Segmentation models

Deep learning segmentation models

---

Here, we list the main models used in the paper and their respective training datasets. The models were trained using ZeroCostDL4Mic StarDist 2D Notebooks.

Model Name	Imaging Modality	Performance	Purpose and Associated Figure	Training Dataset Link
Flow chamber dataset	Brightfield	IoU = 0.813 f1 = 0.933	StarDist model to detect cancer cells in BSA-coated channels. Used to measure perfusion speed inside the channels (Fig S1).	Link
StarDist_Fluorescent_cells	Fluorescence	IoU = 0.646 f1 = 0.877	StarDist model to detect cancer cells from fixed samples. Used in Fig. 1 to count the number of attached cells	Link
StarDist_BF_cancer_cell_dataset_20x	Brightfield	IoU = 0.793 f1 = 0.921	StarDist model capable of segmenting cancer cells on endothelial cells (20x magnification). This model was used to segment cancer cells prior to tracking in Fig 1.	Link
StarDist_BF_Neutrophil_dataset	Brightfield	IoU = 0.914 f1 = 0.969	StarDist model capable of segmenting neutrophils on endothelial cells. This model was used to segment neutrophils prior to tracking in Fig 2.	Link
StarDist_BF_Monocytes_dataset	Brightfield	IoU = 0.831 f1 = 0.941	StarDist model capable of segmenting mononucleated cells on endothelial cells. This model was used to segment mononucleated cells prior to tracking in Fig 2.	Link
StarDist_HUVEC_nuclei_dataset	Fluorescence	IoU = 0.927 f1 = 0.976	StarDist model capable of segmenting endothelial nuclei while ignoring cancer cells. Used to segment endothelial nuclei in Fig 4.	Link
StarDist_BF_cancer_cell_dataset_10x	Brightfield	IoU = 0.882 f1 = 0.968	StarDist model capable of segmenting cancer cells on endothelial cells (10x magnification). This model used in figure 7, 8 + associated supplementary figures.	Link
StarDist_AsPC1_Lifeact	Fluorescence	IoU = 0.884 f1 = 0.967	StarDist model capable of segmenting AsPC1 cells from AsPC1 channel, in addition to segmenting from background, model also segments individual cells from clusters. Used in figure 6.	Link
Stardist_MiaPaCa2_from_CD44	Fluorescence	IoU = 0.884 f1 = 0.950	StarDist model capable of segmenting MiaPaCa2 cells from CD44 channel while ignoring endothelial cells. Used in figure 6.	Link
StarDist_TumorCell_nuclei	Fluorescence	IoU = 0.558 f1 = 0.793	StarDist model capable of segmenting tumor cell nuclei from the nuclei channel while ignoring endothelial nuclei.	Link