论文部分内容阅读
针对腹部CT影像邻近器官对比度较低及因个体肝脏形状差异较大等引起肝脏分割困难的问题,提出了全卷积神经网络肝脏分割模型。首先通过卷积神经网络提取图像深层、抽象的特征,再通过反卷积运算对提取到的特征映射进行插值重构后得到分割结果。由于单纯进行反卷积得到的分割结果往往比较粗糙,因此,在反卷积之前,先融合高层与低层的特征,并且通过增加反卷积的层数、减少反卷积步长,得到了更为精确的分割结果。与传统卷积神经网络的分割方法相比,该模型可以充分利用CT影像的空间信息。实验数据表明该模型能够使腹部CT影像肝脏分割具有较高的精度。
In order to solve the problem of liver segmentation caused by the low contrast of adjacent organs in the abdominal CT images and the difficulty of the liver segmentation due to the large differences in the shape of the individual liver, a liver segmentation model of the full convolution neural network is proposed. Firstly, the deep and abstract features of the image are extracted by convolutional neural network, then the extracted feature map is interpolated and reconstructed by deconvolution to get the segmentation results. Since the segmentation result obtained by simply performing deconvolution is relatively rough, the characteristics of the upper and lower layers are fused prior to deconvolution, and by increasing the number of deconvolution layers and reducing the deconvolution step, more For accurate segmentation results. Compared with traditional convolution neural network segmentation method, this model can make full use of the spatial information of CT images. The experimental data show that the model can make abdominal CT image segmentation with high accuracy.