Atlas-Based Segmentation of the Brain for 3-Dimensional Treatment Planning in Children with Infratentorial Ependymoma

This paper presents a fully automated brain segmentation method that has been applied to a group of patients with infratentorial ependymoma. The purpose of the study was to test the hypothesis that fully-automated atlas-based segmentation methods provide useful normal tissue dosimetry from which dose-volume modeling may be performed in a manner equivalent to dose-volume data obtained from manual contouring. To test this hypothesis, we compared the integrated average dose for three small (chiasm, pituitary, hypothalamus) and three large (temporal lobes and total brain) normal tissue structures from ten patients using automated and manual contouring. There was no significant difference in the calculated average dose for the structures of interest. The greatest difference was noted for smaller structures which were located along the midline and in the gradient of dose. The results of this study form the basis of an ongoing larger study involving similar patients to evaluate automated and manual contouring as well as the clinical significance of any differences using dose-volume modeling.