StarDist_BF_cancer_cell_dataset_10x

This repository includes a StarDist deep learning model and its training dataset designed for segmenting cancer cells perfused over an endothelial cell monolayer captured at 10x magnification. The model was trained on 77 manually annotated images, with the dataset being computationally augmented during training by a factor of 8. The model was trained for 500 epochs and achieved an average F1 Score of 0.968, indicating high accuracy in segmenting cancer cells on endothelial cells.

Specifications





Model: StarDist for cancer cell segmentation on endothelial cells (10x magnification)




Training Dataset:






Number of Images: 77 paired brightfield microscopy images and label masks




Augmented Dataset: Computational augmentation by a factor of 8 during training




Microscope: Nikon Eclipse Ti2-E, 10x objective




Data Type: Brightfield microscopy images with manually segmented masks




File Format: TIFF (.tif)






Brightfield Images: 16-bit




Masks: 8-bit or 16-bit





Image Size: 1024 x 1022 pixels (pixel size: 1.3148 μm)





Training Parameters:






Epochs: 500




Patch Size: 992 x 992 pixels




Batch Size: 2





Performance:






Average F1 Score: 0.968




Average IoU: 0.882





Model Training: Conducted using ZeroCostDL4Mic (https://github.com/HenriquesLab/ZeroCostDL4Mic/wiki)





Reference

Fast label-free live imaging reveals key roles of flow dynamics and CD44-HA interaction in cancer cell arrest on endothelial monolayers


Gautier Follain, Sujan Ghimire, Joanna W. Pylvänäinen, Monika Vaitkevičiūtė, Diana Wurzinger, Camilo Guzmán, James RW Conway, Michal Dibus, Sanna Oikari, Kirsi Rilla, Marko Salmi, Johanna Ivaska, Guillaume Jacquemet

bioRxiv 2024.09.30.615654; doi: https://doi.org/10.1101/2024.09.30.615654