Summary of Dimensionality Reduction and Nearest Neighbors For Improving Out-of-distribution Detection in Medical Image Segmentation, by Mckell Woodland et al.
Dimensionality Reduction and Nearest Neighbors for Improving Out-of-Distribution Detection in Medical Image Segmentation
by McKell Woodland, Nihil Patel, Austin Castelo, Mais Al Taie, Mohamed Eltaher, Joshua P. Yung, Tucker J. Netherton, Tiffany L. Calderone, Jessica I. Sanchez, Darrel W. Cleere, Ahmed Elsaiey, Nakul Gupta, David Victor, Laura Beretta, Ankit B. Patel, Kristy K. Brock
First submitted to arxiv on: 5 Aug 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 abstract describes a study that aims to address the issue of clinically deployed deep learning-based segmentation models failing on data outside their training distributions. The problem is exacerbated by clinicians reviewing segmentations, which can lead to automation bias. To mitigate this issue, the authors propose applying the Mahalanobis distance (MD) post hoc to the bottleneck features of four segmentation models that segmented liver images. They also explore a non-parametric alternative, k-th nearest neighbors distance (KNN), and demonstrate its improved scalability and performance over MD when applied to raw and average-pooled bottleneck features. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The study focuses on improving deep learning-based segmentation models by detecting out-of-distribution images at inference time. This is important because current models tend to perform well most of the time, but can fail outside their training distributions. The authors use a combination of techniques including principal component analysis and uniform manifold approximation and projection to reduce the dimensions of bottleneck features and detect failed images with high performance and minimal computational load. |
Keywords
» Artificial intelligence » Deep learning » Inference » Principal component analysis
