基于全卷积神经网络的前列腺磁共振图像分割

（本科毕业论文题目、摘要、关键词及相应的英文翻译，欢迎做前列腺分割的小伙伴可以加我qq：604395564一起交流呀）

 

基于全卷积神经网络的前列腺磁共振图像分割

 

目的研究从前列腺磁共振图像中自动分割的方法。方法本文提出了基于全卷积神经网络的前列腺磁共振图像分割方法。首先将三维前列腺磁共振图像经切片采样得到二维图像，然后将前列腺图像切片输入全卷积神经网络，该网络结构可实现图像像素级别的语义分割，通过其特有的反卷积结构拓展下采样层提取到的特征图尺寸，得到与输入图像尺寸相同的分割结果。结果\结论大量基于MICCAI 2012公开数据集的实验表明，基于全卷积神经网络的分割方法能自适应前列腺磁共振图像的差异性,使图像分割精度得到明显提高。

关键词: 前列腺分割； 磁共振成像； 卷积神经网络； 全卷积神经网络

 

Prostate Magnetic Resonance Images Segmentation Based on Fully Convolutional Neural Networks

Aim Automatic segmentation methods for segmenting prostate from magnetic resonance images. Method Firstly, the 3D prostate MR images are sliced into 2D images. Then, the prostate MR image slices are inputted to the fully convolution neural network which can realize the pixel-level image semantic segmentation. By its unique deconvolution structure, the size of feature image extracted from down-sampling layer is expanded and the size of segmentation results are the same as that of input images. Results\Conclusion A large number of experiments based on MICCAI 2012 publicly available data show that the segmentation method based on fully convolutional neural network can adapt the difference of prostate MR images and improve the image segmentation accuracy obviously.

Key words: prostate segmentation; magnetic resonance imaging; convolutional neural networks; fully convolutional networks