Researchers have developed a deep learning-based automatic detection algorithm (DLAD) that can detect malignant pulmonary nodules on chest x-rays better than physicians, sharing their findings in a new study published by Radiology.
The authors developed the DLAD by collecting more than 43,000 chest x-rays from more than 34,000 patients obtained from 2010 to 2015 at a single hospital. Its performance was compared to 18 physicians, including nine board-certified radiologists.
Overall, the DLAD had a higher area under the receiver operating characteristic curve (0.92-0.99) than 17 of the 18 physicians. It also had a higher jackknife alternative free-response receiver-operating characteristic figure of merit (0.831-0.924) than 15 of 18 physicians. In addition, all physicians were better at detecting nodules when using the algorithm as a second reader.
“Our study results demonstrated that DLAD could accurately detect malignant pulmonary nodules on chest radiographs with better performance than that of physicians, and that it enhanced performance of physicians when used as a second reader,” wrote author Ju Gang Nam, Seoul National University Hospital and College of Medicine in South Korea, and colleagues. “More specifically, DLAD showed high specificity and was able to detect 100 percent of high conspicuity nodules (score of ≥4), most large (> 3 cm) nodules, and more nodules in overlapped areas than the four groups of physicians in our study.”
The researchers noted that the DLAD could not accurately detect “small and less conspicuous” nodules. Because it was trained with x-rays labeled and annotated by radiologists, they explained, the limitations of those radiologists is present in the algorithm itself.
“In the future, establishing an algorithm supervised by annotations on chest radiographs containing radiologically undetectable nodules by using CT as the reference standard is warranted,” the authors wrote.
Michael has more than 16 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.