AI Software Facilitates 22.1 Percent Reduction in Chest CT Review Time

Could an emerging artificial intelligence (AI) software boost efficiency for cardiothoracic radiologists when reviewing chest computed tomography (CT) scans?

In a recently published prospective study involving 390 patients who had outpatient chest CTs, researchers compared the use of an AI software platform (AI-Rad Companion Chest CT, Siemens Healthineers) versus non-AI assessment. For the three cardiothoracic radiologists who assessed the chest CT scans, the use of AI assistance resulted in a mean interpretation time of 328 seconds in comparison to a mean of 421 seconds for non-AI-assisted interpretation. According to the study authors, the 93 second difference per chest CT scan represented a 22.1 percent reduction in mean interpretation time.

“These findings illustrate the utility of integrating an AI support program into the clinical workflow to improve radiologists’ interpretation efficiency,” wrote Tilman Emrich, M.D., who is affiliated with the Division of Cardiovascular Imaging in the Department of Radiology and Radiological Science at the Medical University of South Carolina, and colleagues. “ … If a radiologist were to interpret an average of 40 chest CT scans per day, then approximately one hour would be saved over the course of a typical workday using AI support.”

Out of the 390 chest CT scans, 200 were performed with contrast enhancement, according to the study. The researchers noted a mean interpretation time reduction of 83 seconds (or a 20 percent decrease) with AI assistance for contrast-enhanced CT. For the 190 chest CT scans without contrast enhancement, Emrich and colleagues reported a 104 second reduction or a 24.2 percent decrease in the mean interpretation time with AI assistance.

The use of the AI software platform also facilitated a 117 second or a 25.7 percent reduction in mean interpretation time for positive chest CT scans without new findings and a 92 second or 20.4 percent reduction in mean interpretation time for positive scans with new findings, according to the study.

In regard to the AI-Rad Companion Chest CT used in the study, Emrich and colleagues said the AI software provides automated detection, segmentation and measurement of lung lesions; assesses pulmonary parenchyma to help quantify emphysema; and quantifies thoracic aortic diameters.

“While incorporating AI into radiologists’ daily routines may require an initial learning curve, it may ultimately provide an effective tool for reducing the time and effort spent on performing repetitive tasks,” suggested Emrich and colleagues.

Noting that the chest CT scans were interpreted during the course of real-world clinical care, the study authors acknowledged the scans in the AI-assisted arm of the study were not the same as those in the non-AI-assisted cohort. However, they said the range of patient and scan characteristics were not significantly different between the two groups. With the data coming from a single institution and a single AI software platform, the researchers said future studies involving other institutions and different AI modalities are necessary prior to general extrapolation of the aforementioned study findings.