Estimating Blood Oxygen Saturation with T2 maps: Evaluation in Pulmonary Hypertension

Background: T2 mapping is an important tool for myocardial tissue characterization. Although not utilized clinically, it is well established that blood pool T2 values reflect blood oxygen saturation (SbO₂) due to the paramagnetic effects of deoxygenated hemoglobin [1], and this relationship has been explored in various patient populations [2]. Here, we study patients with pulmonary hypertension (PH), for whom right-heart catheterization (RHC) is available to measure reference-standard oxygen saturations among other hemodynamics, with the purpose of building a framework for estimating actual right ventricular (RV) SbO₂ using only a conventional T2 map.

Methods: Patients and imaging: This retrospective study included 31 consecutively imaged PH patients with RHC within one month of MRI, which included T2 mapping. Imaging was performed on a 1.5T Siemens scanner, and all patients were enrolled in an IRB-approved study. Mid-ventricular short-axis images were analyzed to measure RV and LV blood pool T2 values (Figure 1a), using the commercial T2 mapping method with 2D bSSFP T2prep (TEs of 12, 25, and 50ms). Circular regions of interest at least 1.2 cm² in size, were placed in papillary and artifact-free areas of the blood pool T2 maps, by a blinded observer.
SbO₂ estimation: We developed a framework (Figure 1) similar to that presented in (4) for transforming blood T2s into SbO_2-CALC using published values for model constants (3) (see Figure 1 and caption). This framework accounts for the fact that clinical T2 mapping uses two TE times (besides TE=0) and thus two tau₁₈₀ exchange times, which means the effective blood T2 is different in the two contrasts generated, perturbing the T2 decay curve.

Results: RV blood-pool T2 showed a strong correlation with RHC-derived right atrial (RA) SbO₂ (r=0.80, p< 0.001), as was RV/LV T2 ratio (r=0.79 p< 0.001) (Figure 2a). When RV blood-pool T2 was converted to SbO_2CALC as described in Figure 1, there was a strong quantitative agreement between SbO_2CALC and RHC-derived RA SbO₂ (r=0.82, p< 0.001, Figure 2b), with Bland-Altman limits of agreement =4.5%±11%. RA Sb0₂ was a moderate to strong correlate of key hemodynamic parameters of PH severity, e.g. mean pulmonary artery pressure (mPAP) (r=-0.66,p< 0.001) and pulmonary vascular resistance (PVR) (r=-0.72,p< 0.001), illustrating its value in PH evaluation. Importantly, MR-derived SbO_2CALC was similarly a correlate of both mPAP (r=0.64,p< 0.001) and PVR (R=-0.71, p< 0.001). Even T2 ratio was similarly correlated (r=-0.65 for MPAP, r=-0.73 for PVR).

Conclusion: In this PH cohort, blood-pool T2 values on conventional T2-mapping were transformed to provide quantitative estimations of right-heart SbO₂. We found that key RHC hemodynamic parameters, PVR and mPAP, were strongly correlated to MR-derived SbO_2-CALC., highlighting this measurement's importance. Potentially, the estimated SbO₂ will be a more robust site-independent MRI metric, vs. RV T2 or T2 ratio, for characterizing PH among other cardiovascular diseases.

Figure 1: Framework for calculating RV blood oxygen saturation (Sb02-CALC) from RV blood T2 values obtained by T2 mapping, using (a) Meiboom-Luz Equation (1), and previously published values (3) for constants. (b) Inputs are the LV blood T2 (assumed to fully oxygenated blood, T20) and RV blood T2. Also note that since the T2 depends on exchange rate and exchange time (e.g. τ_ex and τ_180), each T2prep module (c) with a different TE uses a different τ_180 and thus the T2 changes (d), and this is also included in the modeling. Subjecting the resulting signal vs. TE to an exponential fit (e) leads to a look up table (f), relating measured RV T2 (T2meas) to Sb02-CALC. It is subject specific since each subject has a different T20. Constants used were τ_ex=7.5ms, and αω0=(0.39ppm) ∙(2∙pi ∙1.5T ∙42.58 MHz/T) in rad/sec. Hct=0.46 (assumed since T2 is very insensitive to the factor (Hct)(1-Hct)), τ_180=12ms and 25ms for TEs of 25 and 50ms.
Figure 2: A) Blood T2 ratio is strongly correlated with measured oxygen saturations. B) Using framework with previously published constants, LV and RV T2 values were used to estimate oxygen saturations in the right heart, showing excellent absolute agreement. The Bland-Altman limits of agreement were 4.5%±11%.