Skip to content

usage-guide.md

Latest commit

 

History

History
52 lines (39 loc) · 1.7 KB

usage-guide.md

File metadata and controls

52 lines (39 loc) · 1.7 KB

Data Preprocessing - Usage Guide

Overview

IMC/MIBI data requires preprocessing before analysis including hot pixel removal, normalization, and format conversion.

Prerequisites

pip install readimc tifffile napari scikit-image
# For steinbock pipeline (Docker-based)
docker pull ghcr.io/bodenmillergroup/steinbock:latest

Quick Start

Tell your AI agent what you want to do:

  • "Load MCD files and extract multichannel TIFFs"
  • "Remove hot pixels from my IMC images"
  • "Normalize channel intensities across samples"

Example Prompts

Data Loading

"Read my MCD file and export each acquisition as OME-TIFF"

"Convert my Hyperion MCD file to multichannel TIFF images"

Hot Pixel Removal

"Detect and remove hot pixels from my IMC images using median filtering"

"Clean detector artifacts from my MIBI data"

Normalization

"Apply percentile normalization to my IMC channels"

"Arcsinh transform my intensity values for visualization"

Batch Processing

"Preprocess all MCD files in my experiment folder"

What the Agent Will Do

  1. Load raw data from MCD/TIFF files using readimc or tifffile
  2. Detect hot pixels by comparing to local median filter
  3. Replace hot pixels with local median values
  4. Apply normalization (percentile, z-score, or min-max)
  5. Optionally apply arcsinh transformation for visualization
  6. Save preprocessed images as OME-TIFF

Tips

  • MCD files contain all acquisitions; extract each ROI separately
  • Hot pixels appear as bright single-pixel spots from detector noise
  • Percentile normalization (1st-99th) is robust to outliers
  • Arcsinh transform (cofactor 5) is standard for visualization
  • steinbock provides a complete Docker-based preprocessing pipeline