DeepMRSeg: A convolutional deep neural network for anatomy and abnormality segmentation on MR images

摘要: Segmentation has been a major task in neuroimaging. A large number of automated methods have developed for segmenting healthy and diseased brain tissues. In recent years, deep learning techniques attracted lot attention as result their high accuracy different segmentation problems. We present new based method, DeepMRSeg, that can be applied generic way to variety tasks. The proposed architecture combines advances the field biomedical image computer vision. use modified UNet takes advantage multiple convolution filter sizes achieve multi-scale feature extraction adaptive desired task. Importantly, our method operates on minimally processed raw MRI scan. validated wide range tasks, including white matter lesion segmentation, structures hippocampus segmentation. provide code pre-trained models allow researchers apply own datasets.