Beyond Target Scores: Measuring Off-Target Drift in Diffusion-Based Medical Image Editing
Quick Answer
The study introduces CIB-Med-1, a benchmark for evaluating off-target drift in medical image editing, revealing that diffusion models can inadvertently alter non-target findings.
Quick Take
A constrained diffusion guidance approach shows improved target progression while significantly reducing off-target drift, demonstrating the need for trajectory-level evaluation in medical imaging.
Key Points
- CIB-Med-1 benchmarks directional pleural effusion editing in chest radiography.
- Constrained diffusion guidance reduces median off-target drift from 0.46 to 0.20.
- Human validation shows stronger agreement with intended progression orderings at τ=0.61.
- Standard evaluation metrics may overlook significant off-target changes in medical imaging.
- The study emphasizes trajectory-level semantic control over endpoint score maximization.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Diffusion models can now edit medical images in visually plausible ways, but the standard evaluation question is too narrow: did the target score increase? In clinical imaging, target findings are entangled with co-morbidities, acquisition effects, and selection bias, so a model can appear successful by changing correlated non-target findings rather than isolating the intended pathology. We introduce CIB-Med-1, a trajectory-level benchmark for controlled biomarker editing in chest radiography. CIB-Med-1 evaluates directional pleural effusion editing through calibrated target progression, inversion rate, and off-target semantic drift over 14 clinically motivated nuisance axes. The benchmark exposes a reward-hacking failure mode in which diffusion editors increase effusion scores while simultaneously altering parenchymal, cardiomediastinal, pleural, chronic, or artifact-related findings. We further present a constrained diffusion guidance baseline that optimizes target progression subject to bounded off-target change. Across held-out radiographs, the constrained editor preserves target progression ($\rho_{\mathrm{trend}}=0.88$ vs. $0.90$ for unconstrained guidance) while reducing median off-target drift from $0.46$ to $0.20$ and 90th-percentile drift from $0.98$ to $0.33$. Drift magnitude tracks empirical target--off-target association, supporting the view that semantic instability is structured rather than incidental. A blinded human validation probe with radiology trainees further shows stronger agreement with intended progression orderings ($\tau=0.61$ vs.\ $0.29$ for Pix2Pix). These results argue that medical image editing should be evaluated as trajectory-level semantic control, not as endpoint score maximization.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2607.16291 [cs.CV] |
| (or arXiv:2607.16291v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2607.16291 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Todd Zhou [view email]
[v1]
Sun, 12 Jul 2026 20:42:16 UTC (7,013 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.