Automatic real-time CNN-based neonatal brain ventricles segmentation

Quantitative imaging of brain plays an important role in preterm neonates with a very low birth weight due to the increased risk of developing intraventricular hemorrhage (IVH). In this work, we propose a fully automated method for segmentation of ventricles from two-dimensional (2D) ultrasound (US) scans. The proposed method is based on a Convolutional Neural Network (CNN) that combines the advantages of U-Net and SegNet architectures for ventricles segmentation. Extensive experiments on a dataset consisting of 687 US scans show that the proposed method achieves significant improvements over the state-of-the-art medical image segmentation methods.