MedDiffuseMix: Preserving Diagnostic Evidence with Saliency-Aware Diffusion Medical Image Data Augmentation
Teerath Kumar, Raja Vavekanand, Muhammad Turab · 2026
Read MedDiffuseMix: Preserving Diagnostic Evidence with Saliency-Aware Diffusion Medical Image Data Augmentation by Teerath Kumar, Raja Vavekanand, Muhammad Turab (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/arxiv-2606-28419.txt.
Description
Teerath Kumar, Raja Vavekanand, and Muhammad Turab augment scarce medical images without ruining the evidence doctors read. Ordinary augmentation can warp the very structures that carry a diagnosis, and free-running generative models may invent content that clashes with the labels. Their MedDiffuseMix framework mixes images under the guidance of saliency maps that mark what matters.
A classifier's saliency map splits each image into diagnostic foreground and background. Diffusion-guided mixing then acts mainly on the background, with adaptive strength, smooth blending at the boundary, and a preservation rule that rejects samples dragging the model's attention away from real evidence. Tested on four public benchmarks, from pneumonia chest images to breast tissue slides, and across CNN and transformer classifiers, the method improved accuracy and related scores over standard and generative rivals.
Catalog record
| CREATOR | Teerath Kumar, Raja Vavekanand, Muhammad Turab |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.28419