Segmentation de la fosse postérieure en IRM fœtale : application biométrique

Abstract

Posterior fossa segmentation in fetal MR imaging: biometrical application. In order to complete ultrasound scanning and histological tests, magnetic resonance imaging is now considered for the study of prenatal development and fetal pathologies. Due to the recent implementation the quantitative data are still lacking, particularly concerning the central nervous system. Our study is dealing with fetal cerebral MR imaging, and, more precisely, with the biometrical analysis of the different structures of the posterior fossa. Its malformations and pathologies are often due to a trouble of “container-content” and could require a medical termination of pregnancy. So, the need of quantitative information about the relative growth of the different parts of the posterior fossa during the pregnancy is real. We have therefore developped a segmentation method based on region growing with pixel agregation, which has been tested on 67 sagital cross-section MR images. The good segmentation has permitted us to determine the regions of interest (posterior fossa, vermis, brain stem) and to calculate four relative surface measures between them. After technical and clinical validations, the analysis of our results shows that the relative development of these structures is constant and independent of the pregnancy term. It may thus be possible to find standard index for characterizing their relative growth.