Pulmonary and cardiac diseases are the leading causes of morbidity and mortality in premature babies. Hence, chest MRI has the potential to improve clinical diagnosis and patient care in the neonatal intensive care unit (NICU). However, physiological motion poses a challenge to MRI data acquisition in neonates who have higher heart and respiration rates than adults. In our institution, we are investigating chest MRI on a small-bore (21.8cm) MRI scanner dedicated for neonatal examination within the NICU^1,2. A real-time imaging technique is developed to collect neonatal chest MRI data faster than cardiac and respiratory motion. It is demonstrated that our new approach can provide high-quality cardiac and lung images for functional and structural chest examination in premature babies.

The real-time imaging technique for neonatal chest MRI consists of the following three functional units:

1) A 10-channel receive coil array for parallel imaging (Figure 1a)³: The coil housing consists of two identical polycarbonate shells designed using fused deposition modeling (FDM) technology (Stratasys Ltd., Valencia, CA). When combined, the two shells form a thin cylinder designed to comfortably hold a swaddled baby. Each shell has five coil loops with a diameter of ~5 cm. The five elements provides a 2D field of view (FOV) coverage of ~8x10cm. The two shells are placed within an 18cm transmit-only volume coil during the scan.

2) A pulse sequence for real-time data acquisition (Figure 1b): A single-shot 2D steady-state free precession (SSFP) sequence is used to collect k-space data. The entire k-space is divided into multiple regions. Each region is uniformly undersampled in the phase encoding direction and the undersampling factors increase from the center to the outer k-space regions. This undersampling strategy typically gives a net acceleration factor of 8 for an acquisition matrix of 128x128, providing a temporal resolution of ~50 milliseconds for real-time data acquisition.

3) An iterative algorithm for image reconstruction (Figure 1c): A technique developed in our previous works^4,5, correlation imaging, is used to reconstruct images from undersampled data. The image reconstruction is performed region-by-region from the center to the outer k-space based on the undersampling trajectory used in data acquisition. In each region, the reconstructed data are used as a feedback to improve the calibration of correlation imaging in the subsequent iteration. The image reconstruction is run until all of k-space is covered.

To demonstrate real-time imaging for neonatal chest MRI within the NICU, 5 neonatal patients (2 weeks to 2 months of age and <2.5kg of weight) were scanned with free-breathing and without sedation. Short-axis cardiac imaging data were collected in real-time for ~10 seconds with a temporal resolution of 48 milliseconds (FOV 16x16cm, matrix 128x128, TR/TE 3.7/1.1 ms, slice thickness 5 mm, flip angle 45°). A set of cine images collected with k-space segmentation (8 views per segment) was used as references. Real-time lung images were collected for ~40 seconds with a temporal resolution of 123 milliseconds (FOV 16x16cm, matrix 192x192, slice thickness 6mm, flip angle 10°). A set of static lung images collected using a 3D fast gradient echo sequence was used as references.

Figure 2 shows a cardiac imaging example from a subject with a heart rate of >150 beats per minute. Compared with k-space segmentation, real-time imaging gives better dynamic contrast. The temporal trajectory also demonstrates that real-time imaging gives a higher speed than respiratory and cardiac motion. Figure 3 shows a lung imaging example from a subject with a respiratory rate of ~80 breaths per minute. Compared with static imaging, real-time imaging gives higher signal in lung parenchyma. Using the Fourier decomposition approach described in a previous work⁶, ventilation- and perfusion-weighted images can be generated from the temporal-dimension Fourier transform of the real-time dynamic image series.The neonates in our ongoing study have a respiratory rate of 30-90 breaths per minute and a heart rate of 100-200 beats per minute. In addition, cardiac and respiratory motion shows non-periodic behaviors. As a result, k-space segmentation introduces considerable data inconsistency that may cause a loss of dynamic contrast (Figure 2). In lung imaging, static imaging typically needs ~3 minutes for data acquisition and motion may introduce a considerable signal loss. By collecting data faster than motion, real-time imaging can effectively reduce this loss, providing a signal gain over static imaging within the lung parenchyma (Figure 3).It is experimentally demonstrated that real-time imaging can provide high-quality cardiac and pulmonary images for improved clinical diagnosis and patient care in premature babies within the NICU.