Pulmonary Embolism-dataset-CSV

This dataset provides curated and standardized CSV files derived from the publicly available RSNA Pulmonary Embolism (PE) Detection dataset. The data were reformatted and quality-checked to support explainable AI development and reproducible research in pulmonary embolism detection.
Each record represents metadata extracted from CT pulmonary angiography (CTPA) studies, including study-level and image-level identifiers and diagnostic annotations. The dataset includes quality assurance (QA) indicators, imaging artifacts, and diagnostic labels such as PE presence, chronicity, and localization (left, right, central).
Columns overview:
StudyInstanceUID: Unique identifier for each CT study (exam)SeriesInstanceUID: Unique identifier for each image series within the studySOPInstanceUID: Unique identifier for each image instancepe-present-on-image: Indicates whether pulmonary embolism (PE) is visible in the imagenegative-exam-for-pe: Indicates that the exam contains no PE-positive imagesqa-motion / qa-contrast / flow-artifact: Radiologist quality assurance flags for motion, contrast, and flow artifactsrv-lv-ratio-gte-1 / rv-lv-ratio-lt-1: Indicators of right-ventricle to left-ventricle ratio findingsleftsided-pe / rightsided-pe / central-pe: Indicates the anatomical location of PEchronic-pe / acute-and-chronic-pe: Indicates chronic or mixed PE findingstrue-filling-defect-not-pe: Marks filling defects that were not pulmonary embolismsindeterminate: Indicates studies where no definitive exam-level label could be assigned due to QA issuesSource:
Data derived from the RSNA Pulmonary Embolism Detection dataset, publicly available on Kaggle (https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection).
Ethics:
All data were originally de-identified and shared under the RSNA open data license. No additional identifiable patient data were included in this release.