StyleGAN-based synthetic image augmentation for multi-class otoscopy image classification

Seda Camalan, Carl D. Langefeld, Amy Zinnia, Aaron C. Moberly, Metin N. Gurcan · 2026

Read StyleGAN-based synthetic image augmentation for multi-class otoscopy image classification by Seda Camalan, Carl D. Langefeld, Amy Zinnia, Aaron C. Moberly, Metin N. Gurcan (2026) 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/PMC13500606.txt.

Description

Seda Camalan and co-authors test in 2026 whether synthetic eardrum images generated by an artificial intelligence image maker can train better diagnostic classifiers. Examining the eardrum with an otoscope is central to spotting infections and perforations, yet labelled image sets are small and grading is subjective. The team created artificial and composite images with a StyleGAN3 generator and trained a deep image network. Comparisons ran against classic augmentation tricks such as flipping and rotating, and against no augmentation at all.

Synthetic augmentation lifted mean test accuracy into the mid-ninety percent range with a matching F1 score in the reported runs, ahead of classic augmentation and well ahead of the unaugmented baseline. The authors caution that their setup represents an optimistic upper bound, since generated images may leak information about the test distribution. The gap suggests machine-made training images can relieve the chronic data scarcity in ear imaging.

Catalog record

CREATORSeda Camalan, Carl D. Langefeld, Amy Zinnia, Aaron C. Moberly, Metin N. Gurcan
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE70725
LICENSEhttps://creativecommons.org/licenses/by/4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/article/MED/PMC13500606

Browse the full library · About the Tower of Babel