Segmentation of Gliomas and Prediction of Patient Overall Survival: A Simple and Fast Procedure

This paper proposes, in the context of brain tumor study, a fast automatic method that segments tumors and predicts patient overall survival. The segmentation stage is implemented using a fully convolutional network based on VGG-16, pre-trained on ImageNet for natural image classification, and fine tuned with the training dataset of the MICCAI 2018 BraTS Challenge. It relies on the “pseudo-3D” method published at ICIP 2017, which allows for segmenting objects from 2D color-like images which contain 3D information of MRI volumes. With such a technique, the segmentation of a 3D volume takes only a few seconds. The prediction stage is implemented using Random Forests. It only requires a predicted segmentation of the tumor and a homemade atlas. Its simplicity allows to train it with very few examples and it can be used after any segmentation process. The presented method won the second place of the MICCAI 2018 BraTS Challenge for the overall survival prediction task. A Docker image is publicly available on https://www.lrde.epita.fr/wiki/NeoBrainSeg.