LiveMIEL+: Processing Microscopy Images of Epigenetic Landscape

Distinguish between epigenetic states in living cells directly from raw fluorescence microscopy data using a complete image processing pipeline.

Installation

LiveMIEL+ is configured for immediate use in Google Colab with Google Drive integration.

Setup Steps

1. Upload to Google Drive:

   • Upload the entire LiveMIEL+ package to your Google Drive
   • Ensure python_files/ and Segmentation_features_extraction_clustering.ipynb are in the project root

2. Prepare Data Directory:

   • Inside the LiveMIEL+ directory, create a folder nuclei_images
   • Add your microscopy images or download test data here

3. Run Analysis:

   • Open Segmentation_features_extraction_clustering.ipynb in Google Colab
   • Mount your Google Drive when prompted
   • Install dependencies (automated in notebook)
   • Execute the notebook cells

Technical Details

Supported Image Formats

• Grayscale 8-bit (single channel intensity)
• One-channel RGB (only one channel contains data)
• Uniform RGB (all three channels have identical values)

Recommended Format: TIFF (high-resolution, lossless compression - common in microscopy)

Dependencies: All required packages are listed in requirements.txt and automatically handled in the notebook's Libraries import section.

Usage Example

The notebook Segmentation_features_extraction_clustering.ipynb provides a complete walkthrough of LiveMIEL+ applied to a cell-cycle study dataset. You can download the test data here.

Dataset Description: The dataset consists of fluorescent microscopic images of HEK293 cells from a 24-hour time-lapse experiment, with images captured every 15 minutes.

• Microscope Settings: 20x magnification, 1410 x 1200 px resolution.
• Channels: Images were collected in three channels (Red, Yellow, Blue), each stored in a separate directory named after its color.
• Biological Context:
  • Red Channel: Epigenetic landscape of methylation mark H3K9me3.
  • Yellow & Blue Channels: Cell cycle phase indicators.
    • Nuclei in Yellow only: S phase.
    • Nuclei in Blue only: G1 phase.
    • Nuclei in both Yellow and Blue: G2/M phase.

Tunable Parameters

The segmentation pipeline is highly configurable to adapt to different image qualities and magnifications. It consists of three stages:

1. Bandpass Segmentation: (image * big_gaussian_kernel) - coeff x (image * small_gaussian_kernel) > thresh
2. Watershed Segmentation: Separates adjacent or touching nuclei.
3. False Positive Removal: Filters out objects that do not correspond to true nuclei.

Core Parameters

Parameter	Description	Typical Range (60x, 1360x1024 px)
lowSigm	Std. dev. of the small_gaussian_kernel. Smoothes noise and defines object detail. Lower values preserve more detail; higher values smooth outlines.	[5, 15]
highSigm	Std. dev. of the big_gaussian_kernel. Removes background. Lower values remove more background fragments.	[20, 70]
thresh	Intensity threshold for the bandpass filter. Pixels above this value are considered potential objects.	[0.003, 0.05]
FalsePositBrightness_k	Brightness coefficient k. Objects with mean fluorescence < k * average_image_fluorescence are removed.	[1.0, 2.5]
MinNucleusArea	Minimum area (in pixels). Objects smaller than this are discarded.	[500, 1500]

Scaling Parameters for Different Magnifications/Resolutions

The parameters lowSigm, highSigm, and MinNucleusArea scale linearly with the image dimensions.

Example: For a 60x image with a resolution of 4080 x 3072 px (3x larger in each dimension than the base 1360x1024):

• Scale Factor: (4080 / 1360) = 3
• Area Scale Factor: 3 * 3 = 9

Parameter	Scaled Value / Range
lowSigm	[5*3, 15*3] = [15, 45]
highSigm	[20*3, 70*3] = [60, 210]
MinNucleusArea	1000 * 9 = 9000

Tips for Parameter Tuning:

• Start with the scaled base values and adjust incrementally
• Use the example images above as visual references for expected outcomes

Citing

If you use this package in your publications or in other work, please cite it as follows:

@article{stepanov2024iPSC,
  title={Tracking induced pluripotent stem cell differentiation with a fluorescent genetically encoded epigenetic probe},
  author={Stepanov, A. I. and Shuvaeva, A. A. and Putlyaeva, L. V. and Lukyanov, D. K. and Galiakberova, A. A. and Gorbachev, D. A. and Maltsev, D. I. and Pronina, V. and Dylov and D. V. and Terskikh, A. V. and Lukyanov, K. A. and Gurskaya, N. G.},
  journal={Cellular and Molecular Life Sciences},
  number = {1},
  pages = {381},
  volume = {81},
  doi = {/10.1007/s00018-024-05359-0},
  url = {https://link.springer.com/article/10.1007/s00018-024-05359-0}
  year = {2024}
}