Summary of Cross-task Pretraining For Cross-organ Cross-scanner Adenocarcinoma Segmentation, by Adrian Galdran
Cross-Task Pretraining for Cross-Organ Cross-Scanner Adenocarcinoma Segmentation
by Adrian Galdran
First submitted to arxiv on: 20 Sep 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 proposed solution addresses the COSAS 2024 competition’s challenge in segmenting adenocarcinoma from histopathological image patches, focusing on cross-organ and cross-scanner segmentation. The primary issue lies in the noticeable domain shift when switching microscopes or organs. To tackle this problem, three strategies were tested: standard training for each dataset, pretraining on one dataset and fine-tuning on another (Cross-Task Pre-training), and combining both datasets. Results showed that Cross-Task Pre-training is a more effective approach to bridging the domain shift gap. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The solution is designed to help doctors diagnose cancer better by recognizing different types of cancer cells in images taken from different organs and microscopes. The problem is that these images can look very different when taken with different equipment, making it hard for computers to learn what each type of cell looks like. To solve this problem, the researchers tested three ways to train a computer model: training on one set of images at a time, training on all the images first and then fine-tuning on individual sets, or combining all the images together. The best approach was to use all the images at once. |
Keywords
» Artificial intelligence » Fine tuning » Pretraining
