Radiologist-AI combo has highest potential to improve pulmonary embolism detection, experts charge
A radiologist-artificial intelligence combination represents the highest potential to improve pulmonary embolism diagnoses, experts charge in new research published Wednesday.
PE is a life-threatening condition, with such blockages of one of the lungs’ main arteries causing over 300,000 deaths in the U.S. each year. Radiologists typically diagnose blood clots in the lungs with CT pulmonary angiography, or CTPA, using contrast to create detailed images of the lung’s blood vessels.
“However, growing workloads and a global radiologist shortage may delay interpretation,” experts led by Shlomit Goldberg-Stein, MD, with Northwell Health, write May 13 in the Radiological Society of North America’s Radiology: Artificial Intelligence.
In new Neiman Health Policy Institute Research, scientists explored the use of AI to aid in this task. They retrospectively applied technology from vendor Aidoc to assess images gathered at an integrated hospital network between 2021 and 2023. All CTPA images underwent real-time AI analysis and radiologist interpretation for pulmonary embolism. Any physician-AI disagreements triggered further review by an additional expert thoracic radiologist. The study included a total of 32,500 CT pulmonary angiography exams from over 29,000 adults at an average age of 62.
About 9.9% of examinations were positive for PE, and radiologists and AI agreed on about 97.8% of cases. Of positive instances, 85% were initially identified by AI and confirmed by radiologists, “showing the value of AI triage," Matthew Barish, MD, also with New York-based Northwell Health, said in a statement. However, 15% were detected only with radiologist involvement, “demonstrating the importance of subsequent radiologist review when AI was negative,” he added. Radiologists were correct in about 88.7% of disagreements (or 638 cases), while AI was correct in the other 11.3%. Rad-AI concordance was higher for negative exams (98.2%) than positive ones (93.8%), “underscoring the algorithm’s strength in helping rule out PE."
“AI-informed radiologists,” who used the technology, achieved 99.2% sensitivity for pulmonary embolism detection. Radiologist-AI agreement was highest for acute and central emboli—cases associated with the greatest clinical urgency and mortality risk, the study found.
“This suggests the algorithm is most reliable in precisely the clinical scenarios where triage has the greatest potential to impact patient outcomes” Goldberg-Stein, MD, a professor of radiology and director of artificial intelligence at the Zucker School of Medicine at Hofstra/Northwell, said in an announcement from the Neiman Policy Institute. “Ultimately, these results demonstrate that the combined expertise of radiologists with AI offers the best potential to improve PE identification in clinical practice.”
Read more about the results in RSNA’s Radiology: AI.
Marty Stempniak has covered healthcare since 2012, with his byline appearing in the American Hospital Association's member magazine, Modern Healthcare and McKnight's. Prior to that, he wrote about village government and local business for his hometown newspaper in Oak Park, Illinois. He won a Peter Lisagor and Gold EXCEL awards in 2017 for his coverage of the opioid epidemic.