PCCA-Model: an attention module for medical image segmentation

Abstract

Convolutional neural networks have been increasingly employed in the field of medical image segmentation. Based on the idea that the human visual cortex differs in terms of the size of the receptive field and can sense the stimulus location, we propose the pyramid channel coordinate attention (PCCA) module to fuse multiscale features in the channel direction, aggregate local and global channel information, combine them with the location information in the spatial direction, and then integrate them into the existing semantic segmentation network. We conducted numerous experiments on the datasets, namely LiTS, ISIC-2018, and CX, and obtained state-of-the-art results.

Tasks

Reproductions