A fast and robust algorithm for General Defocusing Particle Tracking

TL;DR

This paper introduces a fast, robust, and segmentation-free 3D particle tracking algorithm based on defocused images, suitable for real-time microfluidic applications and capable of handling overlapping particles and various imaging conditions.

Contribution

The work presents a novel, fast, and segmentation-free GDPT algorithm that improves accuracy and robustness for real-time 3D particle tracking in microfluidics.

Findings

Algorithm achieves high detection rate and low depth uncertainty.

Robust against background and illumination fluctuations.

Suitable for real-time applications with synthetic and experimental data.

Abstract

The increasing use of microfluidics in industrial, biomedical, and clinical applications requires a more and more precise control of the microfluidic flows and suspended particles or cells. This leads to higher demands in three-dimensional and automated particle tracking methods, e.g. for use in feedback-control systems. General defocusing particle tracking (GDPT) is a 3D particle tracking method based on defocused particle images which is easy to use and requires standard laboratory equipment. In this work, we describe in detail a fast and robust algorithm for performing GDPT, which is suitable for automatized and real-time applications. Its key feature is a fast, segmentation-free approach to identify particles and estimate their 3D position. This detection step is followed by a refinement and iteration step to improve accuracy and identification of overlapping particles. We show that the algorithm is versatile and can be applied to different types of images (darkfield and brightfield). We use synthetic image sets of varying particle concentration to evaluate the performance of the algorithm in terms of detected depth coordinate uncertainty, particle detection rate, and processing time. The algorithm is applied and validated on experimental images showing that it is robust towards background or illumination fluctuations. Finally, to test the algorithm on real-time applications, we use synthetic images to set up a simulation framework with experimentally-relevant parameters and where the true particle positions are known.