Deep learning-based segmentation of acute pulmonary embolism in cardiac CT images
-
By
-
Ehsan Amini
-
Georg Hille
-
Janine Hürtgen
-
Alexey Surov
-
Sylvia Saalfeld
-
September 25, 2025
-
Clinical Scorecard: Automated Segmentation of Acute Pulmonary Embolism in Cardiac CT Scans Using Deep Learning Techniques
At a Glance
| Category | Detail |
|---|
| Condition | Acute pulmonary embolism (APE) caused by obstructive substances blocking the pulmonary artery |
| Key Mechanisms | Blockage of pulmonary artery disrupting blood supply to lung tissues, leading to variable symptomatic presentation and mortality |
| Target Population | Patients undergoing cardiac CT scans suspected of APE |
| Care Setting | Radiology and intensive care settings depending on risk stratification |
Key Highlights
- APE diagnosis relies on computer tomographic pulmonary angiography (CTPA) as the gold standard.
- Manual segmentation of APE in CTPA is time-consuming, subjective, and labor-intensive.
- Deep learning models like nnU-Net and VT-UNet enable automated, precise 3D segmentation of APE to aid risk stratification.
Guideline-Based Recommendations
Diagnosis
- Use CTPA imaging to detect and evaluate APE.
- Assess derived parameters such as right ventricle enlargement and epicardial adipose tissue for mortality risk.
- Perform accurate delineation of emboli location, volume, and morphology for risk stratification.
Management
- High-risk patients should receive thrombolytic treatment and close surveillance in intensive care.
- Low-risk patients may be managed with anticoagulation without intensive care.
Monitoring & Follow-up
- Close monitoring in intensive care for high-risk patients based on imaging and clinical parameters.
Risks
- Delayed or inaccurate diagnosis can lead to mortality rates up to 30%.
- Misdiagnosis due to imprecise emboli segmentation may affect treatment decisions.
Patient & Prescribing Data
Patients diagnosed with acute pulmonary embolism via CTPA imaging
Risk stratification based on imaging segmentation guides treatment intensity from anticoagulation to thrombolysis and intensive care.
Clinical Best Practices
- Employ CTPA as the diagnostic gold standard for suspected APE.
- Utilize automated deep learning segmentation tools to reduce reading time and improve accuracy.
- Collaborate with experienced radiologists for manual correction and validation of automated segmentations.
- Apply risk stratification based on emboli characteristics to guide appropriate treatment pathways.
References
