QSAIS ‐Fuse: A Quality‐Guided Saliency–Structure and Adaptive Interactive Spiking Framework for Multimodal Medical Image Fusion
International Journal of Imaging Systems and Technology · 2026-07 · 36 (4)
DOI: 10.1002/ima.70423 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Imaging Systems and Technology, online erschienen am 20.07.2026. Verfasst von Phu‐Hung Dinh. Die Arbeit nennt 50 Literaturangaben.
Abstract (Originalsprache)
Multimodal medical image fusion seeks to integrate complementary structural and functional information from imaging modalities such as MRI and PET into a unified and diagnostically informative representation. Existing approaches, including both traditional decomposition‐based methods and deep learning frameworks, face persistent limitations: traditional methods rely on fixed heuristic weighting rules that fail to adapt to complex cross‐modal interactions, while data‐driven models suffer from poor interpretability, heavy training data dependency, and limited generalization across imaging conditions. To address these challenges, this paper proposes QSAIS‐Fuse, a quality‐guided saliency‐structure and adaptive interactive spiking fusion framework for multimodal medical image fusion. QSAIS‐Fuse operates through a structured four‐stage pipeline. First, a Phase‐consistent Entropy‐aware Directional Enhancement (PCED) method selectively strengthens anatomically consistent MRI structures while suppressing noise. Second, a Hybrid Decomposition Method (HDM) explicitly separates each source image into a base component and complementary detail components. Third, an Adaptive Interactive Spiking Network for Detail Component Fusion (AISN‐DCF) models nonlinear cross‐modal interactions through competitive and self‐regulatory spiking dynamics to ensure fine‐detail continuity and perceptual fidelity. Fourth, Quality‐Guided Saliency‐Structure Optimization for Base Component Fusion (QSSO‐BCF) uses the resulting fused detail components to formulate base fusion as a Newton–Raphson‐guided quality optimization problem that adaptively balances saliency and structural contributions to preserve global brightness and energy consistency. Extensive experiments on a widely used public MRI‐PET dataset demonstrate that QSAIS‐Fuse consistently outperforms recent state‐of‐the‐art methods across multiple quantitative metrics, achieving superior performance in structural detail preservation, contrast enhancement, and information fidelity, while maintaining full interpretability without reliance on training data.
Zitieren
APA
Dinh, P. (2026). QSAIS ‐Fuse: A Quality‐Guided Saliency–Structure and Adaptive Interactive Spiking Framework for Multimodal Medical Image Fusion. International Journal of Imaging Systems and Technology, 36(4). https://doi.org/10.1002/ima.70423
Vancouver
Dinh P. QSAIS ‐Fuse: A Quality‐Guided Saliency–Structure and Adaptive Interactive Spiking Framework for Multimodal Medical Image Fusion. International Journal of Imaging Systems and Technology. 2026;36(4). doi:10.1002/ima.70423
BibTeX
@article{Dinh2026,
author = {Dinh, Phu‐Hung},
title = {QSAIS ‐Fuse: A Quality‐Guided Saliency–Structure and Adaptive Interactive Spiking Framework for Multimodal Medical Image Fusion},
journal = {International Journal of Imaging Systems and Technology},
year = {2026},
volume = {36},
number = {4},
doi = {10.1002/ima.70423}
}
EUVIMED