Ai Matches Radiologists in Detecting Pulmonary Embolism
Ai Matches Radiologists in Detecting Pulmonary Embolism
Publish Date: 2026-05-13 13:50:00
Source Domain: respiratory-therapy.com
-
High Concordance Rates: An AI algorithm for detecting pulmonary embolism achieved a 97.8% agreement rate with radiologist interpretations across more than 32,000 CTPA scans in an integrated healthcare system.
-
Critical Physician Oversight: While the AI tool assisted in triage, radiologists identified 15% of confirmed cases that the software missed, underscoring the importance of human oversight in critical decisions.
-
Urgent Case Reliability: The agreement between the AI and physicians was highest for acute and central emboli, which carry the highest mortality risk.
-
Study Design and Performance Metrics: The study analyzed 32,501 CTPAs over 18 months, showing superior concordance rates for negative exams (98.18%) compared to positive exams (93.75%).
-
Importance of Combined Approach: AI-assisted radiologists achieved a 99.2% sensitivity for pulmonary embolism detection, and showed improved accuracy when used in conjunction with human review.
-
Expert Validation: Expert adjudication by thoracic radiologists confirmed that in cases of disagreement between AI and radiologists, radiologists’ interpretation was correct in 88.7% of cases.
