MSAR-Net: A multi-scale attention residual network for medical image segmentation

Xiaoheng Li, Cheng Chen, Yunqing Chen, Mingan Yu, Ruoxiu Xiao

Abstract: Highlights•A novel multi-scale attention residual network for vital organ segmentation before minimally invasive surgery.•Hierarchical residual connections enhance the model’s perceptual ability in multi-scale refinement of feature representations.•Attention mechanisms and cross-scale residual connections enhance segmentation at object boundaries by aggregating features.•The MSAR Network excels in organ segmentation for parathyroid ablation and atlantoaxial dislocation screw implantation surgeries.