Unified medical image segmentation by learning from uncertainty in an end-to-end manner

作者：

Highlights：

• A deep learning-based network named UG-Net is proposed for medical image segmentation.

• An uncertainty guided module is proposed to learn from uncertainty in an end-to-end manner.

• A feature refinement module with dual attention mechanism is designed for further performance promotion.

• A multi-scale feature extractor is devised to fit into different segmentation tasks.

摘要

•A deep learning-based network named UG-Net is proposed for medical image segmentation.•An uncertainty guided module is proposed to learn from uncertainty in an end-to-end manner.•A feature refinement module with dual attention mechanism is designed for further performance promotion.•A multi-scale feature extractor is devised to fit into different segmentation tasks.

论文关键词：Medical image segmentation,Uncertainty,End-to-end,Deep learning,Feature refinement

论文评审过程：Received 10 June 2021, Revised 8 January 2022, Accepted 10 January 2022, Available online 19 January 2022, Version of Record 3 February 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.108215