Summary of Promise: Promptable Medical Image Segmentation Using Sam, by Jinfeng Wang et al.
ProMISe: Promptable Medical Image Segmentation using SAM
by Jinfeng Wang, Sifan Song, Xinkun Wang, Yiyi Wang, Yiyi Miao, Jionglong Su, S. Kevin Zhou
First submitted to arxiv on: 7 Mar 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The Segment Anything Model (SAM) is a popular foundation model for medical image segmentation (MIS). However, fine-tuning SAM for MIS comes with significant costs, including instability and feature damage. Moreover, methods that transfer SAM to the target domain often disable its prompting capability, limiting its utilization scenarios. To address these issues, we propose an Auto-Prompting Module (APM) that provides SAM-based foundation models with Euclidean adaptive prompts in the target domain. Our experiments show that APM significantly improves SAM’s non-fine-tuned performance in MIS. Additionally, we introduce Incremental Pattern Shifting (IPS), a novel method for adapting SAM to specific medical domains without fine-tuning. Experimental results demonstrate that IPS enables SAM to achieve state-of-the-art or competitive performance in MIS with frozen parameters. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making it easier to use a powerful tool called the Segment Anything Model (SAM) to analyze medical images. Currently, this requires fine-tuning the model for each specific task, which can be time-consuming and lead to mistakes. The authors propose two new methods: one that helps SAM understand what it’s supposed to do in a given image, and another that adjusts SAM to work better with different types of medical images. They test these methods and find that they work well without requiring the model to be fine-tuned for each specific task. |
Keywords
» Artificial intelligence » Fine tuning » Image segmentation » Prompting » Sam
