This repo contains information and code used for analysing images from the GeoMx digital spatial profiler for the T-cell paper (add link). The images were stained with 4 markers.
The image analysis was performed with the open source software QuPath.
The nuclei segmentation was performed with a fine-tuned "nuclei" Cellpose model using the Cellpose GUI. The "cyto" model was fine-tuned for the cytoplasm segmentation.
For running the QuPath classification:
- run_cellpose.groovy runs the nuclei model with cell expansion
- run_QuPath_classifiers_RF.groovy classifies the cells to their cell types
For running the Cellpose classification
- classify_cells_by_cellpose_segmentation.groovy runs segmentation and overlap checking with classification
For exporting images and labeled images with class maps:
- export_labeled_images_with_classMap.groovy The output is later used in calculate_metrics.ipynb to calculate the metrics used for evaluation of the different methods.