Simulation-Trained Deep Learning for Automated Cell-Based HLA Antibody Assay Interpretation in Pre-Transplant Diagnostics - Images

Dataset description This Zenodo record contains anonymized image datasets used for the development, validation, and demonstration of automated image-analysis workflows for cell-based diagnostic assays. The archive includes three complementary image collections: experimental_readout_images.tar.gz This archive contains assay images that were scored during routine diagnostic evaluation. The images represent real-world experimental readouts and were used to assess the performance and generalizability of the automated image-analysis workflow under routine laboratory conditions. human_annotations.tar.gz This archive contains images that were independently reviewed and cross-annotated by human experts. The corresponding annotations provide reference data for evaluating agreement between automated image analysis and human assessment. show_images.tar.gz This archive contains representative validation images intended for testing and demonstrating the image-analysis workflow through the associated web application. In order to reproduce the source code with the figures, extract this to ./show_images. When uploading these images to the web application (https://www.alloviewer.org), the plate layout should be configured as: Scan pattern: Vertical snake Starting well: A1 The order of the images within each folder must be retained because image assignment is based on the acquisition sequence. IMPORTANT! The images have been stripped of all identifying metadata and image-specific metadata were also removed. In order to use the pixel dimensions, contact us.