Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/dataset.json

Name

Augmented RSPECT: Augmented RSNA Pulmonary Embolism CT Dataset

Link

https://github.com/dila-ai/Augmented_RSPECT

Indexing

Keywords: pulmonary embolism, CTPA, bounding boxes, object detection, pulmonary arterial tree, anatomic localization

Content: CH, CT, RS

RadLex: RID35976, RID12778, RID49837, RID33250, RID4834, RID13060

SNOMED: 59282003

Author(s)

Matias F. Callejas

Hui Ming Lin

Thomas Howard

Matthew Aitken

Marc Napoleone

Laura Jimenez-Juan

Robert Moreland

Shobhit Mathur

Djeven P. Deva

Errol Colak

Organization(s)

Department of Medical Imaging, Unity Health Toronto, University of Toronto

Department of Medical Imaging, Faculty of Medicine, University of Toronto

Li Ka Shing Knowledge Institute, St Michael’s Hospital, Unity Health Toronto

Contact

Errol Colak, Department of Medical Imaging, Unity Health Toronto, University of Toronto, 30 Bond St, Toronto, ON, Canada M5B 1W8

Funding

E.C. supported by the Odette Professorship in Artificial Intelligence for Medical Imaging, St Michael’s Hospital, Unity Health Toronto.

Ethical review

No institutional review board approval was obtained as the CTPA studies are part of a publicly available dataset.

Comments

This resource provides anatomy-specific bounding box annotations for pulmonary emboli on a subset of the RSNA STR Pulmonary Embolism CT (RSPECT) training dataset.

Date

Published: 2023-05-03

References

[1] Callejas MF, Lin HM, Howard T, Aitken M, Napoleone M, Jimenez-Juan L, Moreland R, Mathur S, Deva DP, Colak E.. "Augmentation of the RSNA Pulmonary Embolism CT Dataset with Bounding Box Annotations and Anatomic Localization of Pulmonary Emboli". Radiology: Artificial Intelligence. 2023-05-03. doi:10.1148/ryai.230001. PMID: 37293344. PMCID: PMC10245177.

Dataset

Motivation

To provide granular, anatomy-specific localization of pulmonary emboli to facilitate object detection and localization model development.

Sampling

20% of positive training studies from each contributing institution were selected via stratified random sampling.

Missing information

No patient demographics provided; no explicit train/validation/test partitions; license for the GitHub annotations not specified in the article.

Relationships between instances

Bounding boxes are linked to DICOM instances via StudyInstanceUID, SeriesInstanceUID, and SOPInstanceUID, enabling correspondence to original RSPECT labels.

Noise

Studies with substantial motion, poor opacification, or flow artifacts were flagged via QA labels and excluded from the final dataset.

External data

Augments the RSNA STR Pulmonary Embolism Detection Challenge dataset hosted on Kaggle.

Confidentiality

Derived from a publicly available de-identified dataset (RSNA STR Pulmonary Embolism Detection Challenge).