LSSANet: A Long Short Slice-Aware Network for Pulmonary Nodule Detection

Convolutional neural networks (CNNs) have been demonstrated to be highly effective in the field of pulmonary nodule detection. However, existing CNN based detection methods lack ability capture long-range dependencies, which is vital for global information extraction. In computer vision tasks, non-local operations widely utilized, but computational cost could very high 3D computed tomography (CT) images. To address this issue, we propose a long short slice-aware network (LSSANet) nodules. particular, develop new mechanism termed slice grouping (LSSG), splits compact embeddings into short-distance grouped one and long-distance counterpart. This not only reduces burden, also keeps dependencies among any elements across slices whole feature map. The proposed LSSG easy-to-use can plugged many networks. verify performance LSSANet, compare with several recently competitive approaches on 2D/3D CNN. Promising evaluation results large-scale PN9 dataset demonstrate effectiveness our method. Code at https://github.com/Ruixxxx/LSSANet .