SAFFusion: a saliency-aware frequency fusion network for multimodal medical image fusion
Biomedical Optics Express · 01.06.2025 · 16 (6) · 2459
DOI: 10.1364/boe.555458 ↗Kurz gefasst
Fachartikel in der Zeitschrift Biomedical Optics Express, online erschienen am 27.05.2025. Verfasst von Renhe Liu und 4 weiteren Autor:innen. Beteiligte Einrichtung: School of Microelectronics, Tianjin University, Tianjin 300072, China. Die Arbeit wurde bisher 14-mal zitiert (Crossref) und nennt 44 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: Natural Science Foundation of Tianjin Municipality, Research and Innovation Project for Postgraduates in Tianjin, Introduction of talents and doctoral start-up fund of Tianjin Medical University Cancer Institute and Hospital.
Abstract (Originalsprache)
Medical image fusion integrates complementary information from multimodal medical images to provide comprehensive references for clinical decision-making, such as the diagnosis of Alzheimer’s disease and the detection and segmentation of brain tumors. Although traditional and deep learning-based fusion methods have been extensively studied, they often fail to devise targeted strategies that fully utilize distinct regional or feature-specific information. This paper proposes SAFFusion, a saliency-aware frequency fusion network that integrates intensity and texture cues from multimodal medical images. We first introduce Mamba-UNet, a multiscale encoder-decoder architecture enhanced by the Mamba design, to improve global modeling in feature extraction and image reconstruction. By employing the contourlet transform in Mamba-UNet, we replace conventional pooling with multiscale representations and decompose spatial features into high- and low-frequency subbands. A dual-branch frequency feature fusion module then fuses cross-modality information according to the distinct characteristics of these frequency subbands. Furthermore, we apply latent low-rank representation (LatLRR) to assess image saliency and implement adaptive loss constraints to preserve information in salient and non-salient regions. Quantitative results on CT/MRI, SPECT/MRI, and PET/MRI fusion tasks show that SAFFusion outperforms state-of-the-art methods. Qualitative evaluations confirm that SAFFusion effectively merges prominent intensity features and rich textures from multiple sources.
Autor:innen und Institutionen
- School of Microelectronics, Tianjin University, Tianjin 300072, China.
- Tianjin Medical University Cancer Institute & Hospital, Tianjin 300060, China.
Förderung
- Natural Science Foundation of Tianjin Municipality
- Research and Innovation Project for Postgraduates in Tianjin
- Introduction of talents and doctoral start-up fund of Tianjin Medical University Cancer Institute and Hospital
Zitieren
APA
Liu, R., Liu, Y., Wang, H., Li, J., & Hu, K. (2025). SAFFusion: a saliency-aware frequency fusion network for multimodal medical image fusion. Biomedical Optics Express, 16(6), 2459. https://doi.org/10.1364/boe.555458
Vancouver
Liu R, Liu Y, Wang H, Li J, Hu K. SAFFusion: a saliency-aware frequency fusion network for multimodal medical image fusion. Biomedical Optics Express. 2025;16(6):2459. doi:10.1364/boe.555458
BibTeX
@article{Liu2025,
author = {Liu, Renhe and Liu, Yu and Wang, Han and Li, Junxian and Hu, Kai},
title = {SAFFusion: a saliency-aware frequency fusion network for multimodal medical image fusion},
journal = {Biomedical Optics Express},
year = {2025},
volume = {16},
number = {6},
pages = {2459},
doi = {10.1364/boe.555458}
}
EUVIMED