CANet: Context aware network with dual-stream pyramid for medical image segmentation

Owing to the various object types and scales, complicated backgrounds, and similar appearance between tissues in medical images, it is difficult to extract some valuable information from different medical images. In this paper, we propose a context aware network with dual-stream pyramid (CANet) for medical image segmentation, which comprises a dual-stream pyramid module and an encoder–decoder module with context aware. Concretely, the dual-stream pyramid captures numerous complementary features at different layers by adopting multi-resolution input versions and multi-scale convolutional units, which is conductive to learning the local detail features in various scales. The encoder–decoder module with context aware progressively concatenates semantic features of the encoder branch with high-level features of the decoder branch in an efficient manner, which aims to suppress complicated backgrounds and highlight the most attractive object in medical images. Quantitative and qualitative experiments demonstrate that our CANet favorably performs against 13 state-of-the-art object segmentation methods on three publicly available medical image segmentation datasets. The code will be released at: https://github.com/Xie-Xiwang/BSPC2022_CANet.