Skip to content

Cell Analysis Guide

This comprehensive guide covers all aspects of automated cell analysis in mAIcrobe, from basic morphometry to cell classification.

🎯 Overview

The "Compute cells" widget provides:

  • πŸ“ Morphological analysis - Shape and size measurements
  • πŸ’‘ Intensity analysis - Fluorescence quantification
  • 🧠 Cell classification - Deep learning single-cell classification. Custom models or pre-trained.
  • πŸ”— Colocalization analysis - PCC analysis across channels
  • πŸ“Š Report generation - HTML and CSV output

πŸ”¬ Analysis Workflow

Step 1: Prepare Your Data

Before running cell analysis, ensure you have:

  1. 🏷️ Segmented cells - Labels layer either loaded externally or computed from the compute_label widget. (See segmentation guide)
  2. πŸ–ΌοΈ Image channels - Fluorescence channels to be analyzed (as needed)

Step 2: Configure Analysis Parameters

βš™οΈ Settings

Image Selection:

  • Label Image: Segmentation results (required)
  • Fluorescence 1: Fluorescence channel (e.g., Membrane stain)
  • Fluorescence 2: Optional, in case it's not needed set to the same channel as Fluorescence 1 (e.g., Nuclear/nucleoid staining)
  • Pixel size: Physical pixel size (e.g., 0.065 ΞΌm/pixel) (optional)

Subcellular Segmentation:

  • Inner mask thickness: Membrane thickness for cytoplasmic measurements (default: 4)
  • Find septum: Whether to detect division septa
  • Septum algorithm: "Isodata" or "Box" thresholding
  • Find open septum: Detect incomplete septa

Fluorescence Analysis:

  • Baseline margin: Background region size (default: 30)

Cell Cycle Classification:

Check cell classification guide for details.

  • Classify cell cycle: Enable single cell classification
  • Model: Choose appropriate pre-trained model or choose a custom model
    • Pre-trained models:
      • S.aureus DNA+Membrane Epi
      • S.aureus DNA+Membrane SIM
      • S.aureus DNA Epi
      • S.aureus DNA SIM
      • S.aureus Membrane Epi
      • S.aureus Membrane SIM
      • E.coli DNA+Membrane AB phenotyping
    • Custom: Use your own trained model
  • Custom model path: If you selected custom, provide the path to your own model file (.keras)
  • Custom model input: Specify input type for custom model (Membrane, DNA, or Membrane+DNA)
  • Custom model max size: Maximum cell size for custom model (default: 50 pixels)

Other Options:

  • Compute Colocalization: Multi-channel colocalization analysis via PCC's. Only works if two different fluorescence channels are provided.
  • Generate Report: Whether to create a report folder with an HTML report and CSV.
  • Report path: Directory to save the report folder.
  • Compute Heatmap: Spatial analysis visualization of a fluorescence channel. Corresponds to an average intensity heatmap over all cells aligned according to their major axis. Outputs a new Image layer.

Outputs

Morphological Measurements

mAIcrobe computes shape and size parameters using scikit-image regionprops. Furthermore it adds custom measurements relevant for bacterial cells such as subcellular segmentation (membrane, cytoplasm, and septum when requested).

Basic Shape Parameters

Area and Size:

  • Area: Cell area (in pixels or ΞΌmΒ² depending on the pixel size setting)
  • Perimeter: Cell boundary length

Shape Descriptors:

  • Eccentricity: Ellipse eccentricity (0=circle, 1=line)
    • ecc = sqrt(1 - (bΒ²/aΒ²)) where a is the semi-major axis and b is the semi-minor axis

Intensity Analysis

Quantify fluorescence signals in subcellular compartments.

Basic Statistics:

  • Baseline intensity: Local background signal in Fluorescence 1 channel.

IMPORTANT: This value is subtracted from all other intensity measurements

  • Cell Median intensity: Median fluorescence within the entire cell in Fluorescence 1 channel
  • Membrane Median intensity: Median fluorescence in the membrane region in Fluorescence 1 channel
  • Cytoplasm Median intensity: Median fluorescence in the cytoplasmic region in Fluorescence 1 channel
  • Septum Median intensity: Median fluorescence in the septum region (if detected and enabled otherwise 0) in Fluorescence 1 channel
  • Fluorescence Ratios 100%, 75%, 25%, 10% percentiles: Ratios between septum and membrane (if septum detected and enabled otherwise 0) in Fluorescence 1 channel
  • DNA Ratio: Relative DNA content compared to baseline background fluorescence (if Fluorescence 1 channel provided, otherwise 0)

🧠 Cell Classification

Use deep learning models to automatically classify cells. Classification is given in the form of a integer per cell corresponding to the predicted class.

mAIcrobe includes pretrained models or you can use your own custom trained models. Check the Cell Classification Guide for details on the available models. To train your own model check the training notebook and Generate Training Data tutorial for details.

οΏ½ Colocalization Analysis

Quantify spatial relationships between two fluorescence channels.

  • Pearson correlation coefficient

πŸ” Interactive Filtering

Use the "Filter cells" widget for real-time quality control:

  1. 🏷️ Select Labels layer after compute_cells
  2. βž• Add filters for any measured feature
  3. πŸ‘οΈ Preview filtered population in real-time
  4. βœ… Use the filtered results for further analysis The new layer "Filtered cells" contains only the selected cells.

Tips

  • Hide the original Labels layer to visualize only the filtered results.
  • Combine multiple filters to refine your selection. For example, filter by area to remove badly segmented cells and by cell classification to focus on a specific cell cycle phase.
  • Compute septum option works best when combined with filtering by classification by cell cycle phase to focus on cells with actual division septa.

πŸ“š Further Reading

πŸ”— Technical References

  • scikit-image regionprops: Documentation
  • napari-skimage-regionprops plugin: GitHub - mAIcrobe internally uses this plugin to add regionprops tables to the GUI.

Next: Explore deep learning cell classification in the Cell Classification Guide πŸ§