The research, led by Xin Miao of Children’s Hospital Los Angeles and the Keck School of Medicine of the University of Southern California, together with colleagues across multiple institutions, set out to test whether a single eight-minute scan on a commercial 0.55-tesla MRI system could simultaneously capture both the structure and the function of diseased infant lungs. Eleven preterm infants diagnosed with bronchopulmonary dysplasia were recruited from the neonatal intensive care unit, all of whom were already scheduled for clinically indicated brain MRI. After their routine brain scan, the infants were transferred to an adjacent 0.55-tesla scanner for the research lung examination. The cohort spanned a wide range of clinical circumstances: gestational ages at birth ranged from 23 to 32 weeks, post-menstrual ages at scanning ranged from 40 to 57 weeks, and body weights ranged from 2.3 to 6.4 kilograms. Nine infants had severe bronchopulmonary dysplasia and two had moderate disease, with five on non-invasive oxygen support and six on mechanical ventilation at the time of imaging.

The technical heart of the study is a prototype sequence called bSTAR, short for balanced steady-state free precession with half-radial dual-echo readout. The choice of sequence and field strength is not arbitrary. Lung MRI has historically been notoriously difficult because air-filled tissue is sparse in hydrogen protons and riddled with microscopic air-tissue interfaces that rapidly dephase the magnetic resonance signal. At conventional field strengths of 1.5 and 3 tesla, researchers have relied on ultrashort echo time sequences to catch the signal before it decays, but the parenchymal signal-to-noise ratio remains a fundamental constraint, especially in preterm infants whose tiny lungs demand high spatial resolution. At 0.55 tesla, the lung’s transverse relaxation time, T2*, stretches roughly tenfold, from about one millisecond at 1.5 tesla to approximately ten milliseconds, and susceptibility artifacts are markedly reduced. That longer signal lifetime gives the scanner far more room to work with.

Lower field strength also unlocks a sequence that has long been underused in the chest. Balanced steady-state free precession offers inherently high signal-to-noise efficiency, but at conventional field strengths it has been hobbled by banding artifacts and by specific absorption rate limits that constrain how fast the sequence can run. Both problems ease substantially at 0.55 tesla, allowing the full advantage of the sequence’s contrast to be exploited for lung parenchymal imaging. In adults, this combination has already produced submillimeter isotropic images with CT-like depiction of lung tissue. The bSTAR sequence in this study acquired three-dimensional data at 1.3-millimeter isotropic resolution over an eight-minute acquisition, using a 340-millimeter field of view, a flip angle of 25 degrees, and echo and repetition times of 0.2, 2.3, and 2.5 milliseconds respectively. Crucially, the infants were not sedated for the lung scan and did not need to hold their breath. Those on non-invasive support breathed freely through a nasal cannula, while ventilated infants breathed under controlled respiratory patterns, with clinically ordered respiratory support maintained throughout.

Handling motion from a freely breathing, sometimes crying infant is where the signal processing becomes as important as the physics. The team extracted respiratory self-navigation waveforms directly from the center of k-space, sampled every 2.5 milliseconds, then filtered and selected the channel with the highest variance as the breathing signal. Data contaminated by bulk motion, identified from waveform irregularity, were discarded. The remaining data were sorted into three respiratory motion states, and each state and each echo was reconstructed independently using non-Cartesian compressed sensing with wavelet regularization, implemented through the Berkeley Advanced Reconstruction Toolbox. Reconstructing a single three-dimensional volume took roughly an hour on a 64-core Linux server. To derive functional information, the end-exhalation volume was deformably registered to the end-inhalation volume, and fractional ventilation was calculated at every lung voxel from the signal change between the two states, following a formula originally established by Zapke and colleagues. Two quantitative biomarkers emerged from these maps: ventilation defect percentage, the share of lung volume with ventilation below 0.75 times the median, and ventilation inhomogeneity, the coefficient of variation of ventilation values across the lung.

The structural results were striking. Two pediatric radiologists, one with nine years of post-fellowship experience and one fellow, independently scored the images using the Ochiai system, a scoring scheme originally developed for chest CT that evaluates eight features including hyper-expansion, mosaic attenuation, bullae and blebs, triangular subpleural opacities, bronchovascular distortion, intercostal bulging, and consolidation, with total scores from 0 to 18. The MRI captured every BPD-associated feature, and in the three patients who also had a clinical chest CT within a week of the scan, feature-by-feature comparison showed largely consistent scoring, with a mean total score difference of 2.8. Inter-reader agreement for the total score was substantial, with a weighted Cohen’s kappa of 0.80, and for individual features such as hyper-expansion, bronchovascular distortion, and bleb size the kappa values of 0.91, 0.94, and 0.85 were numerically higher than those typically reported for CT-based scoring of the same features. Notably, the researchers applied the original, unmodified Ochiai system, including intercostal bulging, which prior MRI studies had excluded and which proved to be among the more frequently observed features in this cohort.

Every one of the eleven scans produced diagnostic image quality, regardless of whether the infant was breathing spontaneously or on a ventilator. The median overall image quality score was 4.0 on a five-point scale, and nine cases scored above the diagnostic threshold with minimal to mild motion blurring. The two remaining cases, blurred by multiple crying episodes, still remained diagnostic. The Ochiai structural score correlated significantly with clinically meaningful variables: it was higher in infants on mechanical ventilation at the time of scanning compared with those on non-invasive oxygen support, a difference that reached statistical significance at P equals 0.035, and it correlated inversely with birth weight, with a Spearman coefficient of minus 0.70, and positively with the duration of mechanical ventilation at discharge, with a coefficient of 0.82. In a cohort where nine of eleven infants had severe disease, these associations suggest that what the scanner sees tracks what the clinical team experiences.

Perhaps the most intriguing findings came from the ventilation maps, which repeatedly resolved ambiguities that structural images alone could not. In one infant whose two lungs showed diffusely different signal intensities, the ventilation map identified markedly reduced ventilation in the right lung, confirming hyper-expansion as the underlying pathology. In another, hyper-expansion in the anterior lung was detectable on structural images only through subtle intercostal bulging, with no corresponding signal change, while the ventilation map provided clear functional confirmation. Elevated signal in the posterior lungs of two infants, which could have represented either consolidation or a normal gravity-induced density gradient, was clarified by matching low ventilation, favoring consolidation. In one case the reverse occurred: a region that mimicked a bleb on structural images showed relatively high ventilation, suggesting preserved function rather than true air trapping. Ventilation defect percentage and inhomogeneity were numerically higher in infants with severe disease and in those still on mechanical ventilation at discharge, though these differences did not reach statistical significance in this small pilot cohort, and the two biomarkers were strongly correlated with each other.

The authors are careful about the limits of the work. The cohort was small and skewed toward severe disease, most subgroup comparisons involved only two or three subjects, and the ventilation biomarker definitions were adapted from other fields with empirically chosen post-processing parameters that can influence absolute values. Absolute fractional ventilation also cannot be fully interpreted without normative models of how infant lung ventilation changes with development, which do not yet exist. Still, the study establishes something genuinely new: that a single free-breathing, radiation-free, sedation-free acquisition on a compact 0.55-tesla scanner can deliver both high-resolution structural images and quantitative three-dimensional ventilation maps in critically ill preterm infants. Because lung density changes rapidly with normal development and overlaps with the course of bronchopulmonary dysplasia, pairing structural and functional measures may prove especially powerful for distinguishing disease from normal physiology. Larger studies spanning the full severity spectrum will determine whether combined structural-functional MRI can improve diagnostic confidence beyond structural imaging alone, and whether it can finally give clinicians an objective, repeatable window into the lungs of their tiniest patients.

Subject of Research: Low-field MRI for structural and functional lung imaging in preterm infants with bronchopulmonary dysplasia

Article Title: Feasibility of free-breathing 0.55-T lung MRI for structural and functional characterization in preterm infants with bronchopulmonary dysplasia

Article References: Miao, X., Nguyen, H. N., Lee, N. G., Uribe, J., Doyle, E. K., Capaldi, D. P., Arrioja, L., Kato, R. M., Durazo-Arvizu, R. A., Nayak, K. S., & Iyer, N. P. (2026). Feasibility of free-breathing 0.55-T lung MRI for structural and functional characterization in preterm infants with bronchopulmonary dysplasia. Pediatric Radiology. https://doi.org/10.1007/s00247-026-06793-x

Image Credits: AI Generated

DOI: 10.1007/s00247-026-06793-x

Keywords: bronchopulmonary dysplasia, preterm infants, 0.55-T MRI, lung imaging, bSSFP, bSTAR, ventilation mapping, neonatal intensive care, Ochiai score, fractional ventilation, pediatric radiology, radiation-free imaging