In neonates, ultrasound is the initial neuroimaging modality used to detect and follow up on intracranial pathologies, since MRI is challenging due to immobilization, cost and sedation issues. Ultrasound is typically acquired in 2D and interpretations are performed slice by slice. However, recent reports suggest a longitudinal follow-up of structure volumes and shapes may be relevant for investigation of neurodevelopmental disorders. The first objective of this study is to test the diagnostic efficiency of the 3D ultrasound technology to measure total brain volume as well as lateral ventricular volume compared to volumetric measurements obtained from MRI. Subsequently, an algorithm automatically estimates the volume of the brain in 3D ultrasound with the method validated by MRI images. Then a segmentation of the lateral ventricles is achieved with an image registration and a deformable mesh. This makes it possible to automatically calculate the ventricle volume and to quantify its relative dilatation compared to the brain by extracting the ventricle / brain volume ratio.
This will allow us to do 3D volume segmentation of the ventricles with 3D ultrasound images, which is easier than using MRI from a clinical perspective.
