Summary of Mitigating Annotation Shift in Cancer Classification Using Single Image Generative Models, by Marta Buetas Arcas et al.
Mitigating annotation shift in cancer classification using single image generative models
by Marta Buetas Arcas, Richard Osuala, Karim Lekadir, Oliver Díaz
First submitted to arxiv on: 30 May 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 study aims to develop a high-accuracy cancer risk prediction model for breast mammography, which is capable of distinguishing benign from malignant lesions. To mitigate the impact of annotation shifts in multiclass classification performance, particularly for malignant lesions, the authors propose a training data augmentation approach based on single-image generative models. This approach requires only four in-domain annotations to significantly reduce the effects of annotation shift and addresses dataset imbalance. Additionally, an ensemble architecture is developed by training multiple models under different data augmentation regimes, further improving performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study uses artificial intelligence (AI) to help radiologists detect and diagnose breast cancer more accurately. The goal is to develop a reliable model that can identify benign or malignant lesions in mammography images. However, the quality of available data affects AI’s performance, and limited annotation procedures lead to shifts in how data is labeled. This research explores these challenges by developing a risk prediction model that can handle these issues. By using single-image generative models, the authors show that only a few annotations are needed to improve AI’s accuracy. |
Keywords
» Artificial intelligence » Classification » Data augmentation
