Venous oxygen saturation estimation from multiple T2 maps with varying inter-echo spacing

Background

Dependence of blood T2 on O2 saturation has led to non-invasive MRI-based techniques for determining venous O2 saturation (SvO₂) [1-3]. However, applying a general calibration factor derived from in vitro experiments can lead to inaccurate and largely varying SvO₂ estimates in the target population. We aim to show that based on the Luz-Meiboom relation 1/T2 = 1/T2o + Hct(1-Hct) τ_ex[(1-%SO₂/100)αω₀]²(1-2*τ_ex/τ₁₈₀ tanh(τ₁₈₀/2*τ_ex))[4], individual SvO₂ can be determined from multiple T2 maps, each acquired at a specific inter-echo spacing (τ₁₈₀).

Methods

Three T2 prepared SSFP quantitative T2 maps (τ₁₈₀=10, 12 and 15 ms, TR >3000 ms, FA = 40⁰, 2.8 × 2.8 × 10 mm³, 2NEX, free breathing) were acquired in seven volunteers (age: 32.6 ± 12.2 yrs) on a 3T MRI system (Tim Trio, Siemens Healthcare). T2 preparation involved an MLEV refocusing pulse train with 2, 4, 8 and 12 composite pulses. Venous and arterial blood T2 were measured in each map in an ROI in the ventricles. For each subject, the multiple τ₁₈₀ measurements were processed jointly to estimate SvO₂ along with other nuisance parameters (T2o and τ_ex) via constrained nonlinear least squares fitting in Matlab (The Mathworks Inc, Natick, MA, USA). The values of hematocrit (Hct), arterial O2 saturation (SaO₂), and α were fixed at 41%, 97%, and 0.4 ppm respectively.

Results

Figure 1 shows T2 maps acquired in a volunteer. Table 1 shows mean ± SD of venous and arterial T2 and estimated parameters. Venous and arterial blood T2 agree with previously reported values at 3T [2,3]. Estimated SvO₂ for all volunteers falls within the normal physiological range (60-80%).

Figure 1

T2 prepared SSFP quantitative T2 maps (four-chamber view) acquired in a volunteer at different τ ₁₈₀ (10, 12 and 15 ms).

Table 1

Conclusions

The measured T2 of blood is dependent on Hct, O2 saturation, and τ₁₈₀. We have shown that if these parameters are known, SvO₂ can be non-invasively determined from arterial and venous blood T2 maps acquired at multiple τ₁₈₀. This provides in-vivo, patient-specific calibration, and may reduce the uncertainty and error in SvO2 estimation from applying a general calibration factor to the entire patient population. Although in this preliminary study we have assumed a fixed Hct and SaO₂ for all subjects, greater accuracy may be achieved by measuring individual Hct and SaO₂ from a blood sample and a pulse oximeter respectively. Nevertheless, our results show that the proposed approach is feasible, giving reasonable SvO₂ estimates. Future studies will involve validation of the single assumed parameter, α, in the T2-SO₂ model, optimization of the set of τ₁₈₀ times, and further evaluation of the accuracy and precision of T2 mapping in determining blood O2 saturation. This would lead to the development of rapid, accurate, non-invasive, in-vivo quantification of SvO₂ from T2 maps, which would be highly beneficial for heart failure and congenital heart disease patients.