Summary of Constructing and Exploring Intermediate Domains in Mixed Domain Semi-supervised Medical Image Segmentation, by Qinghe Ma et al.
Constructing and Exploring Intermediate Domains in Mixed Domain Semi-supervised Medical Image Segmentation
by Qinghe Ma, Jian Zhang, Lei Qi, Qian Yu, Yinghuan Shi, Yang Gao
First submitted to arxiv on: 13 Apr 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
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 proposed Mixed Domain Semi-supervised medical image Segmentation (MiDSS) addresses the challenges of limited annotation and domain shift in medical image segmentation. This novel scenario involves handling data from multiple medical centers with limited annotations available for a single domain and a large amount of unlabeled data from multiple domains. The authors employ Unified Copy-Paste (UCP) between images to construct intermediate domains, facilitating knowledge transfer from labeled data to unlabeled data. A symmetric Guidance training strategy (SymGD) is proposed to merge pseudo labels from intermediate samples, offering direct guidance to unlabeled data. Additionally, a Training Process aware Random Amplitude MixUp (TP-RAM) incorporates style-transition components into intermediate samples. Compared to state-of-the-art approaches, the method achieves a 13.57% improvement in Dice score on the Prostate dataset, as demonstrated on three public datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary MiDSS is a new way to solve medical image segmentation problems. This method helps doctors and computers work together better by using images from many different places. The computer gets help from some labeled pictures, but most of the time it has to figure things out on its own. To make this happen, MiDSS uses three cool techniques: Unified Copy-Paste (UCP), symmetric Guidance training strategy (SymGD), and Training Process aware Random Amplitude MixUp (TP-RAM). These tools help the computer understand what’s important in an image and get better at identifying different parts of a picture. This method is really good, with a 13.57% improvement on some public datasets. |
Keywords
* Artificial intelligence * Image segmentation * Semi supervised
