Automatic Measurement of the Volume of Brain Ventricles in Preterm Infants from 3D Ultrasound Datasets

In this work we evaluate and validate the lateral ventricular brain volume obtained using transfontanellar three-dimensional ultrasound (3D US). The goal is to improve diagnosis of excessive dilation of ventricles after germinal matrix- intraventricular hemorrhage in the preterm infant, a complication known as posthemorrhagic ventricular dilatation, with a computer-aid method that is not operator-dependent. To this end, a software has been developed for automatic segmentation of 3D US using conventional image processing techniques. It significantly reduces the processing time compared to the state-of-the-art methods. The results are discussed with respect to the measurements obtained using manual methods of segmentation that are considered as the gold standard.