Post

New research · Internal Medicine
European journal of radiology open · 3h
AI / informaticsEuropean journal of radiology open · 2026

Artificial intelligence for pulmonary embolism detection: Is it comparable to radiology residents?

Periktioni Crina Chrysostomou, Alexandra Platon, Cyril Jaksic … Pierre-Alexandre Poletti
Read paper
Internal MedicineAI / informatics

Artificial intelligence software detected pulmonary embolism with lower sensitivity than radiology residents

Artificial intelligence for pulmonary embolism detection: Is it comparable to radiology residents?

Periktioni Crina Chrysostomou … Pierre-Alexandre Poletti
European journal of radiology open · 2026
Purpose

To assess the diagnostic performance of a commercial artificial intelligence (artificial intelligence) software in detecting pulmonary embolism (pulmonary embolism) in emergency settings, compared with on-call radiology residents.

Methods

All consecutive emergency CT pulmonary angiographies (CT pulmonary angiographies) performed over a 3-month period in the emergency department of a university hospital in patients with suspected pulmonary embolism, initially interpreted by the radiology residents during the emergency workflow and subsequently verified and approved by a board-certified radiologist, were concomitantly analyzed by an artificial intelligence software for the presence of pulmonary embolism.

n = 82 patients
Results

Artificial intelligence correctly identified 89% of pulmonary embolism cases, versus 97.6% for residents.

97.6%
Sensitivity
89%
Specificity
More results

Among 594 CT pulmonary angiographies examinations, pulmonary embolism was present in 82 patients (13.8% prevalence), including 41 (50%) proximally located emboli.

More results

For proximal pulmonary embolism, sensitivity was 97.6% for artificial intelligence and 100% for residents (p = 1), whereas for peripheral pulmonary embolism, artificial intelligence sensitivity was 80.5% versus 95.1% for residents (p = 0.08).

“
Conclusion · 1 of 2

Radiology residents showed an overall better performance in detecting pulmonary embolism compared to the artificial intelligence software.

Conclusion · 2 of 2

The artificial intelligence software achieved good diagnostic performance for pulmonary embolism detection, particularly for proximal located pulmonary embolism.

Read paper
0 comments

No comments yet. Be the first.

Related papers

LatestFoundational
Cohort Study
13.5%
Central retinal artery peak blood-flow speed was 13.5% lower in diabetic patients.
Cohort Study
4.5 minutes
Each visual-field test averaged 4.5 minutes per eye.
Review
46.7%
nearly half of thyroid patients had Graves' disease, the most common overactive-thyroid condition
Cross-sectional
MYR 88,346.12
Average annual out-of-pocket expenditure for patients with cancer in Malaysia.
Case-control
5.4% Reduction in Odds of Neuropathy
each rise in zinc levels lowers the chance of nerve damage
Cross-sectional
SIR 9.93
risk climbs sharply when relatives cluster and clotting starts young
Cohort Study
67.8%
one-year survival was far worse than the intermediate-risk group's, roughly one in three died
Systematic Review
“Current evidence does not support universal thromboprophylaxis following ACLR.”