Model-free 3D Localization with Precision Estimates for Brightfield-imaged Particles

Volumetric imaging and 3D particle tracking are becoming increasingly common and have a variety of microscopy applications including in situ fluorescent imaging, in-vitro single-molecule characterization, and analysis of colloidal systems. While recent interest has generated discussion of optimal schemes for localizing diffraction-limited fluorescent puncta, there have been relatively few published routines for tracking particles imaged with bright-field illumination. To address this, we outline a simple, look-up-table based 3D tracking strategy, which can be adapted to most commercially available wide-field microscopes, and present two image processing algorithms that together yield high-precision localization and return estimates of statistical accuracy. Under bright-field illumination, a particle's depth can be determined based on the size and shape of its diffractive pattern due to Mie scattering. Contrary to typical "super-resolution" fluorescence tracking routines, which typically fit a diffraction-limited spot to a model point-spread-function, the lateral (XY) tracking routine relies on symmetry to locate a particle without prior knowledge of the form of the particle. At low noise levels (signal:noise > 1000), the symmetry routine estimates particle positions with accuracy better than 0.01 pixel. Depth localization is accomplished by matching images of particles to those in a pre-recorded look-up-table. The routine presented here optimally interpolates between LUT entries with better than 0.05 step accuracy. Both routines are tolerant of high levels of image noise, yielding sub-pixel/step accuracy with signal-to-noise ratios as small as 1, and, by design, return confidence intervals indicating the expected accuracy of each calculated position. The included implementations operate extremely quickly and are amenable to real-time analysis at frame rates exceeding several hundred frames per second.