基于HU的前景掩蔽用于3D医学掩蔽图像建模
HU-based Foreground Masking for 3D Medical Masked Image Modeling
AI总结：
针对3D医学图像建模中随机掩蔽忽略解剖密度的问题，提出基于HU值的前景掩蔽策略，在五个公开数据集上提升分割Dice分数，验证了领域中心MIM的有效性。
摘要：
虽然掩蔽图像建模（MIM）已经彻底改变了计算机视觉领域，但其在3D医学图像计算中的应用一直受到随机掩蔽的限制，随机掩蔽忽略了解剖对象的密度。为了解决这一限制，我们通过一种简单而有效的掩蔽策略增强了预文本任务。利用亨斯菲尔德单位（HU）测量，我们实现了基于HU的前景掩蔽，该掩蔽聚焦于内脏器官的强度分布，并排除缺乏诊断意义特征的空气和液体等非组织区域。在五个公开的3D医学影像数据集上的大量实验表明，我们的掩蔽方法在分割质量和Dice分数方面持续提高了性能（BTCV：~84.64%，Flare22：~92.43%，MM-WHS：~90.67%，Amos22：~88.64%，BraTS：~78.55%）。这些结果强调了以领域为中心的MIM的重要性，并为医学图像分割中的表示学习指明了一个有前景的方向。实现可在github.com/AISeedHub/SubFore/获取。
英文摘要：
While Masked Image Modeling (MIM) has revolutionized fields of computer vision, its adoption in 3D medical image computing has been limited by the use of random masking, which overlooks the density of anatomical objects. To address this limitation, we enhance the pretext task with a simple yet effective masking strategy. Leveraging Hounsfield Unit (HU) measurements, we implement an HU-based Foreground Masking, which focuses on the intensity distribution of visceral organs and excludes non-tissue regions, such as air and fluid, that lack diagnostically meaningful features. Extensive experiments on five public 3D medical imaging datasets demonstrate that our masking consistently improves performance, both in quality of segmentation and Dice score (BTCV:~84.64\%, Flare22:~92.43\%, MM-WHS:~90.67\%, Amos22:~88.64\%, BraTS:~78.55\%). These results underscore the importance of domain-centric MIM and suggest a promising direction for representation learning in medical image segmentation. Implementation is available at github.com/AISeedHub/SubFore/.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
