order of the Reaction with Respect to the Solution: Order of the Reaction with Respect to the Solvent. J Chem Phys. 1963

2. Wright et al. Estimating Oxygen Saturation of blood in vivo with MR imaging at 1.5T. JMRI. 1991

3. Lu et al. Calibration and validation of TRUST MRI for Estimation of Cerebral Blood Oxygenation. MRM 2012

4. Lu H, Ge Y. Quantitative Evaulation of Oxygenation in Venous Vessels using T2 Relaxation Under Spin Tagging MRI. MRM. 2008

Figures

1/T2b demonstrates a (1-Y)^{2} dependent that is also related to HCT level.

Highly linear relationship between modeling constant, K and HCT suggesting additional HCT dependent effect.

Measured vs Predicted 1/T2b. Empirical human blood model is statistically different from bovine blood model.

Difference between human blood model and bovine blood model derived saturation shows nonlinear bias with respect to HCT.

Average saturation derived in healthy controls and patients with chronic anemia using our model and bovine model after TRUST MRI of the sagittal sinus. Bovine model leads to overestimates of saturation in healthy controls.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1484