Thompson PM, Giedd JN, Woods RP, MacDonald D, Evans AC, Toga AW
Laboratory of Neuro Imaging, Dept. Neurology, Division of Brain Mapping,
UCLA School of Medicine, Los Angeles CA 90095, USA, and
Child Psychiatry Branch, NIMH, and
Montreal Neurological Institute, McGill University, Canada
Summary. We report the creation of the first, high-resolution, 4-dimensional quantitative maps of growth patterns in the developing human brain. Both regressive (tissue loss) and progressive (tissue growth) processes are characterized, across time spans of several weeks to four years. Their heterogeneity, local biases, and fine-scale characteristics are delineated. Growth patterns are detected with a tensor mapping strategy that produces a variety of maps, reflecting the magnitude and principal directions of tissue dilation or contraction, rates of strain, and the local divergence and gradient of the growth processes detected in the dynamically changing brain architecture. The resulting tensor maps provide spatially-detailed information on local growth. They reveal striking systematic patterns, as well as dramatic directional biases in the profiles of growth and tissue loss.
Methods. Several time-series of 3D (256x256x124 resolution; 24 cm FOV; 1.1 mm slice thickness) T1- weighted fast SPGR (spoiled GRASS) MRI volumes were acquired from young normal subjects (mean age: 9.9+/- 0.9 yrs.; 3 girls, 2 boys) covering intervals from 2 weeks to 4 years. Pairs of scans were selected to determine patterns of structural change across the interval between the two scans. After correcting both scans for radio-frequency inhomogeneities, the initial scan was rigidly registered to the target using automated image registration software, with chirp-Z resampling . High-resolution cortical models were automatically extracted  from each of the mutually registered histogram-matched scans. 3D parametric surface mesh models [3,4] were made to represent a comprehensive set of deep sulcal, callosal, caudate and ventricular surfaces at each time-point. Models were averaged across multiple trials (N=6) to minimize error . These model surfaces provided anatomic constraints for an elastic image registration algorithm [4,5], which reconfigures the anatomy at the earlier time-point into the shape of the later scan. The deformation field required to match the surface anatomy of one scan with the other was extended to the full volume using a continuum-mechanical model based on the Cauchy-Navier operator of linear elasticity . In the resulting high-dimensional transformation (0.1 billion degrees of freedom), the validity of the cortex-to-cortex transformation was guaranteed by forcing a large system of gyral and sulcal curves to match exactly [5,6]. Deformation processes recovered by the warping algorithm were analyzed with vector field operators to produce a variety of tensor maps, representing the growth processes recovered by the transformation.
Results. The resulting tensor maps revealed the phenomenal complexity of tissue growth in late brain development, indicating prominent foci of rapid growth and tissue elimination. Before and during puberty, highest growth rates were attained in temporo-parietal fiber systems, while highly-localized tissue losses were detected at the caudate head. A typical 4-year caudate tissue loss (up to 50% locally) contrasted with a 20-30% internal capsule growth, a 5-10% superior ventricular horn dilation, and an approximate equilibrium in total cerebral volume. Rostral callosal fiber systems were relatively stable, in sharp contrast with rapid focal growth at the callosal isthmus and its temporo-parietal projection systems (up to 80% growth in 4 years). These associative networks may undergo myelination over more prolonged periods than rostral fiber systems, with growth spurts spanning the 6 to 11 year age range.
The additional spatial detail and detection sensitivity offered by dynamic growth maps may help to generate powerful quantitative descriptors of developmental and disease processes, based on increasingly sensitive imaging strategies, and based on dynamic rather than static criteria.
References. . Woods et al., J. Comp. Assist. Tomogr. 16:620-633 (1992); . MacDonald et al., Proc SPIE 2359:160-169 (1994); [3-7]: Thompson et al.: . J. Neurosci. 16(13):4261-4274 (1996); . IEEE Trans. Med. Imag. 15(4):1-16 (1996); . Chapter 19 in Brain Warping, Toga AW [ed.], Academic Press (1998); . J. Comp. Assist. Tomogr. 21(4):567-581 (1997).
|RESUME| E-MAIL ME| PERSONAL HOMEPAGE| PROJECTS|