基于ResUnet对抗网络的脑瘤图像分割方法

A Method for Segmentation Brain Tumor in MRI Based on ResUnet Embedded in GAN

定位病灶并将病灶分离出来一直是图像医学研究的热点,为快速准确地将脑肿瘤从脑部磁共振图像中分离出来,在了解传统脑肿瘤分割方法不足之处后,提出了基于ResUnet对抗网络的磁共振图像脑肿瘤分割方法。它的总框架是生成对抗网络,在对抗网络的生成器部分内嵌ResUnet。这种结构使得此语义分割的神经网络具有对抗网络无需在学习过程中进行推断的优点,具有残差网络的梯度不易消失的优点,同时能充分发挥Unet的特点。最后,以某医院提供的磁共振成像图片为样本经行训练,在与两种传统方法比较结果后,证明此方法有一定优势。

Locating and separating the lesions has always been a hot spot in imaging medical research.In order to quickly and accurately separate the brain tumors from brain MRI images,after understanding the shortcomings of traditional brain tumor segmentation methods,this paper proposes a brain tumor segmentation method based on ResUnet in GAN.Its overall framework is the generative adversarial network,with ResUnet being embedded in the generator section of the network.This structure makes the semantic segmentation neural network have the advantage that does not have to infer in the learning process because of GAN,and have the advantage that the gradient does not easy to disappear because of the residual network.Thus,it can give full play to the characteristics of the Unet.Finally,magnetic resonance imaging images provided by a hospital are used as samples for training,and the results are compared with those of two traditional methods,which proves that this method has certain advantages.