BSDA: Bayesian Random Semantic Data Augmentation for Medical Image Classification

Yaoyao Zhu, Xiuding Cai, Xueyao Wang, Xiaoqing Chen, Zhongliang Fu, Yu Yao · 2024

Read BSDA: Bayesian Random Semantic Data Augmentation for Medical Image Classification by Yaoyao Zhu, Xiuding Cai, Xueyao Wang, Xiaoqing Chen, Zhongliang Fu, Yu Yao (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-11644067.txt.

Description

Medical image classifiers starve for data, and one remedy is to shift images in feature space, inventing new examples with the same label. Yaoyao Zhu and co-authors warn that shifting too far silently changes the label. Their fix, BSDA, treats the shift amount with Bayesian care, sampling it in a way that respects the data's own spread.

The module plugs into ordinary networks without surgery. Across two- and three-dimensional datasets and several imaging types, it beat rival augmentation schemes and lifted baseline models it was attached to. For small, precious medical datasets, a principled nudge beats a random one.

Catalog record

CREATORYaoyao Zhu, Xiuding Cai, Xueyao Wang, Xiaoqing Chen, Zhongliang Fu, Yu Yao
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
VERSION1.0
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC11644067

Browse the full library · About the Tower of Babel