ResAttenGAN: Simultaneous segmentation of multiple spinal structures on axial lumbar MRI image using residual attention and adversarial learning

作者：

Highlights：

• Proposing ResAttenGAN to segment simultaneously multiple spinal structures on the axial lumbar MRI image

• The full feature fusion module is proposed to fully fuse multi-scale and multi-hierarchy features.

• The residual refinement attention module is proposed to refine implicit borders.

• Applying adversarial learning to effectively smooth the higher-order spatial consistency for avoiding overfitting

摘要

•Proposing ResAttenGAN to segment simultaneously multiple spinal structures on the axial lumbar MRI image•The full feature fusion module is proposed to fully fuse multi-scale and multi-hierarchy features.•The residual refinement attention module is proposed to refine implicit borders.•Applying adversarial learning to effectively smooth the higher-order spatial consistency for avoiding overfitting

论文关键词：Simultaneous segmentation,Multiple structures,Axial MRI image,Feature fusion,Attention module,GAN

论文评审过程：Received 24 November 2020, Revised 18 June 2021, Accepted 3 January 2022, Available online 8 January 2022, Version of Record 17 January 2022.

论文官网地址：https://doi.org/10.1016/j.artmed.2022.102243