bbCARMEN Automated Image Analysis Pipeline

This repository contains code and workflows for automated image analysis and downstream data processing of bbCARMEN assays. The pipeline leverages CellProfiler 3.5 and a custom Jupyter notebook to identify, track, and classify beads in multiplexed viral detection experiments.

Overview

The bbCARMEN analysis pipeline performs:

• Bead detection and filtering using CellProfiler
• Color channel correction and droplet masking
• Timepoint-based bead tracking
• Bead clustering and virus assignment via k-means
• FAM fluorescence quantification and classification
• Visualization and QC reporting

Workflow Summary

1. Bead Detection & Filtering (CellProfiler)

• Illumination correction by background subtraction (per channel)
• Color bleedthrough correction via image subtraction
• Droplet masking to exclude well edge artifacts
• Bead filtering:
  • By shape: solidity, eccentricity
  • By proximity: exclusion of beads near neighbors
  • By droplet content: exclusion of multi-bead droplets

2. Bead Feature Extraction

• Each accepted bead’s mask is expanded by 5 pixels to define a donut region
• FAM (blue channel) intensity is measured in this donut region

3. Bead Tracking Across Time

• Uses Linear Assignment Problem (LAP) framework to track beads across timepoints

4. Bead Classification (Jupyter Notebook)

• Normalized RGB features are computed per bead:
  • Normalized = channel intensity / sum of RGB intensities
• K-means clustering on normalized red, green, and yellow values
• Ternary plot visualization with cluster coloring
• Each cluster matched to a known virus based on proximity to predefined centroids

5. FAM Fluorescence Quantification

• Median blue intensity in the donut region used to classify virus presence per sample
• Classification thresholding based on:
  • Fold-difference from within-well negative controls
  • or number of standard deviations above the negative control median
• FAM fluorescence kinetics visualized over time

Outputs

• Ternary cluster plots for bead classification
• Virus detection calls per sample
• Kinetic plots of FAM fluorescence
• QC plots and metrics

Code Repository

👉 https://github.com/broadinstitute/bbCARMEN_analysis

Authors

Tien G. Nguyen
Rebecca Senft
David R. Stirling
Sameed M. Siddiqui
Nicole L. Welch
Cheri M. Ackerman
Paul C. Blainey
Pardis C. Sabeti
Cameron Myhrvold

License

This project is licensed under the MIT License. See the LICENSE file for details.