Project work (group 1) from Defragmentation Workshop from October 2022

Group members:

• Daniel Waiger
• Camille Loiseau
• Tatiana Woller
• Anna Agafonova
• Gayathri Nadar

This project would not have been successful without the friendly collaboration of the team members :)

Analysis Goal

• Segment 'only' the cytoplasm (neurons) in confocal single slice mouse brain images with no nuclei stain.

Workflow

1. save_tiff.py: save the images saved in propietary format into tiff for easy accessibility by other programs. The channels are split and saved separately. Works on a folder of images.
2. cytoplasm_segmentation.ipynb:
   • Read images in folder, randomly pick one for training an Object classifier in napari-devbio package.
   • Initially training could be done by adding labels in napari window.
   • Option to also load labels from image file - possibly saved using napari > labels > save as image.
   • We train a classifier to identify cytoplasm region in cells.
   • The classifier is saved to disk.
3. batch_prediction_using_classifier.ipynb:
   • Read images in folder, read classifier model
   • Apply to images in the folder in a loop and save label maps of the segmentation.

Images

• Images used for training: 2D, single channel containing the cytoplasm