![thoracic](http://media.assettype.com/healthday%2Fimport%2F32019209%2Forigin.jpg?w=480&auto=format%2Ccompress&fit=max)
WEDNESDAY, Sept. 27, 2023 (HealthDay News) -- Artificial intelligence (AI) tools show moderate-to-high sensitivity for detecting airspace disease, pneumothorax, and pleural effusion on chest radiographs, according to a study published online Sept. 26 in Radiology.
Louis Lind Plesner, M.D., from Herlev and Gentofte Hospital in Copenhagen, Denmark, and colleagues examined the diagnostic accuracy of four commercially available AI tools for detection of airspace disease, pneumothorax, and pleural effusion on chest radiographs in a retrospective study. A total of 2,040 patients comprised the dataset, and of these, 669 (32.8 percent) had target findings.
The researchers found that the AI tools demonstrated areas under the receiver operating characteristic curve ranging from 0.83 to 0.88, 0.89 to 0.97, and 0.94 to 0.97 for airspace disease, pneumothorax, and pleural effusion, respectively. The corresponding sensitivities ranged from 72 to 91 percent, 63 to 90 percent, and 62 to 95 percent. For all target findings, negative predictive values ranged from 92 to 100 percent. Specificity was high for chest radiographs with normal or single findings (range, 85 to 96 percent, 99 to 100 percent, and 95 to 100 percent for airspace disease, pneumothorax, and pleural effusion, respectively) and was markedly lower for chest radiographs with four or more findings (corresponding ranges, 27 to 69 percent, 96 to 99 percent, and 65 to 92 percent).
"Future studies could focus on prospective assessment of the clinical consequence of using AI for chest radiography in patient-related outcomes," the authors write.
Several authors disclosed ties to the pharmaceutical and medical device industries.