MURA Shoulder Radiograph Quality Label Dataset (Axillary, Grashey/"True AP" views)

Dataset Description and Citation 732 labels for Shoulder Radiograph Quality on axillary and Grashey/"True AP" views. The included JSON file includes the de-identified accession name and corresponding label in the MURA dataset. Note that the actual radiographs are not included here, as those are owned by the Center for Artificial Intelligence in Medicine & Imaging group at Stanford. If using these labels, please be sure to cite the following articles: MURA: Large Dataset for Abnormality Detection in Musculoskeletal Radiographs (original dataset) Automated Shoulder Radiograph Quality Review to Support Efficient Workflow in the Emergency Department: The SQUIRE Deep Learning Ensemble (study which these labels were generated for) MURA Shoulder Radiograph Quality Label Dataset (Axillary, Grashey/"True AP" views) (this label dataset) Notes on contents: MURA_Quality_Dataset.json: contains all radiograph-image pairings of Training data. SQUIRE Manuscript -- All Model Weights foundationCXRMLP: contains weights for all five folds of foundation chest x-ray MLP after training grid_search_weight_decay: contains all optimal model weights and logs when weight decay was set to be between 2.5e-4 to 2.5e-1. weight_decay_large: contains all optimal model weights and logs when weight decay was set to 2.5. weight_decay_larger: contains all optimal model weights and logs when weight decay was set to 25. weight_decay_zero: contains all optimal model weights and logs when weight decay was set to 0.