MA-SAM: A Multi-atlas Guided SAM Using Pseudo Mask Prompts without Manual Annotation for Spine Image Segmentation pdf

We only give the framework of the training procedure, please fulfill relevant parts (e.g., the spine_dataset.py, the properties.py, and the relevant codes in train.py) before actually conducting the training process.

1. Train the coarse segmentation sub-network first to obtain coarse segmentation results of input image.
2. Use registration.py to obtain warped label maps of atlas as prompts.
3. Train segmentation sub-network using input image and warped label maps of atlas.

  @article{fan2025ma,
  title={MA-SAM: A Multi-atlas Guided SAM Using Pseudo Mask Prompts without Manual Annotation for Spine Image Segmentation},
  author={Fan, Dingwei and Zhao, Junyong and Li, Chunlin and Wang, Xinlong and Zhang, Ronghan and Zhu, Qi and Wang, Mingliang and Si, Haipeng and Zhang, Daoqiang and Sun, Liang},
  journal={IEEE Transactions on Medical Imaging},
  year={2025},
  publisher={IEEE}
}