DeepCellMap

Algorithms associated with the submission of the paper titled "Spatiotemporal mapping of human microglia during brain development with advanced spatial statistics assisted by deep-learning" to Nature Methods.

This directory contains all the source code needed to reproduce the results of the paper using different notebooks (doc/src/notebooks). Some intermediate results have been provided (doc/data) to speed up the calculation of spatiotemporal statistics on several pre-defined regions of interest.

Overview

• Data
  • Input
  • Intermediate results provided
• Code
  • Notebooks description
    • DeepCellMap_Pipeline_1 - Image preprocessing and cells segmentation & classification
    • DeepCellMap_Pipeline_2 - Selection different ROIs and computation statistics
    • DeepCellMap_Pipeline_3 - Spatiotemporal analysis
    • DeepCellMap_x_Cellpose_region_segmentation
    • DeepCellMap_x_Manual_region_segmentation
    • MAYBE - DeepCellMap_Training_set_construction
  • Python files description
    • General
    • Region of interest (central)
    • Image processing & Deep learning
    • Statistics
    • Cellpose
    • Spatiotemporal analysis

Data

Input

• 3 IHC images of human fetal brain at 17,19 and 20 pcw.
• Model classification microglial cells

Intermediate results provided

• Segmented & classified cells for the three IHC images.

Resultat de l'application du model de segmentation et de classification des cellules microgliales à l'ensemble des images. Cet output peut être généré dans DeepCellMap_notebook_1.d mais prend du temps (entre 7h/image et 30h/image selon la machine et la taille de l'image). - Masks of the 4 anatomical regions (striatum, neocortex, cordical boundary, ganglionic eminence) for the three times. - Mask of a manually segmented region. Ce Resultat peut être reproduis avec le notebook

Code

Notebooks description

Notes sur les notebooks : Les DeepCellMap notebooks 1/2/3 servent à guider pas à pas dans l'application du pipeline à toutes les images.

DeepCellMap_Pipeline_1 - Image preprocessing and cells segmentation & classification

Performs the different steps

• a.Image downscaling
• b.Mask extraction
• c.Tiling
• d.Cells classification on the entire images

DeepCellMap_Pipeline_2 - Selection different ROIs and computation statistics

Performs the different steps - a.Selection ROIs in tissue - b.Visualisation ROI and cells - c.Cell-cell colocalisation - d.Cell-Border colocalisation - e.DBSCAN-based clusters analysis - f.Neighbors analysis

DeepCellMap_Pipeline_3 - Spatiotemporal analysis

Performs the different steps : - a.Selection of regions to compare over time - b.Generation of statistical figures

DeepCellMap_x_Cellpose_region_segmentation

Performs the different steps : - a.Cellpose best model selection - b.Cellpose application on the entire images - c.Anatomical region segmentation using image processing

DeepCellMap_x_Manual_region_segmentation

Performs the different steps : - a.Manual selection of a ROI

MAYBE - DeepCellMap_Training_set_construction

Performs the different steps : - a.Random selection of microglial cells in an image - b.Annotation of the different cells - c.Visualisation and correction

Python files description

Python files used in the notebooks

General

const.py:	Contains all project constants and parameters
const_roi.py:	ROI definition
util.py:	Displaying functions, paths/image manipulation, measurement calculation time

Region of interest (central)

region_of_interest.py:	Central - Contains the RegionOfInterest class for reconstructing an ROI and performing calculations on it

Image processing & Deep learning

slide.py:	Used in image pre-processing - IHC image manipulation, downscaling
filter.py:	Used in image pre-processing - Tissue extraction, filtering functions
tiles.py:	Used in image pre-processing - Images tiling, generation of summary html
segmentation.py:	Microglial cell segmentation functions
training_set_constitution.py:	Used to create the training database
train_classification_model.py:	and train the Unet Deep-Learning classification model
classification.py:	Use to classify microglial cells on an entire image
training_set_constitution.py:

Statistics

colocalisation_analysis.py:	Algorithms for Cell-cell colocalisation and Cell-region's border colocalisation analysis
dbscan.py:	Algorithms for cluster analysis based on DBSCAN
neighbours_analysis.py:	Algorithms for analysing neighbour-neighbour relationships

Cellpose

util_cellpose.py:	Algorithms for region segmentation based on nuclei density obtained by cellpose

Spatiotemporal analysis

deepcellmap.py:	Defines ROIs and applies the pipeline on them, gathers the results in dataframes for spatiotemporal analysis
display_statistics.py:	Generates statistical figures

---