基于深度卷积神经网络的低剂量CT肺部去噪

Low Dose CT Lung Denoising Model Based on Deep Convolution Neural Network

为了降低低剂量CT肺部噪声对肺癌筛查后期诊断的影响,该文提出一种基于深度卷积神经网络的低剂量CT肺部去噪算法。以完整的CT肺部图像作为输入,池化层对输入图像进行降维处理;批规范化解决随着网络深度的增加性能降低的问题;引入残差学习,学习模型中每一层的残差,最后输出去噪图像。与经典去噪算法实验结果对比,所提方法在解决去噪方面达到了很好的滤波效果,同时也较好地保留了肺部图像的细节信息,大大优于传统的去噪算法。

In order to reduce the effect of low dose CT lung noise on the late diagnosis of lung cancer screening,a denoising model of low-dose CT lung based on deep convolution neural network is proposed.The input of the model is the complete CT lung image.The pooling layer reduces the dimension of input.Batch normalization works out the poor performance with the increase of network depth.The residuals of each layer are learned with residual learning.Finally,the denoised image is produced.Compared with classical methods,the proposed method achieves good filtering effect in solving the denoising method,and also retaining the details of lung image information,which is much better than the traditional filtering algorithm.