PO.BCS01.13 · 生物信息与计算
频域信号分离网络改善磁粒子成像中动脉瘤区域的双示踪剂检测
Frequency-domain signal separation network improves dual-tracer detection of aneurysm regions in magnetic particle imaging
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
动脉瘤是以显著结构异质性为特征的血管异常，其破裂风险与区域扩张、局灶性薄弱和炎症活动密切相关。因此，可靠地区分正常血管与受动脉瘤影响的区域对于动脉瘤检测、区域可视化和病灶评估至关重要。与磁共振成像（MRI）和计算机断层扫描（CT）等现有医学成像技术相比，磁粒子成像（MPI）具有独特优势，包括零组织背景、高灵敏度、实时成像和出色的定量能力，使其非常适合多示踪剂血管成像。不同的超顺磁性氧化铁（SPIO）示踪剂在磁性和磁化动力学上存在固有差异，为多示踪剂MPI提供了物理基础；然而，实际测量中的噪声、串扰和信号失衡会掩盖这些差异，阻碍准确的双示踪剂分离。为提高多示踪剂MPI的可靠性，我们提出一种频域双示踪剂分离框架，即频域信号分离网络（FSS-Net）。FSS-Net将原始MPI信号映射为二维谐波频率表征，通过分离模块预测示踪剂特异性掩膜，并使用频域解码器重建无干扰的双通道谐波信号。为验证其性能，我们构建了一个双示踪剂动脉瘤模体，其中正常血管区域和动脉瘤病灶用不同的SPIO示踪剂标记，模拟病变与健康血管系统中异质的纳米颗粒分布。将FSS-Net与两种成熟的多色MPI方法——系统矩阵拼接（SM Cat）和MKZ进行了比较。使用PSNR和SSIM进行的定量评估表明，FSS-Net在信号保真度和结构保持方面显著优于这两种方法，有效减少了串扰并改善了正常血管和受动脉瘤影响区域的可视化。总体而言，FSS-Net实现了高质量的双示踪剂MPI信号分离，为动脉瘤区域检测、区域可视化和纳米颗粒分布分析提供了可靠的方法，展现了在血管病理成像和基于纳米颗粒的生物医学研究中的强大潜力。
查看英文原文 English abstract
Aneurysms are vascular abnormalities characterized by marked structuralheterogeneity, and their rupture risk is closely associated with regional dilation, focalweakening, and inflammatory activity. Therefore, reliably distinguishing normalvessels from aneurysm-affected regions is essential for aneurysm detection, regionalvisualization, and lesion assessment. Compared with existing medical imagingtechniques such as magnetic resonance imaging (MRI) and computed tomography(CT), Magnetic Particle Imaging (MPI) offers unique advantages including zero tissuebackground, high sensitivity, real-time imaging, and excellent quantitative capability,making it well suited for multi-tracer vascular imaging. Different superparamagneticiron oxide (SPIO) tracers exhibit inherent differences in magnetic properties andmagnetization dynamics, providing the physical basis for multi-tracer MPI; however,noise, cross-talk, and signal imbalance in practical measurements can obscure thesedifferences and hinder accurate dual-tracer separation. To enhance the reliability ofmulti-tracer MPI, we propose a frequency-domain dual-tracer separation framework,the Frequency-Domain Signal Separation Network (FSS-Net).FSS-Net maps raw MPI signals into a two-dimensional harmonic-frequencyrepresentation, predicts tracer-specific masks through a separation module, andreconstructs interference-free dual-channel harmonic signals using afrequency-domain decoder. To validate its performance, we constructed a dual-traceraneurysm phantom in which normal vessel regions and aneurysm lesions were labeledwith different SPIO tracers, simulating heterogeneous nanoparticle distribution indiseased versus healthy vasculature. FSS-Net was compared with two establishedmulti-color MPI methods-System Matrix Concatenation (SM Cat) and MKZ.Quantitative evaluation using PSNR and SSIM showed that FSS-Net significantlyoutperformed both methods in signal fidelity and structural preservation, effectivelyreducing cross-talk and improving visualization of normal vessels andaneurysm-affected regions.Overall, FSS-Net enables high-quality dual-tracer MPI signal separation and offers areliable approach for aneurysm region detection, regional visualization, andnanoparticle distribution analysis, demonstrating strong potential for vascularpathology imaging and nanoparticle-based biomedical research.
利益披露 Disclosure
Z. Chen, None..
G. Shi, None..
J. Ye, None..
Z. Zhang, None..
X. Feng, None..
Y. An, None..
J. Tian, None.