基于多层卷积注意力感知的肺部图像分割研究

Research on Lung Image Segmentation Based on Multi-Layer Convolutional Attention Perceptron

胸腔部位X光片检查已被广泛应用于各种肺部疾病的检测中,针对使用U-Net网络进行肺实质分割时遇到的噪声敏感问题,提出一种便于集成在卷积神经网络中的多层次注意力感知机,使网络降低对于分割任务无关区域的关注,并在原U-Net的基础上增加了数据增强操作。实验结果表明,与传统U-Net相比,该方法在保证计算效率的同时,能够获得更好的分割效果。

X-ray examination of the chest area is widely used in the detection of various lung diseases, aiming at the noise sensitivity problems encountered when using U-Net network for lung parenchymal segmentation, a multi-level attention perceptron method that is convenient for integration in convolutional neural networks is proposed, so that the network can focus on extracting multi-level features of the region of interest. And add data augmentation operations on the basis of the original U-Net. Experiments show that the proposed method obtains a better segmentation effect while ensuring computational efficiency compared with the traditional U-Net.