Detection and Triage of Cancer-associated Incidental Pulmonary Embolism
model2025-12-03https://doi.org/10.1148/atlas.1764792733523
102

Overview

Schema Version

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

Name

Detection and Triage of Cancer-associated Incidental Pulmonary Embolism

Link

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10698599

Indexing

Keywords: Cancer-associated Incidental Pulmonary Embolism, Pulmonary Embolism, Artificial Intelligence, Cancer, CT Imaging
Content: CT, CH, OI
RadLex: RID45946, RID13060, RID4834

Author(s)

Peder Wiklund, MD, PhD
Koshiar Medson, MD, PhD

Organization(s)

Department of Radiology, Region Halland
Department of Radiology and Functional Imaging, Karolinska University Hospital
Department of Physiology and Pharmacology, Karolinska Institutet

Version

1.0

Contact

es.dnallahnoiger@dnulkiw.redep

Funding

Authors declared no funding for this work.

Ethical review

Approved by the Swedish Ethical Review Authority; informed consent was waived.

Date

Updated: 2023-11-01
Published: 2023-10-18
Created: 2022-12-14

References

[1] Wiklund P, Medson K. "Use of a Deep Learning Algorithm for Detection and Triage of Cancer-associated Incidental Pulmonary Embolism". Radiology: Artificial Intelligence. 2023 Nov;5(6):e220286.. 2023-11-01. doi:10.1148/ryai.220286. PMID: 38074784. PMCID: PMC10698599.

Model

Architecture

Commercial deep learning algorithm optimized for study-level classification of pulmonary embolism on CT; details of training/validation previously published.

Availability

Commercial, cloud-based solution for iPE/PE detection and triage (Aidoc BriefCase; Aidoc Medical).

Clinical benefit

Increased detection rate of incidental pulmonary embolism in cancer patients and significantly reduced report turnaround time and time to treatment in clinical practice.

Clinical workflow phase

Triage and workflow optimization within radiology reading workflow; clinical decision support for expedited patient evaluation.

Degree of automation

Data upload and AI analysis fully automated; results presented in a separate widget at radiologist and radiographer CT workstations; radiologist reviews all positive AI results and initiates communication.

Indications for use

Detection and triage of incidental pulmonary embolism in adult patients with cancer undergoing contrast-enhanced chest (± abdominal) CT performed for cancer evaluation in a nonurgent setting within a hospital radiology department.

Input

Contrast-enhanced chest CT (and chest+abdominal CT) examinations performed for cancer staging/treatment evaluation.

Instructions

All positive AI results are immediately evaluated by a radiologist; in true-positive iPE cases, the radiology team contacts the referring physician/department or on-call physician to initiate prompt clinical evaluation; communication with patients managed by radiographers with radiologist support.

Limitations

Single-center retrospective cross-sectional study; cases that were both AI- and report-negative were not rereviewed, so true sensitivity could not be evaluated; two reported iPE cases were AI false-negative; three AI-positive but report-negative cases were reclassified as true-positive on retrospective review; overlap with COVID-19 era, though COVID-19–specific referrals were excluded.

Output

CDEs: RDE1704, RDE2934, RDE2607
Description: Study-level classification/alert indicating suspicion of pulmonary embolism; triage notification presented in a dedicated widget.

Recommendation

Use as a triage tool to expedite detection and management of incidental PE in oncology CT workflows, with radiologist review of AI alerts.

Sustainability

Mean AI processing time 10.8 minutes per study.

Use

Intended: Detection, Triage
Out-of-scope: Other

User

Intended: Radiology technologist, Radiologist