zFISHer: Automated 3D Registration, Detection, and Colocalization with Interactive Curation for Sequential Multiplexed FISH

zFISHer is an open-source, napari-based application that automates the 3D registration, detection, and colocalization analysis of sequential multiplexed FISH data while providing interactive curation tools to overcome the labor-intensive bottlenecks of manual analysis.

The Gist

Imagine you are trying to solve a massive, 3D jigsaw puzzle, but the pieces are tiny glowing dots inside cells, and the picture keeps shifting every time you look at it. This is the daily reality for scientists using a technique called "sequential multiplexed FISH." They want to see where specific molecules are located inside cells by taking pictures in layers (like slicing a loaf of bread). However, because the cells move slightly between each photo, lining up these layers to see which dots actually belong together is incredibly tedious and time-consuming work.

Enter zFISHer, a new free software tool designed to be the ultimate assistant for this job. Think of it as a smart, interactive workshop built on top of a popular viewing platform called "napari."

Here is how zFISHer tackles the problem, using some everyday comparisons:

• The Automatic Aligner: Imagine taking a stack of slightly wobbly photos and needing to stack them perfectly on top of each other. zFISHer does this automatically. It uses a clever math trick (called RANSAC) to find the best way to slide the layers into place, and if the layers are a bit warped or bent, it can gently stretch and reshape them (like smoothing out a crumpled map) to make them fit perfectly.
• The 3D Detective: Once the layers are aligned, the software needs to find the tiny glowing dots (puncta). It doesn't just look at one flat slice; it builds a complete 3D map of the cell's nucleus.
• The "Fishing Hook" Tool: This is one of the coolest features. Usually, finding the exact center of a dot in a 3D space requires scrolling up and down through hundreds of image slices, like searching for a needle in a haystack. zFISHer introduces a "Fishing Hook" raycasting algorithm. Imagine casting a fishing line straight down from your camera view; the hook automatically snags the brightest spot of light along that line. This instantly tells you the exact 3D center of the dot without you having to scroll manually. It even fine-tunes the location to be more precise than a single pixel.
• The Interactive Editor: Sometimes, computers make mistakes. zFISHer lets you step in and fix them. It's like a video game where you can drag and drop dots to the right spot. As you move them, the software instantly checks if you accidentally placed two dots on top of each other (a "collision"), acting like a digital traffic cop to keep everything organized.
• The Party Planner: Finally, the software analyzes which dots are hanging out together. It counts how often different colored dots appear in the same spot (colocalization), giving scientists a clear statistical report on which molecules are friends and which are strangers.

For scientists who have hundreds of these experiments to run, zFISHer also has a "batch mode." Think of this as setting a coffee maker to brew while you sleep; you can load up multiple datasets, hit start, and let the software run unattended while it processes everything automatically.

In short, zFISHer takes a process that used to be a slow, manual struggle of aligning and hunting for dots, and turns it into a fast, automated, and interactive experience, all while keeping the data in perfect 3D focus.

Technical Summary

Technical Summary of zFISHer

Problem Statement
Sequential multiplexed fluorescence in situ hybridization (FISH) has emerged as a powerful technique for spatially resolved molecular profiling within cell monolayers. However, the analysis of these datasets faces a significant bottleneck: the labor-intensive process required to analyze puncta colocalization across three-dimensional (3D) volumes. Specifically, aligning multiple imaging rounds, detecting puncta in unaligned z-stacks, and manually curating results to ensure accuracy are time-consuming tasks that hinder high-throughput analysis.

Methodology
To address these challenges, the authors developed zFISHer, an open-source application built on the napari viewer platform. The software is designed to provide a complete, end-to-end workflow for processing paired sequential FISH datasets. The methodology integrates automated computational pipelines with interactive user-curation tools, structured as follows:

• Preprocessing and Segmentation: The pipeline begins with nuclear segmentation and automated puncta detection performed on unaligned z-stacks.
• Image Registration: Multi-round image registration is achieved using a translation-constrained RANSAC (Random Sample Consensus) algorithm. This is complemented by optional B-spline deformable warping to correct for non-rigid distortions between imaging rounds.
• Coordinate Transformation: Once the images are aligned, the system precisely transforms detected puncta coordinates into the aligned reference space.
• Consensus Generation: The software generates consensus nuclei to facilitate consistent analysis across rounds.
• Interactive Curation and the "Fishing Hook" Algorithm: A core innovation is the "Fishing Hook" raycasting algorithm. This tool allows users to locate puncta at their true 3D centroids by identifying intensity maxima along the camera ray. This mechanism eliminates the need for manual z-slice navigation. The interface supports real-time collision detection and sub-voxel volume optimization, enabling precise interactive editing of detection results.
• Colocalization and Statistics: The system performs pairwise and tri-channel colocalization analysis, generating relevant statistics.
• Batch Processing: To support scalability, zFISHer includes a batch processing mode for the unattended, high-throughput analysis of multiple experimental datasets.

Key Contributions
The primary contribution of this work is the integration of full automation with interactive curation in a single open-source framework. Key technical contributions include:

1. End-to-End Automation: A unified pipeline covering segmentation, registration, coordinate transformation, and colocalization analysis.
2. The Fishing Hook Algorithm: A novel raycasting approach that resolves 3D centroid localization without manual slice-by-slice inspection, significantly reducing user effort.
3. Hybrid Workflow: The combination of robust automated registration (RANSAC with B-spline warping) and real-time interactive tools (collision detection, sub-voxel optimization) within the napari ecosystem.
4. Scalability: The inclusion of batch processing capabilities to handle large-scale experimental datasets without continuous user intervention.

Results and Significance
The paper presents zFISHer as a solution to the specific bottleneck of 3D puncta colocalization analysis in sequential multiplexed FISH. By automating the registration and detection phases while retaining interactive tools for quality control, the software aims to reduce the manual labor currently required for these analyses. The significance of the work lies in its ability to facilitate high-throughput, accurate spatial profiling by making the complex 3D analysis of sequential FISH data more accessible and efficient. The authors position zFISHer as a practical tool that bridges the gap between raw imaging data and reliable molecular profiling statistics, specifically targeting the needs of researchers working with cell monolayers.