Chest radiography (CXR) is the primary tool to visualize thoracic pathologies in the ICU. However, accurate interpretation of CXRs is challenging, even when imaging is performed under optimal conditions.

This results in both a considerable variability of reports and errors, possibly causing harm to patients.

These problems are aggravated in the intensive care unit (ICU) by the suboptimal quality of the images due to varying patient positioning and the superimposition of multiple conditions and internal and external devices.

While AI for CXR reporting in the ICU lags behind the performance of other AI radiology systems and is not (yet) ready for prime time, it is urgently needed and holds great promise to support critical decision-making.

Read More