Search papers, labs, and topics across Lattice.
The paper introduces IHF-Harmony, a novel invertible hierarchy flow framework for multi-modality MRI harmonization using unpaired data, addressing limitations in scalability and reliance on traveling subjects. IHF-Harmony employs hierarchical subtractive coupling to remove artifact-related features and anatomy-fixed feature modulation via artefact-aware normalization (AAN) to transfer target characteristics, ensuring bijective mapping and lossless reconstruction. Experiments demonstrate that IHF-Harmony outperforms existing methods in anatomical fidelity and downstream task performance across multiple MRI modalities.
Achieve high-fidelity MRI harmonization without anatomical distortion using invertible hierarchical flows, even with unpaired data.
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 will be released upon acceptance.
