Summary of Label Dropout: Improved Deep Learning Echocardiography Segmentation Using Multiple Datasets with Domain Shift and Partial Labelling, by Iman Islam (1) et al.
Label Dropout: Improved Deep Learning Echocardiography Segmentation Using Multiple Datasets With Domain Shift and Partial Labelling
by Iman Islam, Esther Puyol-Antón, Bram Ruijsink, Andrew J. Reader, Andrew P. King
First submitted to arxiv on: 12 Mar 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); 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 Echocardiography (echo) is a widely used imaging modality for assessing cardiac function. Researchers have proposed deep learning models to automate echo segmentation, but these models need to be robust to various images from different scanners and operators with varying expertise. To achieve this, the models must be trained on multiple diverse datasets. However, training with partially labelled data can lead to shortcut learning, where the model associates label presence with domain characteristics, resulting in a performance drop. The authors propose a novel label dropout scheme to break this link, demonstrating a significant improvement in echo segmentation Dice score (62% and 25%) for two cardiac structures when training on multiple diverse partially labelled datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Doctors use a special tool called echocardiography (echo) to check how well the heart is working. They want to make this process better using computer models, but these models need to work well with many different types of images. To do this, they train the models on lots of different pictures taken from different machines and by people with varying levels of skill. However, when training with incomplete information, the model might get confused and think it’s not doing its job. The researchers came up with a new way to fix this problem, which makes their computer models better at segmenting heart structures. |
Keywords
* Artificial intelligence * Deep learning * Dropout
