Quantifying Quality: Introducing Bead Analyzer for 3D PSF Characterization

In high-end microscopy — confocal, light-sheet, or custom-built optical engines — your data is only as good as your Point Spread Function (PSF). If the PSF is aberrated, tilted, or wider than the diffraction limit, your resolution suffers, and downstream analysis like deconvolution becomes a game of "garbage in, garbage out."

Despite its importance, characterizing a microscope by measuring fluorescent beads is often a chore. Researchers frequently find themselves manually clicking on beads in ImageJ or struggling with scripts that aren't robust to noisy backgrounds.

To solve this, I've released Bead Analyzer — an open-source tool designed to make 3D PSF characterization fast, accurate, and reproducible.

When we characterize a system, we aren't just looking for a single number. We need to understand how the system performs across the entire field of view. A single bead in the center might look perfect, but what happens at the corners? Is there field curvature? Is the light-sheet properly aligned?

Bead Analyzer automates the extraction of these metrics:

• FWHM (Full Width at Half Maximum): Calculated via both prominence and Gaussian fitting for sub-pixel precision
• 3D Characterization: Measurements for X, Y, and Z axes to detect axial stretching or misalignment
• Spatial Heatmaps: Automatically generates 3×3 grids showing how FWHM varies across the field — essential for identifying alignment drift
• Quality Assurance: Built-in SNR and symmetry filtering ensures that "bad" beads (clumps or out-of-focus spots) don't corrupt your statistics

One of the core challenges in bead analysis is detection. Depending on your system, beads might be tiny diffraction-limited spots or larger spheres. Bead Analyzer supports five different detection modes:

• Blob & Trackpy: Classical methods robust for most standard slides
• StarDist & Cellpose: Deep-learning backends for dense fields or complex backgrounds where classical math struggles
• Manual Mode: For those "one-off" validations where you need absolute control

We didn't just want a CSV of numbers. Bead Analyzer generates publication-quality summary figures, including:

• The "Average Bead": An upsampled, denoised 3D stack of all accepted beads
• Profile Plots: X, Y, and Z intensity profiles with Gaussian overlays
• Diagnostic Plots: A per-bead breakdown so you can verify exactly why a specific bead was accepted or rejected

Reproducibility is at the heart of what we do at Arnas Technologies. By making Bead Analyzer open-source, we hope to provide a standardized way for labs to report system performance.

The tool is available now on GitHub and can be used via a Graphical User Interface (GUI) or a Command Line Interface (CLI) for automated workflows.

Get started here: GitHub — natearnas/bead-analyzer