[论文解读] Polyp-SAM: Transfer SAM for Polyp Segmentation

Yuheng Li, Mingzhe Hu+1 more|arXiv (Cornell University)|Apr 29, 2023

Colorectal Cancer Screening and Detection被引用 15

一句话总结

Polyp-SAM 对 Segment Anything Model (SAM) 进行微调以用于结肠息肉分割，并比较两种迁移策略，在若干数据集上实现了最先进的结果，Dice 分数超过 88%。

ABSTRACT

Colon polyps are considered important precursors for colorectal cancer. Automatic segmentation of colon polyps can significantly reduce the misdiagnosis of colon cancer and improve physician annotation efficiency. While many methods have been proposed for polyp segmentation, training large-scale segmentation networks with limited colonoscopy data remains a challenge. Recently, the Segment Anything Model (SAM) has recently gained much attention in both natural and medical image segmentation. SAM demonstrates superior performance in several image benchmarks and therefore shows great potential for medical image segmentation. In this study, we propose Poly-SAM, a finetuned SAM model for polyp segmentation, and compare its performance to several state-of-the-art polyp segmentation models. We also compare two transfer learning strategies of SAM with and without finetuning its encoders. Evaluated on five public datasets, our Polyp-SAM achieves state-of-the-art performance on two datasets and impressive performance on three datasets, with dice scores all above 88%. This study demonstrates the great potential of adapting SAM to medical image segmentation tasks. We plan to release the code and model weights for this paper at: https://github.com/ricklisz/Polyp-SAM.

研究动机与目标

• 促进自动息肉分割，以减少误诊并提高结肠镜检查中的标注效率。
• 通过在息肉数据上进行微调，利用 Segment Anything Model (SAM) 进行医学图像分割。
• 比较对 SAM 的迁移学习策略：有无编码器微调。
• 在多个公开息肉数据集上评估 Polyp-SAM 以评估泛化能力。

提出的方法

• 对 SAM 进行息肉分割的微调（Polyp-SAM）。
• 比较两种迁移学习策略：带编码器微调的 SAM 与不微调编码器的 SAM。
• 在五个公开息肉数据集上进行评估，Dice 分数为主要指标。
• 与最先进的息肉分割模型进行比较。

实验结果

研究问题

• RQ1 是否可以通过微调有效地将 SAM 适配到息肉分割？
• RQ2 哪种迁移策略（微调编码器 vs. 固定编码器）在息肉分割性能上表现更好？
• RQ3 与现有方法相比，Polyp-SAM 在多样化的公开数据集上的表现如何？

主要发现

• Polyp-SAM 在两个数据集上达到最先进的性能。
• Polyp-SAM 在三个数据集上显示出令人印象深刻的性能。
• 在评估的数据集上 Dice 分数超过 88%。