Organizations: ECE Department, UCLA, 405 Hilgard Ave., Los Angeles, California 90095, USA · Pinpoint Photonics, Inc., 803 Marine Bldg., 4-23 Kaigan-dori, Naka, Yokohama 231-0002, Japan · Adventure Photonics, 889 N. Douglas Ave, El Segundo, California 90245, USA
Deep learning has achieved remarkable success in medical image analysis, yet its performance remains highly sensitive to the heterogeneity of clinical data. Differences in imaging hardware, staining protocols, and acquisition conditions produce substantial domain shifts that degrade model generalization across institutions. Here we present a physics-based data preprocessing framework based on the PhyCV (Physics-Inspired Computer Vision) family of algorithms, which standardizes medical images through deterministic transformations derived from optical physics. The framework models images as spatially varying optical fields that undergo a virtual diffractive propagation followed by coherent phase detection. This process suppresses non-semantic variability such as color and illumination differences while preserving diagnostically relevant texture and structural features. When applied to histopathological images from the Camelyon17-WILDS benchmark, PhyCV preprocessing improves out-of-distribution breast-cancer classification accuracy from 70.8% (Empirical Risk Minimization baseline) to 90.9%, matching or exceeding data-augmentation and domain-generalization approaches at negligible computational cost. Because the transform is physically interpretable, parameterizable, and differentiable, it can be deployed as a fixed preprocessing stage or integrated into end-to-end learning. These results establish PhyCV as a generalizable data refinery for medical imaging-one that harmonizes heterogeneous datasets through first-principles physics, improving robustness, interpretability, and reproducibility in clinical AI systems.
Figures & tables
Figure 1: Examples of physics-inspired computer vision algorithms in PhyCV. The PhyCV framework includes modules for diverse imaging tasks: super-resolution (PhSAR) enhances structural detail in low-resolution MRI scans; edge and orientation extraction (PAGE) reveals vascular and textural organization in retinal imagery; and exposure enhancement (VEViD) restores visibility in low-contrast microscopy images. Together, these algorithms demonstrate the versatility of physics-based models for image refinement and feature extraction across modalities.
Figure 2: Conceptual overview of PhyCV for data refinement and standardization. During training (top), heterogeneous data from multiple hospitals are refined into standardized feature representations before being used to train a neural network. During inference (bottom), unseen data undergo the same refinement to ensure consistent predictions.
Figure 3: PhyCV improves robustness under non-uniform illumination. Top: Non-uniform illumination is studied by considering 6 levels of illumination (linear) on a single induced pluripotent stem cell (iPS) test image picture from [ 16 ] . For each of the 6 illumination levels, center region is used for texture analysis. Bottom: corresponding image entropy showing that PhyCV maintains information content even under severe illumination degradation. When information is fully lost in the case of image 6, there is a noteacble loss in the PhyCV output, demonstrating that it can act as a delineator for defective or untrustworthy images.
Figure 4: Cross-institutional variability and PhyCV-based refinement. Example histopathology patches from five hospitals showing differences in staining and contrast (top). After PhyCV refinement (bottom), features are standardized, enhancing consistency across sites while preserving tissue structure.
Method
OOD Validation
OOD Test
ERM (baseline)
85.8(1.9)
70.8(7.2)
PAIR
84.3(1.6)
74.0(7.2)
Fish
83.9(1.2)
74.7(7.1)
LISA
81.8(1.2)
77.1(6.9)
ERM + target augmentation
92.7(0.7)
92.1(3.1)
ContriMix
91.9(0.7)
94.6(1.2)
Table 1: Out-of-distribution classification accuracy (%) on the Camelyon17-WILDS dataset. Mean ± standard deviation computed across 10 random seeds. All methods use the same network (DenseNet-121) and data split for fair comparison.
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 framework for medical DG. PhaseAT forms phase-perturbed training views in the Fourier domain by iteratively updating a bounded phase perturbation while keeping the amplitude spectrum unchanged, thereby stressing spatial organization under matched appearance statistics. Perturbations are applied only to the luminance channel in YCbCr color space to avoid chromatic artifacts. Additionally, a simple phase-saliency mask concentrates updates on the most influential frequencies. The model is trained with a weighted combination of losses on clean and phase-perturbed samples, supporting both single-source and multi-source DG. We validate our method on two challenging medical datasets and demonstrate that PhaseAT achieves over 20% improvement in single-source domain generalization, outperforming several state-of-the-art DG methods. The code implementation is available at: https://github.com/ahmed-sharshar/PhaseAT.
Ahmed Sharshar, Asif Hanif, Naveen Kumar Kummari +2
Division of Computing and Mathematical Sciences Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, UAE
Cross-site generalizability in medical AI is fundamentally compromised by selection bias, a structural mechanism where patient demographics (e.g., age, severity) non-randomly dictate hospital assignment. Conventional Domain Generalization (DG) paradigms, which predominantly target image-level distribution shifts, fail to address the resulting spurious correlations between site-specific variations and diagnostic labels. To surmount this identifiability barrier, we propose CIV-DG, a causal framework that leverages Conditional Instrumental Variables to disentangle pathological semantics from scanner-induced artifacts. By relaxing the strict random assignment assumption of standard IV methods, CIV-DG accommodates complex clinical scenarios where hospital selection is endogenously driven by patient demographics. We instantiate this theory via a Deep Generalized Method of Moments (DeepGMM) architecture, employing a conditional critic to minimize moment violations and enforce instrument-error orthogonality within demographic strata. Extensive experiments on the Camelyon17 benchmark and large-scale Chest X-Ray datasets demonstrate that CIV-DG significantly outperforms leading baselines, validating the efficacy of conditional causal mechanisms in resolving structural confounding for robust medical AI.
Shaojin Bai, Yuting Su, Weizhi Nie
School of Electrical and Information Engineering, Tianjin University, Tianjin, China
Hardware shifts, color variations, and changing patient characteristics between development and deployment routinely break trained medical image classifiers. Existing remedies fall short: standard color jittering provides insufficient diversity, while deep generative style transfer algorithms hallucinate features, destroy clinically relevant structures, and waste massive compute resources. To address this, we revisit classical statistical color matching and repurpose it as Colorist, a highly efficient data augmentation strategy that applies global mean-standard deviation matching directly in the RGB color space. We demonstrate that this training-free, fully interpretable approach safely generates structurally intact domain variations, outperforming deep generative models in structural fidelity and color alignment. Across out-of-distribution histopathology, peripheral blood, dermatology, and retinal datasets, it improves balanced accuracy by up to +9% over state-of-the-art domain generalization regularizers and by +13% over an unaugmented baseline. Moreover, by avoiding neural networks in the augmentation loop, Colorist preserves anatomical structure, minimizes carbon footprint, and integrates seamlessly into standard dataloaders. Together, these findings establish statistical matching as a safe, interpretable, yet overlooked alternative to deep architectures for clinical robustness. Source code is available at https://github.com/sdoerrich97/colorist.
Sebastian Doerrich, Francesco Di Salvo, Shyam Nandan Rai +2
xAILab Bamberg, University of Bamberg, Bamberg, Germany