RelativeFlow: Taming Medical Image Denoising Learning with Noisy Reference

2026-04-20 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Medical Imaging · Depth: Expert, medium

Summary

RelativeFlow is a novel flow matching framework designed to overcome the "noisy reference problem" in medical image denoising (MID), where absolutely clean images are unavailable for supervision. Existing methods like Simulated-Supervised Discriminative Learning (SimSDL), Self-Supervised Learning (SSL), and Simulated-Supervised Generative Learning (SimSGL) either treat noisy references as clean targets, leading to suboptimal convergence or reference-biased learning, or impose restrictive noise assumptions. RelativeFlow reformulates flow matching by decomposing the absolute noise-to-clean mapping into relative noisier-to-noisy mappings. It achieves this through two main components: Consistent Transport (CoT), a displacement map that ensures relative flows compose a unified absolute flow, and Simulation-based Velocity Field (SVF), which constructs a learnable velocity field using modality-specific degradation operators. Experiments on Computed Tomography (CT) and Magnetic Resonance (MR) denoising demonstrate that RelativeFlow significantly outperforms current methods.

Key takeaway

For Computer Vision Engineers developing medical image denoising solutions, RelativeFlow offers a robust approach to overcome the limitations of noisy training data. By reformulating flow matching with Consistent Transport and Simulation-based Velocity Fields, you can achieve superior denoising performance on modalities like CT and MR, even when only heterogeneous noisy references are available. Consider integrating this framework to improve the consistency and quality of your denoised medical images.

Key insights

RelativeFlow uses flow matching to learn unified denoising from heterogeneous noisy medical image references.

Principles

• Decompose absolute noise-to-clean mapping into relative noisier-to-noisy mappings.
• Constrain relative flows to progressively compose a unified absolute flow.
• Utilize modality-specific degradation operators for velocity field construction.

Method

RelativeFlow employs Consistent Transport (CoT) to unify relative flows and a Simulation-based Velocity Field (SVF) based on medical image degradation processes to learn denoising across modalities like CT and MR.

In practice

• Apply RelativeFlow to CT denoising tasks.
• Implement RelativeFlow for MR image denoising.
• Use provided code for training and evaluation.

Topics

• Medical Image Denoising
• Noisy Reference Problem
• RelativeFlow Framework
• Flow Matching
• Consistent Transport

Code references

• Deliver0/RelativeFlow

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• RelativeFlow: Taming Medical Image Denoising Learning with Noisy Reference — RelativeFlow improves medical image denoising by learning relative noise mappings from heterogeneous noisy references.
• RA-CMF: Region-Adaptive Conditional MeanFlow for CT Image Reconstruction — RA-CMF enhances CT image quality and radiomic consistency by adaptively refining heterogeneous regions using conditional MeanFlow and reinforcement learning.
• Test-Time Adaptation in Optical Coherence Tomography Using Trajectory-Aligned Time-Independent Flow — The method uses flow-matching and histogram alignment for test-time adaptation to denoise OCT images, improving AMD biomarker segmentation.
• Multi-Contrast MRI Motion Correction via Parameter-Informed Disentanglement and Adaptive Experts — A unified MRI motion correction framework disentangles contrast and adapts to severity for robust, generalizable artifact removal.
• Look-Closer-Then-Diagnose: Confidence-Aware Ultrasound VQA via Active Zooming — A novel VLM framework integrates lesion-focused reasoning and uncertainty-aware rewards for improved ultrasound diagnosis.
• Flow Map Denoisers: Traversing the Distortion-Perception Plane for Inverse Problems — Flow Map Denoisers use a single parameter to continuously balance distortion and perception in image restoration and inverse problems.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.