Image Shuffling¶

Back to MIB | User interface | Ribbon | Home

Overview¶

The Image Shuffling tool in Microscopy Image Browser (MIB) shuffles files for blind modeling or classification, ensuring users don't know which file corresponds to which condition. Models from shuffled files can later be restored to their original filenames for analysis.

Demonstration

Image Shuffling Demo

Shuffle Images¶

This mode renames and shuffles files from input directories into multiple output subdirectories.

Prerequisites

Files for each condition must be in separate folders.
Images in each folder must have the same width and height.
Recommended: Each file contains a single image (no stacking).
For models, masks, or annotations: Only one file per type per directory (e.g., one *.model, *.mask, or *.ann file).
For models: All model files must contain the same materials.

Setup¶

Copy images from different conditions into separate directories (e.g., Control for control images, another folder for treatment images). There may be multiple conditions processed at the same time. The only requirement is to have separate conditions in their own directory.

Settings¶

Add folders: populate the list of input directories with images from different conditions.
Remove folders: remove a selected directory from the list.
Include model: shuffle existing models (one *.model file per folder).
Include mask: shuffle existing masks (one *.mask file per folder).
Include annotations: shuffle existing annotations (one *.ann file per folder).
Include measurements: shuffle existing measurements (one *.measure file per folder).
Random seed: a positive number setting the random number generator seed; same value ensures consistent shuffling.
Filename extension: specify the extension for image filenames (e.g., TIF).
Filename template: define the naming pattern for shuffled images.
Output directory: set the destination directory for shuffled files. Use ... to browse.
Number of output sub-directories: number of subdirectories (e.g., Subset_001, Subset_002) to create under the output directory.
generate Excel file with shuffling details: in addition to the automatically generated .mibShuffle MATLAB-compatible file with shuffling details, generate an Excel sheet where the shuffling order is stored in human-readable format.
Rename and Shuffle: start the shuffling process. An automatically generated .mibShuffle MATLAB-compatible file with sorting details is saved to the output directory.

Restore Order¶

This mode restores shuffled models and masks to match their original image filenames.

Prerequisites

Images in each folder must have the same width and height.
For models, masks, or annotations: only one file per type per directory.

Steps¶

Press ... to select the project file (Project file (*.mibShuffle)), saved in the original shuffle output directory.
The Directories with shuffled images and Destination directories lists populate automatically.

Right-click on any directory entry for a context menu with options to:
- Update directory: manually set a new path for the selected entry
- Update parent folder for selected directories: remap all selected entries under a new parent folder
- Copy path to clipboard: copy the directory path
- Open directory in file explorer: browse the directory in Windows Explorer

Optionally, save updated directory settings with Save as.
Select which layer types to restore: Include model, Include mask, Include annotations, Include measurements.
Press Restore shuffled to start.

Output¶

Restored models: Labels_RestoreRand_YYMMDD.model (e.g., Labels_RestoreRand_250405.model for April 5, 2025).
Restored masks: Mask_RestoreRand_YYMMDD.mask.
YYMMDD reflects the restoration date.

Usage Tips¶

Shuffling for Blind Analysis

To anonymize files:

Organize conditions into separate folders (e.g., Control, Treatment).
Set Number of output sub-directories (e.g., 3 for Subset_001 to Subset_003).
Shuffle with a consistent Random seed.

Back to MIB | User interface | Ribbon | Home