Skip to content

regionprops_analysis.md

Latest commit

 

History

History
96 lines (71 loc) · 3.96 KB

regionprops_analysis.md

File metadata and controls

96 lines (71 loc) · 3.96 KB

Regionprops Analysis Documentation

Extract Regionprops to CSV

This processing function extracts region properties (regionprops) from all label images in a folder and saves them to a single CSV file. It is dimension-agnostic, meaning it can handle images with different numbers of dimensions and treats non-spatial dimensions (like Time or Channel) as grouping variables.

Features

  • Automatic dimension detection: Works with 2D (YX), 3D (ZYX), 4D (TZYX, CZYX), and 5D (TCZYX) label images
  • Comprehensive properties: Extracts area, centroid, bounding box, solidity, extent, and more
  • Grouping by dimensions: For time-series or multi-channel data, adds T, C, or Z columns to track regions across frames
  • Single CSV output: All regions from all files are saved to one convenient CSV file
  • Flexible spatial dimensions: Can process either 2D (YX) or 3D (ZYX) as the spatial unit

Usage in Napari

  1. Open napari and go to Plugins > T-MIDAS > Image Processing
  2. Browse to your folder containing label images (.tif, .npy, etc.)
  3. Select "Extract Regionprops to CSV" from the processing functions dropdown
  4. Configure parameters:
    • max_spatial_dims: Set to 3 for 3D analysis (ZYX) or 2 for 2D analysis (YX)
    • overwrite_existing: Set to True to overwrite existing CSV files
  5. IMPORTANT: Set thread count to 1 (this function processes entire folders at once)
  6. Click "Start Batch Processing"

The function will create a CSV file named <folder_name>_regionprops.csv in the parent directory.

Output CSV Structure

The output CSV contains one row per labeled region with the following columns:

Identifier Columns

  • filename: Name of the source label image file
  • T: Time index (for 4D/5D time-series data)
  • C: Channel index (for multi-channel data)
  • label: The label ID of the region

Spatial Properties

  • area: Number of pixels/voxels in the region
  • centroid_x, centroid_y, centroid_z: Center coordinates
  • bbox_min_x, bbox_max_x, etc.: Bounding box coordinates

Shape Properties (when available)

  • eccentricity: How elongated the region is (2D only)
  • solidity: Ratio of region area to convex hull area
  • extent: Ratio of region area to bounding box area
  • perimeter: Perimeter length (2D only)

Example: Time-Series Analysis

For a folder containing 4D label images (TZYX), the CSV will include a T column:

filename,T,label,area,centroid_z,centroid_y,centroid_x,...
cell_tracking.tif,0,1,5000,10.5,120.3,85.7,...
cell_tracking.tif,0,2,4800,12.1,200.5,150.2,...
cell_tracking.tif,1,1,5200,10.8,122.1,86.5,...
cell_tracking.tif,1,2,4900,12.3,202.3,151.0,...

This makes it easy to track how properties change over time in tools like pandas, R, or Excel.

Programmatic Usage

You can also use the function directly in Python:

from napari_tmidas.processing_functions.regionprops_analysis import (
    analyze_folder_regionprops,
)

# Analyze all label images in a folder
df = analyze_folder_regionprops(
    folder_path="/path/to/label/images",
    output_csv="/path/to/output.csv",
    max_spatial_dims=3,  # 3 for ZYX, 2 for YX
)

# Now analyze the results with pandas
print(f"Total regions: {len(df)}")
print(df.groupby("T")["area"].mean())  # Average area per timepoint

Notes

  • Requires pandas to be installed: pip install pandas
  • The function uses a cache to avoid processing the same folder multiple times in one session
  • Call reset_regionprops_cache() if you need to reprocess a folder
  • Some properties (like eccentricity and perimeter) are only available for 2D regions

Use Cases

  • Cell tracking: Extract cell properties across time to analyze growth, movement, or division
  • Multi-channel analysis: Compare properties of objects in different channels
  • High-throughput analysis: Process entire experiments at once and analyze in your favorite data analysis tool
  • Quality control: Check segmentation quality by examining region properties