AN AUTOMATIC SEGMENTATION OF BRAIN MRIS OF 2-YEAR-OLDS INTO 83 REGIONS OF INTEREST Ioannis S. Gousias, United Kingdom    - Biography English - 2007-10-07

PLAY PRESENTATION	( 19 slide(s) )
PODCAST

VIEW THE SLIDE LIBRARY FOR THIS PRESENTATION

Background and aims: Three-dimensional atlases and databases of the brain at different ages facilitate the description of neuroanatomy and the monitoring of cerebral growth and development.

Methods: We have developed an algorithm for automatic segmentation of young children’s brains into 83 regions of interest, applied to an exemplar group of 33 2-year-old subjects born prematurely. We validated this algorithm by comparing the automatic approach with three representative manually segmented volumetric regions using similarity indices (SI).

Results: 33 atlases for Magnetic Resonance data sets of 2-year-old subjects were created (Figure_a). The group of structures covers the whole brain (cortical and subcortical areas). Structure volumes were compatible and slightly smaller than those of older subjects (e.g. 2-yearolds/ 8-year-olds, left caudate 3.57/3.65 cm³, right hippocampus 1.89/3.00 cm³)(Peterson, et al., 2000). SI results for automatic versus manual segmentations for caudate nucleus, pre-central gyrus and hippocampus were 0.90 ± 0.01, 0.90 ± 0.01 and 0.88 ± 0.03 respectively. Mean SI for intra-rater variability was 0.95 ± 0.01. Partitioned versions of the atlases were created based on the grey and white matter partitions of the 2-year-old data sets. The performance of our method was also high in producing atlases of data sets with prominent anatomical abnormalities (Figure_b).

Conclusions: An atlas of such detailed segmentation can be a useful tool in the monitoring of developmental growth of different brain regions in longitudinal studies and in group comparisons between normal controls and pathological cases of the same age, since this can be used to study brain growth in health and disease.