CarveMix: A simple data augmentation method for brain lesion segmentation

Xinru Zhang, Chenghao Liu, Ni Ou, Xiangzhu Zeng, Zhizheng Zhuo, Yunyun Duan, Xiaoliang Xiong, Yizhou Yu, Zhiwen Liu, Yaou Liu, Chuyang Ye

Abstract: Highlights•We proposed a data augmentation approach CarveMix for brain lesion segmentation.•CarveMix mixes pairs of annotated images to generate synthetic training images.•The image mixing is performed according to the location and shape of the lesions.•CarveMix was validated on multiple public and private datasets.•The results show that CarveMix improves the quality of brain lesion segmentation.