Enhancing 3-D Magnetic Particle Imaging Spatial Resolution via Fusion of 2-D NIR-II Fluorescence Molecular Imaging

Abstract

Precise localization and high-resolution imaging of deep-seated microtumors remain key technical challenges in molecular imaging. Magnetic particle imaging (MPI) has strong potential for tumor detection because it provides depth-independent imaging. However, the spatial resolution of MPI is limited under low selection-field (SF) gradients, while high SF gradients may raise biological safety concerns. To address this challenge, we introduce near-infrared II fluorescence molecular imaging (NIR-II FMI) as a complementary imaging modality and integrate it with MPI for multimodal fusion imaging, leveraging its high spatial resolution to compensate for the limited resolution of low-SF-gradient MPI. We construct an NIR-II FMI-MPI fusion imaging system, and based on this system, we propose a multitask fusion framework, Swin-FMnet, which decomposes the fusion process into modality-specific measurement data preprocessing and fine-grained cross-modal fusion. The preprocessing modules suppress measurement artifacts and enhance informative features, while the fine-grained fusion module leverages local cross-attention to enable effective cross-modal interaction, with a sliding-window strategy maintaining global–local feature balance. Through this fusion strategy, the spatial resolution, morphological fidelity, and localization accuracy of MPI are substantially enhanced without increasing the SF gradient. The simulation and phantom experiments demonstrated that Swin-FMnet achieved a dice coefficient of 75.5 %, a localization error (LE) of 0.233 mm, and a full-width at half-maximum (FWHM) of 0.809 mm, corresponding to 17.2 % higher dice and 38.7 % lower FWHM than the second-best fusion results. In vivo imaging on mice with glioma further validated its improved biomedical application potential. Overall, our MPI-NIR-II FMI fusion imaging framework provides an effective solution for enhancing the spatial resolution of MPI and enables high-precision imaging of deep-seated microtumors.