Development of the human brain is incomplete at birth. During the first six years of life, the axonal pathways are progressively enveloped by a dense myelin sheath (i.e. myelination) to complete neuronal connectivity.¹ Abnormal and disrupted myelination is attributable to a variety of genetic disorders and neurologic conditions. The current clinical standard for determining myelin age and integrity is qualitative comparison of MR images to a reference atlas.² Quantitative MRI techniques to objectively evaluate myelin may improve detection of delayed myelination and provide a definitive characterization of disease effects on white matter.

Cross-relaxation imaging (CRI) is a quantitative MRI technique that captures the change in magnetization of water protons (free pool) induced by the transfer of magnetization from semi-solid macromolecular protons (bound pool).^3-8 Fast bound-pool fraction imaging (FBFI) was derived from CRI to provide voxel-based quantification of the bound-pool fraction, f, throughout the brain.⁹ FBFI has been validated with histology to measure myelin density and disease-associated changes in an animal model of glioma.⁹ Subsequently, FBFI was used to measure age-related degradation of white matter in adult patients with Fabry disease, a lysosomal storage disorder associated with stroke.¹⁰ However, integration of specialized pulse sequences (e.g. B₁ mapping techniques, MT pulse modification) and overall scan time, particularly in pediatric patients, remain barriers to clinical applications. In this study, we sought to translate FBFI to a clinical scanner without system modifications for the time-efficient direct measurement of myelin density in the healthy pediatric brain.

To correct for field heterogeneity at 3T, B₀ and B₁ maps are routinely obtained during determination of the bound-pool fraction.^10,11 However, sequences for acquisition of whole-brain B₀ and B₁ maps are not available on clinical scanners and implementation through research key access increases scan time. Previously acquired data¹¹ from adults with Fabry disease and age/gender matched controls were utilized to characterize error in the absence of correcting for B₀ and B₁ heterogeneity. Effects of B₀-field heterogeneity were negligible (<1%) on determination of f (Figure 1). B₁-field heterogeneity had a larger effect, but remained small (<5%) and tended to systematically alter measurements based on anatomic location (Figure 1). This finding suggests accuracy more so than precision is a consequence of errors associated with B₁ heterogeneity, which enables comparison of specific anatomic structures within (i.e. serial studies) and between patients imaged on the same scanner.

Inability to optimize MT pulse delivery timing parameters on clinical scanners lengthens TR in pediatric patients due to SAR limitations. To evaluate feasibility of measuring f utilizing the matrix model of pulsed magnetization transfer¹² on a 3T whole-body pediatric clinical scanner (GE HD, Milwaukee, WI) an agarose-based phantom was constructed. A single Z-spectroscopic data point utilizing the maximum allowed offset frequency (∆=1.6kHz; duration 14ms) was acquired with a 3D SPGR sequence (TE=6ms, α=10°). Z-spectroscopic images were acquired with TR=50 and 90ms, representing the shortest allowable TRs for a 45 kg adult and 2 kg infant, respectively, using the preset MT pulse parameters. A complementary R₁ map necessary for parameter fitting was obtained using the variable flip angle (VFA) method with a 3D SPGR sequence (TR/TE = 22/6 ms, α=4, 10, and 30°). A synthetic reference image for normalization of Z-spectroscopic data was derived as previously described⁷ from:

$$$S_{\alpha}=PD\frac{1-e^{-R_{1}TR}}{1-cos\alpha e^{-R_{1}TR}}sin\alpha$$$

The MT ratio was higher for TR=50ms compared to TR=90ms in the 2% agarose gel (23.4% vs. 21.1%, respectively), but similar in the 1% agarose gel (13.7% vs. 13.8%, respectively). Measurements of f were comparable between TRs and both were able to effectively discriminate differences in agarose concentrations (Figure 2). This finding indicates that although the longer TR increases scan time, it does not compromise measurement of f.

Subsequently, a time-efficient protocol was developed for capturing f in the pediatric brain utilizing a single Z-spectroscopic data point (∆ = 1.6 kHz; duration 14 ms; TR/TE=90/6ms) and a complementary R₁ map as described above (Figure 3). Pediatric patients (ages 2 months – 7 years) receiving a clinical brain MRI for headaches or seizures were recruited for the study. All images were acquired with FOV=22×22cm², matrix=128×128, acquisition resolution of 0.86×0.86mm², and 28 slices (slice thickness of 2.5mm). Total scan time was <7 minutes. Evidence of progressive myelination with age was present on f maps (Figure 4). Age-progressive myelination in the posterior white matter corresponded strongly to a bounded exponential growth curve (Figure 5).