Real-Time 3D Tracking of Multi-Particle in the Wide-Field Illumination Based on Deep Learning

Area of Science:

• Optics and Photonics
• Biophysics
• Materials Science

Background:

• Accurate particle localization and motion analysis are crucial in fields like optical tweezers, cell tracking, and drug delivery.
• While deep learning shows promise for particle tracking, a lack of suitable datasets has been a significant barrier.
• Existing algorithms face challenges in precise 3D particle orientation and real-time analysis.

Purpose of the Study:

• To address the need for robust datasets in deep learning-based particle tracking.
• To develop and validate a high-precision 3D real-time particle positioning network.
• To make generated synthetic datasets publicly available for research advancement.

Main Methods:

• Generation of a synthetic dataset tailored for 3D particle tracking applications.
• Development of a 3D real-time particle positioning network utilizing the CenterNet architecture.
• Experimental validation of the network's performance on real-world particle tracking data.

Main Results:

• The developed network achieved a horizontal positioning error of 0.0478 μm.
• A z-axis positioning error of 0.1990 μm was recorded, demonstrating depth accuracy.
• The system proved capable of high-precision, real-time tracking of diverse particles near the focal plane.

Conclusions:

• The proposed methodology effectively generates synthetic datasets that enhance deep learning model performance in 3D particle tracking.
• The 3D real-time particle positioning network offers a precise and efficient solution for analyzing particle dynamics.
• The public release of datasets facilitates further research and development in particle tracking technologies.