Automatic Template-based Brain Extraction in Fetal MR Images

Youssef Taleb (1) , Marc Schweitzer (2) , Colin Studholme (3) , Mériam Koob (4) , Jean-Louis Dietemann (5) , François Rousseau (1)

1 LSIIT - Laboratoire des Sciences de l'Image, de l'Informatique et de la Télédétection
2 LSIIT
3 BICG, Dept. of Pediatrics
4 Service de Radiologie
5 Service de radiologie [Strasbourg]

MRI is currently a very powerful imaging technique to study the human fetal brain in-utero. However, 3D MRI-based fetal brain studies remain challenging due to the low resolution of the images and the fetal motion that occurred during the acquisition. Although efficient image analysis algorithms have been developed to reconstruct 3D high-resolution fetal brain images [1,2,3], one of the first processing steps that consists in detecting the fetal brain in the images is usually manually performed. In [4], relying on priors such as image contrast, morphological and biometrical information, Anquez et al. have proposed a detection strategy that first estimates locations of eyes and then the fetal brain. More recently, Ison et al. [5] have presented a supervised method based on a random forest classifier, and an approximate high-order Markov random field solution. In [5], the complete brain lies inside the final bounding box in only 28% to 53% of cases. To develop a more robust fetal brain extraction algorithm, we investigate the use of a template-based technique that makes use of geometrical information of acquired scans to automatically create a mask of the brain in fetal MR images.

Domains: Signal and Image processing; Signal and Image Processing

Youssef Taleb, Marc Schweitzer, Colin Studholme, Mériam Koob, Jean-Louis Dietemann, et al.. Automatic Template-based Brain Extraction in Fetal MR Images. 2013. ⟨hal-00879089⟩