Universal Frequency Domain Perturbation for Single-Source Domain Generalization

liu chuang, Yichao Cao, Haogang Zhu, Xiu Su

Abstract: In this work, we introduce a novel approach to single-source domain generalization (SDG) in medical imaging, focusing on overcoming the challenge of style variation in out-of-distribution (OOD) domains without requiring domain labels or additional generative models. We propose a \textbf{Uni}versal \textbf{Freq}uency Perturbation framework for \textbf{SDG} termed as \textit{\textbf{UniFreqSDG}}, that performs hierarchical feature-level frequency domain perturbations, facilitating the model's ability to handle diverse OOD styles. Specifically, we design a learnable spectral perturbation module that adaptively learns the frequency distribution range of samples, allowing for precise low-frequency (LF) perturbation. This adaptive approach not only generates stylistically diverse samples but also preserves domain-invariant anatomical features without the need for manual hyperparameter tuning. Then, the frequency features before and after perturbation are decoupled and recombined through the Content Preservation Reconstruction operation, effectively preventing the loss of discriminative content information. Furthermore, we introduce the Active Domain-variance Inducement Loss to encourage effective perturbation in the frequency domain while ensuring the sufficient decoupling of domain-invariant and domain-style features. Extensive experiments demonstrate that \textit{\textbf{UniFreqSDG}} increases the dice score by an average of 7.47\% (from 77.98\% to 85.45\%) on the fundus dataset and 4.99\% (from 71.42\% to 76.73\%) on the prostate dataset compared to the state-of-the-art approaches.

Primary Subject Area: [Experience] Multimedia Applications

Relevance To Conference: Our submission to the ACM Multimedia Conference (ACMMM) introduces a pioneering approach to medical imaging with an adaptive frequency domain perturbation method, directly aligning with ACMMM’s focus on Multimedia Applications. This work enhances adaptability and robustness in handling style variations in out-of-distribution domains, a challenge central to multimedia research. Demonstrated improvements in medical imaging, not only set new benchmarks in this domain but also highlight the practical implications of our research. This alignment with the ACMMM’s mission to promote innovative research in multimedia systems and applications ensures our work’s relevance and potential impact on the field.

Submission Number: 4543