Radiologists connect via AI ‘hives’ to enhance pneumonia diagnosis
A new technique that connects a small group of radiologists together using artificial intelligence (AI) algorithms performed better than individual doctors or algorithms alone in detecting pneumonia on x-rays.
The method, dubbed “hive mind,” uses Swarm AI technology to pool radiologists together and harness their combined, individual perspectives in real-time to determine an optimal diagnosis. The study was presented Monday, Sept. 10 at the Society for Medical Imaging informatics in Medicine (SIIM)’s Machine Intelligence in Medical Imaging conference.
“Diagnosing pathologies like pneumonia from chest x-rays is extremely difficult, making it an ideal target for AI technologies,” said Matthew Lungren, MD, and an author involved in the study, in a statement. “The results of this study are very exciting as they point towards a future where doctors and AI algorithms can work together in real-time, rather than human practitioners being replaced by automated algorithms.”
Initially, a group of eight radiologists at separate locations individually analyzed 50 chest x-rays. After a few seconds, the group worked as a unit, or “intelligent swarm,” the authors wrote, to come to a diagnostic conclusion. After a set of 50 probabilities were generated, the same x-rays were fed into CheXNet—currently the top automated system for diagnosing pneumonia on chest x-ray.
Overall, the new AI method reduced the average error rate by 33 percent compared to diagnoses made by individuals. Additionally, the Swarm AI system proved 22 percent more accurate in binary classification than the CheXNet system.
“We find that a Swarm AI system that combines real-time human input with intelligence algorithms is significantly more accurate in diagnosing pneumonia than a state-of-the-art software-only ML system,” the study authors concluded. “This suggests that Swarm AI may be a powerful tool for establishing Ground Truth for use in training and for validating machine learning systems.”
Matt joined Chicago’s TriMed team in 2018 covering all areas of health imaging after two years reporting on the hospital field. He holds a bachelor’s in English from UIC, and enjoys a good cup of coffee and an interesting documentary.