PhaseAT: Fourier Phase Adversarial Training for Medical Image Domain Generalization

Excerpt

arXiv:2610.01807v1 Announce Type: cross Abstract: Reliable clinical deployment of deep medical image models is hindered by distribution shifts across scanners, sites, and acquisition protocols. Existing domain generalization (DG) methods often focus on style or intensity diversification, but they can still leave networks dependent on domain-specific texture correlations. Inspired by evidence that Fourier phase encodes semantic structure, we introduce PhaseAT, a phase-aware adversarial training f