A team of Standford researchers has developed a new algorithm that is excellent for diagnosing pneumonia. In fact, the new algorithm, called CheXNet, can diagnose up to 14 different medical conditions and is better at diagnosing pneumonia than radiologists working alone.
As Pranav Rajpurkar, a graduate student in the Machine Learning Group at Stanford and co-lead author of the study explains, interpreting X-rays to diagnose conditions such as pneumonia is very challenging. At least partly, this is because there is a lot of variability in the diagnoses that radiologists arrive at. So naturally, the team became interested in developing machine learning algorithm that could learn from chest X-ray diagnoses and make more accurate diagnoses, says Rajpurkar.
Matthew Lungren, MD, MPH, assistant professor of radiology at the School of Medicine, explains that the motivation behind developing an algorithm that can diagnose diseases is to aid in the interpretation task that is often clouded by an intrinsic limitation of human perception and bias. Ultimately, the goal is to reduce errors and improve diagnoses. Lungren adds that a deep-learning model such as CheXNet could improve health care delivery generally, “across a wide range of settings.”.
Stanford University Medical Center via ScienceDaily (https://phys.org/news/2017-11-algorithm-pneumonia-radiologists.html)