IHF-Harmony: Multi-Modality Magnetic Resonance Images Harmonization using Invertible Hierarchy Flow Model

Pengli Zhu; Yitao Zhu; Haowen Pang; Anqi Qiu

cs.CVFeb 25, 2026

IHF-Harmony: Multi-Modality Magnetic Resonance Images Harmonization using Invertible Hierarchy Flow Model

Authors: Pengli Zhu, Yitao Zhu, Haowen Pang, Anqi Qiu

Organizations: Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong · School of Integrated Circuits and Electronics, Beijing Institute of Technology, China · Mental Health Research Center, The Hong Kong Polytechnic University, Hong Kong · Department of Biomedical Engineering, Johns Hopkins University, USA

Abstract

Retrospective MRI harmonization is limited by poor scalability across modalities and reliance on traveling subject datasets. To address these challenges, we introduce IHF-Harmony, a unified invertible hierarchy flow framework for multi-modality harmonization using unpaired data. By decomposing the translation process into reversible feature transformations, IHF-Harmony guarantees bijective mapping and lossless reconstruction to prevent anatomical distortion. Specifically, an invertible hierarchy flow (IHF) performs hierarchical subtractive coupling to progressively remove artefact-related features, while an artefact-aware normalization (AAN) employs anatomy-fixed feature modulation to accurately transfer target characteristics. Combined with anatomy and artefact consistency loss objectives, IHF-Harmony achieves high-fidelity harmonization that retains source anatomy. Experiments across multiple MRI modalities demonstrate that IHF-Harmony outperforms existing methods in both anatomical fidelity and downstream task performance, facilitating robust harmonization for large-scale multi-site imaging studies. Code is available at https://github.com/Idea89560041/IHF-Harmony.

Figures & tables

Figure 1: Overview of IHF-Harmony. a , IHF-Harmony works in an invertible manner. The blue arrows indicate the forward pass for feature extraction, while the red arrows denote the backward pass for image reconstruction. b , invertible hierarchy flow (IHF) performs hierarchical channel-wise subtraction to enable learnable spatial feature transformation and multi-scale fusion. c , artefact-aware normalization (AAN) aligns source feature statistics to the target template using anatomy-guided affine parameters.

Figure 2: Multi-modality MRI harmonization across vendors. Representative results for a , T1-weighted images, b , T2-weighted images, and c–d , diffusion-weighted images. The leftmost column displays original unharmonized images, followed by results harmonized to GE, Siemens, and Philips scanners, respectively.

Figure 3: Numerical evaluation of harmonization outcomes in structural MRI. a–b , visual inspection and histogram distribution comparison for multi-site harmonization. c , quantitative comparison of different harmonization methods. Multiple metrics were used for evaluation (* p $<$ 0.0001, two-sided $t$ -test) and mean values are marked by bullseye symbols and standard errors by error bars.

Figure 4: Validation of diffusion MRI harmonization via parameter fitting consistency. The left and right panels show parameter maps used for qualitative and quantitative analysis of scanner- and channel-specific harmonization results, respectively. (B/S: scanner; 20/64ch: channel)

	RMSE $\downarrow$	MS-SSIM $\uparrow$	PSNR $\uparrow$	LPIPS $\downarrow$
w/o ${\cal L}_{\text{ac}}$ & ${\cal L}_{\text{art}}$	0.4196 $\pm$ 0.0012	0.5844 $\pm$ 0.0043	7.54 $\pm$ 0.026	0.4761 $\pm$ 0.0033
w/o ${\cal L}_{\text{art}}$	0.1157 $\pm$ 0.0012	0.8844 $\pm$ 0.0012	18.75 $\pm$ 0.09	0.1416 $\pm$ 0.0014
w/o ${\cal L}_{\text{ac}}$	0.0935 $\pm$ 0.0007	0.9079 $\pm$ 0.0011	20.58 $\pm$ 0.06	0.0570 $\pm$ 0.0009
Full model	0.0612 $\pm$ 0.0004	0.9690 $\pm$ 0.0004	24.59 $\pm$ 0.05	0.0406 $\pm$ 0.0009

Table 1: Ablation results of IHF-Harmony with different losses

Explore similar work

A Target-Free Harmonization Method for MRI
May 2, 2026Minjun Kim, Dong Ju Mun, Hwihun Jeong +4Medical Image Analysis Magnetic Resonance Imaging
Department of Electrical and Computer Engineering, Seoul National University, Seoul, Republic of Korea · Department of Electrical and Computer Engineering, Seoul National University, Seoul, 08826, Republic of Korea · Department of Psychiatry, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA +2
HARP: HARmonizing in-vivo diffusion MRI using Phantom-only training
Mar 5, 2026Hwihun Jeong, Qiang Liu, Kathryn E. Keenan +7Diffusion MRI Medical Image Analysis
Department of Psychiatry, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, United States of America · College of Engineering, Northeastern University, Boston, Massachusetts, United States of America · Physical Measurement Laboratory, National Institute of Standards and Technology, Boulder, Colorado, United States of America +6
UniField: A Unified Field-Aware MRI Enhancement Framework
Mar 10, 2026Yiyang Lin, Chenhui Wang, Zhihao Peng +1Medical Image Super-Resolution Magnetic Resonance Imaging
Department of Electronic Engineering, the Chinese University of Hong Kong · The Institute of Science and Technology for Brain-inspired Intelligence, Fudan University

May 2, 2026J/K move · Enter open · S save

A Target-Free Harmonization Method for MRI

Minjun Kim, Dong Ju Mun, Hwihun Jeong, Hangyeol Park+3

Department of Electrical and Computer Engineering, Seoul National University, Seoul, Republic of Korea · Department of Electrical and Computer Engineering, Seoul National University, Seoul, 08826, Republic of Korea · Department of Psychiatry, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA+2

Medical Image Analysis Magnetic Resonance Imaging Brain MRI

Open paper AskPDF arXiv

In MRI, variations in scan parameters, sequence, or hardware can lead to discrepancies in image appearance, even for the same subject. These inconsistencies, known as domain shifts, can hinder image analysis and degrade the performance of deep learning models trained on data from specific target domains. MRI image harmonization aims to address these issues by aligning source domain images to the target domain images while preserving biological information such as anatomical structures. However, most existing harmonization approaches require access to both source and target domain data in training or test time. This dependence induces data sharing between institutions, raising concerns about patient privacy and substantially limiting the harmonization approaches that can be practically deployed in clinical settings. To overcome these limitations, we introduce TgtFreeHarmony, the harmonization framework tailored for target-free scenarios, eliminating the need for target domain data and any data sharing, enabling privacy-preserving harmonization directly within the source institution. Our approach estimates the target domain style by searching the manifold of MRI domain style constructed via a disentanglement-based generator using Bayesian optimization guided by the performance of a downstream task model, which is trained on target domain data. We evaluated our method on the brain tissue segmentation task across multiple institutes and demonstrated that it effectively harmonizes source images into target images, leading to improved downstream task performance. By enabling harmonization without any access to target-domain data, TgtFreeHarmony establishes a new direction of harmonization preserving data privacy that can be realistically deployed within clinical environments.

Abstract

Explore similar work

A Target-Free Harmonization Method for MRI

HARP: HARmonizing in-vivo diffusion MRI using Phantom-only training

UniField: A Unified Field-Aware MRI Enhancement Framework