Skip to content

Cell Classification Guide

This guide provides comprehensive information about napari-mAIcrobe's cell classification system, including model selection, usage, and custom model integration.

🧠 Overview

napari-mAIcrobe uses deep learning models to automatically classify cell cycle phases based on fluorescence features. The plugin includes 6 pre-trained models optimized for S. aureus under different imaging conditions, and supports user-trained custom models.

οΏ½ Pre-trained Models

DNA + Membrane - S.aureus DNA+Membrane Epi (Epifluorescence) - S.aureus DNA+Membrane SIM (SIM)

DNA Only - S.aureus DNA Epi - S.aureus DNA SIM

Membrane Only - S.aureus Membrane Epi - S.aureus Membrane SIM

Values in the original confusion matrices indicate high diagonal accuracy; membrane-only often performs comparably to dual-channel models, DNA-only slightly lower.

🧭 Model Selection (Decision Tree)

Do you have both DNA and membrane staining? - Yes β†’ DNA+Membrane - Super-resolution? β†’ S.aureus DNA+Membrane SIM - Standard resolution? β†’ S.aureus DNA+Membrane Epi - No β†’ Single-channel - DNA only β†’ DNA SIM (SIM) or DNA Epi (Epi) - Membrane only β†’ Membrane SIM (SIM) or Membrane Epi (Epi)

πŸ› οΈ Custom Model Integration

Model requirements - Format: Keras .keras - Input: Membrane, DNA, or Membrane+DNA - Max size: preprocessing crop size (default 50 px)

Build your dataset with Compute pickles (Plugins β†’ mAIcrobe β†’ Compute pickles). See Tutorials β†’ Generate Training Data. Train using the provided Jupyter notebook.

πŸ§ͺ Using in Compute cells

  1. Enable "Classify cell cycle"
  2. Select pre-trained model or choose Custom and select the .keras file
  3. Results appear in the Labels properties (e.g., Cell Cycle Phase)

πŸ“š Further Reading

  • Tutorials β†’ Generate Training Data
  • API Reference