Blood Oxygen Saturation in Great Vessels: Measurement by Magnetic Resonance BOLD Sequence—Accuracy, Challenges, and Potential Solutions

Background: Blood oxygen level-dependent (BOLD) magnetic resonance (MR) sequences are capable of measuring oxygen saturation (SO₂) but are rarely used in cardiovascular imaging due to significant technical and physiological challenges. Noninvasive assessment of SO₂ via MR could improve hemodynamic evaluation.

Methods: Sixty patients (age 63.5–1.8 years) with congenital heart disease underwent MR studies with BOLD sequences between May 2023 and August 2024. In 16/60 patients, cardiac catheterization was performed within 72 hours before or after the BOLD measurement. In these 16 patients, SO₂ measurements were performed using the BOLD sequence in 77 vessels or chambers, including caval veins, aorta, pulmonary arteries, and both ventricles. BOLD measurements were conducted on a clinical 3T MR scanner using a strongly T2-prepared (TE = 100 ms) single-shot bFFE sequence (“label”), and a “control” image without T2 preparation. A preceding, conventional time-resolved flow measurement (2D qFlow) was used to identify the optimal cardiac phase with minimum flow. SO₂ quantification was achieved via a piecewise linear approximation of the Luz–Meiboom formula. SO₂ values were derived from the signal intensity ratio of the aorta and respective vessel. Aortic SO₂ values were derived from a routine pulse oximeter.

Results: Oxygen saturation measured by BOLD in different great vessels significantly correlated with saturations obtained during cardiac catheterization within 72 hours of the MR study (p < 0.0001; r ² = 0.61, y = 0.8x + 16.8). Bland–Altman analysis revealed a bias of 0.3% with confidence limits of ±17%. Interobserver variability between two observers for 20/77 measurements was 2.4%, with confidence limits of −7% to +11% in Bland–Altman analysis. The greatest source of error in BOLD measurement was bias due to flow incoherence within the vessel area. The best agreement with cardiac catheter data was observed during BOLD measurements in late diastole, although in some vessels, trigger intervals had to be adjusted using simultaneous flow images. No measurements were feasible near metal artifacts, as they distorted field homogeneity, even when not obvious in magnitude images.

Conclusion: Noninvasive oxygen saturation in the great vessels can be measured using MR BOLD sequences. However, significant bias due to flow disturbances persists, which necessitates individually adapted trigger intervals to improve accuracy.