Volumetric multimodality neural network for brain tumor segmentation

Brain lesion segmentation is a challenging biomedical problem. Here we present a convolutional neural network that produces a semantic segmentation of brain tumors, capable of processing volumetric information from multiple MRI modalities at the same time. This results in the ability to learn from small training datasets and highly imbalanced data. We present a new architecture with three parallel contracting pathways that receive inputs in different resolution and then merges their results using three fully connected layers. We tested our method over the 2015 BraTS Challenge dataset, reaching an average dice coefficient of 84%.