AI-based prioritization of chest X-rays in the lung cancer diagnostic pathway: the LungIMPACT – Nature Medicine randomized controlled trial

This large, multisite, randomized trial demonstrated that AI prioritization of CXRs does not improve the speed of the lung cancer diagnostic pathway. This means that this element of AI functionality, which adds additional complexity and cost to AI installation and potentially makes clinical workflows more time-consuming, is not necessary to accelerate the lung cancer diagnostic pathway. A detailed health economic assessment will be published separately, but the costs are both significant and avoidable given the results reported in this study. Several other key findings from the secondary endpoints and exploratory analyzes are discussed below.

CXR, often the first test for lung cancer patients referred for diagnosis, is the most commonly performed imaging test. More than 7 million tests are carried out in England each year, including around 2.2 million CXRs referred from primary care alone²⁰. After an abnormal CXR or presentation in a high-risk patient, the NOLCP prescribes a rapid…