获取途径

作者

Famin Wang Yongyi Tan Jingyi Gu Jiao Li Huadong Zheng Xinxing Xia Yunhai Zhang Gang Wen Yun Xiao

作者单位

⁶Academy of Biomedical Engineering, Kunming Medical University, Kunming, 650500, China ⁴Clinical Imaging Research Centre, Centre for Translational Medicine, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, 117599, Singapore ⁵Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Science, Suzhou, 215163, China ³State Key Laboratory of Precision Measurement Technology and Instruments, Tianjin University, Tianjin, 300072, China ¹School of Mechanical Engineering and Automation, Shanghai University, Shanghai, 200444, China ²Center for Nuclear Medicine and Molecular Imaging, Shandong Cancer Hospital and Institute, Shandong First Medical University and Shandong Academy of Medical Sciences, Jinan, 250117, China

刊名

Engineering Applications of Artificial Intelligence

年份

2026

卷号

Vol.175

页码

114620

ISSN

0952-1976

摘要

Single-molecule three-dimensional real-time tracking is crucial for addressing key scientific questions, including viral infection mechanisms, interactions between nanodrugs and target cells, and the laws governing interactions among nucleic acid macromolecules. However, reconstructing 3D multi-molecule motion trajectories remains a major challenge in this technology, which restricts its applications in molecular biology and cell biology. A novel artificial intelligence-based image processing t...更多

Single-molecule three-dimensional real-time tracking is crucial for addressing key scientific questions, including viral infection mechanisms, interactions between nanodrugs and target cells, and the laws governing interactions among nucleic acid macromolecules. However, reconstructing 3D multi-molecule motion trajectories remains a major challenge in this technology, which restricts its applications in molecular biology and cell biology. A novel artificial intelligence-based image processing technique has been developed to address the prevalent issues of suboptimal detection accuracy, tracking drift, and the labour-intensive processes associated with conventional nanoparticle tracking methodologies. First, a generative adversarial network is employed to enhance low signal-to-noise ratio images by suppressing background noise while preserving fine structural details. Then, You Only Look Once version 8 , a real-time object detection algorithm, is applied to accurately locate nanoparticles in each frame. Finally, Better Tracking by Association associates detected particles across consecutive frames to reconstruct continuous motion trajectories. This deep neural network-based multimolecular 3D motion trajectory reconstruction algorithm features high-fidelity image denoising, accurate target detection, and precise particle tracking and matching. In this study, a 3D particle imaging technique based on double-helix phase modulation is used to observe nanoparticles suspended in a glycerol solution, effectively capturing foundational images of particle dynamics. Subsequently, the proposed GAN–YOLOv8–Track algorithm is employed to reconstruct the 3D trajectories of these nanoparticles. Experimental results show that our model outperforms traditional methods in tracking stability, detection accuracy, and operational simplicity.收起

文献类型

期刊

浏览量

1