Brain Lacunae Segmentation from Fair Sequence Based On Fully Convolutional Neural Network

Neuroimaging in the context of brain disease is becoming more and more important. Brain detection and segmentation are two fundamental steps in neuroimage analysis. Because the cost of manual segmentation of the brain is too much, more and more researchers have developed the semi-automatic or automatic brain tumor segmentation methods. However brain lacunae segmentation are different form brain tumor segmentation, the shape and size of brain lacunae are smaller than brain tumor. This paper presents a deep fully convolutional neural network model for brain lacunae segmentation. The experimental results show that deep fully convolutional neural network for brain lacunae segmentation performs well. In addition, the deep fully convolutional neural network for brain lacunae segmentation with preprocessing of histogram equalization, batch normalization, and dropout layers improves the experimental speed and dice coefficient.