Motion-Robust Iron Quantification of Newborn Brains using Radial MRI

Iron plays a vital role in brain development, contributing to processes such as synaptogenesis, myelination, and neurotransmitter synthesis. Perinatal iron deficiency has been linked to delayed nerve conduction, impaired recognition memory, motor difficulties, and lower overall developmental scores. Preterm births affect about 1 in 10 U.S. infants, and 25 - 85% of them show signs of iron deficiency during infancy. Accurate measurement and timely supplementation of brain iron during this early critical phase is therefore essential for optimizing neurodevelopmental outcomes. Qutantitatve susceptibility (QSM) and R2* MR imaging have proven to be important tools for the non-invasive quantitative measurement of brain iron levels. However, these techniques have not yet been robustly deployed in infant studies, especially those preterm infants. The key challenges include: 1) Traditional Cartesian imaging techniques are lengthy (>10 mins) for neonates and highly motion-sensitive; 2) Motion-induced geometric and field distortions introduce additional phase errors, compromising quantitative accuracy; 3) Standard QSM captures average susceptibility within each voxel, making it difficult to separate iron (positive susceptibility) from myelin (negative susceptibility) in the neonatal brains. To address these challenges and enable the reliable use of quantitative MRI techniques to study iron deficiency of neonatal brains, we propose motion-robust 3D kooshball data acquisition and advanced reconstruction strategies for rapid high-resolution QSM and R2* imaging of neonatal brains. In Aim 1, we will extend an existing 3D radial sequence to incorporate multi-echo readouts, adapting previously developed motion estimation algorithms that extract brain motion and motion-induced field maps from the data itself. We will further design an image space-based reconstruction algorithm to generate high-quality MR images from undersampled radial data. In Aim 2, we will develop and validate a motion-corrected, deep-learning- regularized model-based nonlinear reconstruction algorithm that directly estimates quantitative maps from highly undersampled radial data using our rapid protocol from Aim 1, significantly reducing the imaging time. In Aim 3, we will apply source separation techniques to distinguish iron (positive) and myelin (negative) susceptibility from the estimated QSM maps using R2* and R2 information. We will perform evaluation of the utility of the developed technologies in assessing brain iron development in preterm neonatal subjects with iron deficiency compared to full-term neonates. To that end, our novel technologies would speed up the quantitative MR exams and allow for accurate and reliable iron quantification of newborn brains.