MRI Maps Heart Blood Flow Without Contrast Dye in Young Heart Patients

For children and young adults living with congenital heart disease, the cardiac MRI exam has long been a familiar ritual: an intravenous line, a syringe of gadolinium-based contrast agent, and a series of scans that can stretch across an hour or more. A new prospective study published in Pediatric Radiology now offers strong evidence that one of the exam’s most powerful components, four-dimensional flow imaging, can deliver clinically reliable measurements without any contrast at all, and on the widely used 1.5 Tesla scanners found in most pediatric hospitals. The finding could reshape how imaging centers approach hemodynamic assessment in this growing population of patients.

Four-dimensional flow MRI is a technique that captures blood flow in three dimensions across the entire cardiac cycle, producing a volumetric movie of moving blood that can be interrogated after the scan in any plane a clinician chooses. Its two-dimensional predecessor, phase-contrast MRI, remains the clinical reference standard, but it requires operators to prescribe a separate imaging plane for each vessel of interest, one slice at a time. The 4D approach acquires the whole volume at once, allowing retrospective measurement of net flow, forward flow, regurgitant fraction, and peak velocity in the aorta, pulmonary arteries, and venous vessels from a single acquisition. Until now, however, most validation work has relied on contrast-enhanced imaging, which boosts the signal from blood and improves the fidelity of velocity measurements.

The research team, led by investigators at Seattle Children’s Hospital with collaborators at Lurie Children’s, Lucile Packard Children’s, and Texas Children’s Hospitals, set out to test whether non-contrast 4D flow could hold its own at 1.5 Tesla, the field strength most commonly used in pediatric cardiac imaging. They prospectively enrolled thirty patients aged eight to twenty years with suspected or confirmed congenital heart disease, a diagnostically diverse group that included nine patients with repaired tetralogy of Fallot, six with bicuspid or dysplastic aortic valves, four with pulmonary hypertension, and four with atrial septal defects, among other conditions. All scans were performed awake, without sedation, and any 4D flow and 2D phase-contrast acquisitions were completed before contrast was given for the clinical portion of the exam.

The technical comparison was rigorous. Two-dimensional phase-contrast images were acquired in the proximal ascending aorta and main pulmonary artery, with branch pulmonary arteries added when clinically indicated, using free-breathing acquisition with three signal averages and a mean scan time of roughly 2.7 minutes per vessel. The non-contrast 4D flow sequence used standard Cartesian sampling with a respiratory navigator, a voxel size of 2.4 to 2.5 millimeters, and a mean acquisition time of 16.2 minutes. Velocity encoding was tailored to each patient based on recent echocardiographic velocities, ranging from 150 to 400 centimeters per second. Post-processing was performed in cvi42, with background correction and anti-aliasing applied by a single experienced operator before flow planes were positioned to match the 2D acquisitions.

The headline result is a strong correlation between the two techniques. Accounting for repeated measurements within patients and across five vessels, non-contrast 4D flow correlated with 2D phase-contrast at r equals 0.76 for net flow, r equals 0.9 for peak velocity, and r equals 0.75 for regurgitant fraction, all highly statistically significant. Bland-Altman analysis revealed systematic but modest biases: net flow ran about 10.1 percent lower on the non-contrast technique, while peak velocity ran about 18.9 percent higher. Notably, the higher peak velocities may not represent an error at all. Because 4D flow interrogates the full volume rather than a single plane, it can capture velocity components that a planar 2D acquisition misses, and prior work has shown 4D flow agrees better with Doppler echocardiography, the modality that actually guides clinical decisions about stenosis severity.

To anchor the flow data against an independent physiological benchmark, the team compared forward flow in the great vessels with ventricular stroke volumes measured from standard cine imaging. Ascending aortic flow from 4D was on average 18.1 percent lower than left ventricular stroke volume, and main pulmonary artery flow was 13.9 percent lower than right ventricular stroke volume. Yet the reference technique showed the same directional shortfall, with 2D phase-contrast also underestimating stroke volume by roughly 6 to 8 percent, differences that did not reach statistical significance. This pattern suggests that part of the apparent gap reflects known limitations of planar flow measurement itself, including through-plane motion and spatial resolution, rather than a unique weakness of the contrast-free method.

Perhaps the most clinically meaningful findings concern internal consistency and image quality. In a subset of seven patients with branch pulmonary artery data, the sum of right and left branch flows correlated with main pulmonary artery flow at r equals 0.96 for the non-contrast technique, actually exceeding the r equals 0.87 achieved by 2D phase-contrast. Signal-to-noise and contrast-to-noise ratios were statistically indistinguishable between the two methods, even though the raw relative signal intensity was markedly lower for 4D flow, a consequence of acquiring a single large slab without the inflow of fresh unsaturated blood that thinner-slab techniques enjoy. Vessel dimensions measured from the magnitude images correlated at r equals 0.91, confirming that vessel borders remained sharp enough for reliable contouring.

Reproducibility held up as well. In a blinded interobserver analysis of ten patients covering thirty-six discrete vessel measurements, a second reader with only one year of experience produced a mean bias of just minus 5.2 percent relative to the original observer, with 91.7 percent of measurements falling within the limits of agreement and an intraclass correlation coefficient of 0.95. The authors are candid about the technique’s limits, however: the limits of agreement between modalities remain wide, at roughly plus or minus 45 to 51 percent, and one patient with exceptionally high velocity encoding showed substantial flow underestimation. Contrast-enhanced imaging likely retains the edge for very small vessels, obesity, complex shunts, metallic artifact, and advanced parameters such as wall shear stress.

The broader context makes the result timely. Gadolinium deposition in the brain, the logistical burden of intravenous access in children, rare contrast reactions, and growing concern over gadolinium’s ecological footprint in water systems have all fueled interest in contrast-free cardiac MRI. Meanwhile, accelerating technologies, including compressed sensing, non-Cartesian sampling, view sharing, and deep learning reconstruction, are steadily shrinking 4D flow scan times and improving resolution, even opening the door to imaging unsedated neonates and, eventually, fetal circulation. The 16-minute acquisition used here is comparable to the total time of multiple 2D acquisitions in a complex patient, and the technique is already a standard product sequence on many modern platforms.

For the populations that dominate congenital heart follow-up care, patients with repaired tetralogy of Fallot facing pulmonary valve decisions, those with bicuspid aortic valves, and those with pulmonary hypertension, the study suggests that a fully non-contrast hemodynamic assessment is now within reach at 1.5 Tesla. Prior work has shown that 4D flow yields similar clinical decisions to 2D phase-contrast in repaired tetralogy patients despite similar percent differences, and the authors argue that internal consistency across a dataset may matter more for clinical interpretation than exact agreement with any single reference measurement. With parallel advances in contrast-free tissue characterization such as MR fingerprinting and virtual native enhancement, the prospect of a complete cardiac MRI without a single drop of gadolinium is moving from aspiration toward routine practice.

Subject of Research: Validation of non-contrast 4D flow cardiac MRI against 2D phase-contrast imaging in pediatric and young adult congenital heart disease at 1.5 Tesla

Article Title: Establishing the reliability of non-contrast 4D flow in congenital heart disease at 1.5T

Article References: Establishing the reliability of non-contrast 4D flow in congenital heart disease at 1.5T. (n.d.). https://doi.org/10.1007/s00247-026-06787-9

Image Credits: AI Generated

DOI: 10.1007/s00247-026-06787-9

Keywords: 4D flow MRI, congenital heart disease, cardiac MRI, non-contrast imaging, phase-contrast MRI, pediatric radiology, gadolinium, blood flow quantification, tetralogy of Fallot, 1.5 Tesla, stroke volume, regurgitant fraction