Hemocyte Recruitment Analysis

Workflow to analyze hemocyte localization in confocal microscopy images of cryosectioned Drosophila melanogaster abdomens, as described in publication (Chasse et al., Frontiers in Immunology, 2024)

Aims

Are hemocytes recruited to clear the dying ovarian follicle epithelium during starvation?
Does hemocyte localization patterns change in starved abdomens compared to well-fed controls?

Software

• FIJI ImageJ v2.1.0/1.53c
• QuPath v0.3.2
• Python v3.11.5

Workflow overview

QuPath

• Define ROI manually
• Find centroid of defined polygon
• Identify fluorescence as discrete objects
• Find centroid of each object
• Obtain Delauney triangulation metrics

Custom Python notebooks and FIJI macro

• Preprocess DAPI and FITC images to remove background noise and merge channels
• Summarize and consolidate abdomen ROI measurements and hemocyte coordinates into dataframes for downstream analysis
• visualize distributions of abdomen area, hemocyte detections
• Determine hemocyte sparsity in well fed and starved abdomens

Methods

1. Preprocess images

raw images are thresholded based on Yen's method implemented in scikit-image https://scikit-image.org/docs/stable/api/skimage.filters.html#skimage.filters.threshold_yen and then merged into green and blue channels in FIJI

# Create new python virtual environment and install necessary libraries
python3 -m venv hemocyte_quant
source hemocyte_quant/bin/activate

pip3 install ipykernel 
pip3 install numpy 
pip3 install pandas 
pip3 install matplotlib 
pip3 install seaborn
pip3 install scipy 
pip3 install statsmodels
pip3 install sklearn 
pip3 install scikit-image

# Create new kernel with libraries installed
python -m ipykernel install --user --name=hemocyte_quant --user
jupyter notebook 

# Run the following notebook to perform preprocessing 
preprocess_FITC.ipynb 

Once images are preprocessed, run FIJI macro merge_scikit.ijm to apply LUT and merge DAPI and FITC channels

2. ROI definition, Segmentation, centroid measurements, delauney triangulation

• Set up input and output directories
  1. create input directory with merged DAPI and FITC .tif images to be quantified
  2. create output directory named results with 3 following sub-directories -
     1. annotation_measurements for .txt files containing ROI area measurements
     2. detection_measurements- for .txt files containing hemocyte object detection measurements
     3. annotated_images- for storing QuPath processed images containing polygon annotation, positive cell detections, and delauney triangulation measurements.
• Open merged input image to be quantified in QuPath
• Set image type as Fluoresence in the auto-prompt dialogue box
• Use the brush tool to draw ROI boundaries around the abdomen, excluding debris and thorax
• Open the script editor - automate > show script editor and open the file quantify_hemocyteFITC_intensity.groovy
• Modify the paths to the variables annotation_measurements, detection_measurements, annotated_images to include the full paths to the directories created in step 2.b.
• Run the script using keys Cmd+R on Mac OSX or Ctrl+R on Windows

3. Stastiscal analysis and visualization of hemocyte distribution

To perform downstream analysis on hemocyte detections, run the following notebooks in the order mentioned-

scale_coordinates_midsections.ipynb For consolidating feature vectors into dataframes

exploratory_analysis_midsections.ipynb For visualization of distributions of area, hemocyte detections, delaunay triangulation

nearest_neighbors_midsections.ipynb For Nearest-neighbors G-function analysis

Results

Figures and consolidated csv files are in the results/v2_allbatches folder

References

1. Bankhead, P., Loughrey, M.B., Fernández, J.A. et al. QuPath: Open source software for digital pathology image analysis. Sci Rep 7, 16878 (2017). https://doi.org/10.1038/s41598-017-17204-5
2. Spatstat R library
   1. Spatial Point Patterns: Methodology and Applications with R