An AI-based cell segmentation pipeline for studying immune markers from brightfield microscopy images

In our work, we developed a fully automated pipeline for cell segmentation of whole slide brightfield microscopy images. Images are created by immunohistochemistry applying CD3, CD4, CD8, CD20, CD163 and Iba1 antibodies. The pipeline consists of two steps: 1) Filtering the tumor areas by a residual network and 2) segmenting and counting counterstained cells and cells with immunostaining using an U-Net model.

Here we provide you an AI-based cell segmentation pipeline for analysis of wholeslide brightfield microscopy images and the training data and code for training the filtering and segmentation models. 

Content:

• README
• cell_segmentation_pipeline.zip: cell segmentation pipeline, code for read images used by the pipeline and code for creating heatmaps with cell ratios used from the pipeline
• models.zip: filtering and segmentation models used in the pipeline
• filtering_training_code.zip: training code for cross validation models and model on whole data for each staining (CD3, CD4, CD8, CD20, CD163, Iba1) and for all staining together 
• data_filtering.zip
• segmentation_training_code.zip: code for preparation trainng data and code for training the segmentation classifier
• data_segmentation_1.zip,  data_segmentation_2.zip, data_segmentation_3.zip, data_segmentation_4.zip, data_segmentation_5.zip, data_segmentation_6.zip, data_segmentation_7.zip