Integrating a commercial chest CT computer-aided detection (CAD) system into a clinical radiology reporting workflow reduced reading times by as much as 44 percent, according to new research published in Academic Radiology.
A majority of automated detection and measurement systems have been implemented as standalone workstations, forcing radiologists to view studies in CAD applications and their clinical workflow system separately, wrote lead author Matthew Brown, PhD, with University of California, Los Angeles, and colleagues.
“As such the use of these standalone CAD systems are likely to increase the time to generate a report, i.e., decrease radiologist efficiency,” they wrote. “This lack of integration and proven benefit in terms of radiologist efficiency may be a contributing factor to the lack of adoption of CAD in current clinical practice.”
Brown and colleagues integrated a commercial chest CT CAD software which automatically detected and measured lung nodules, ascending and descending aorta and pleural effusion with a commercial radiology report dictation system.
The integrated system automatically prepopulated reporting templates and was evaluated using 40 scans from a public lung nodule database.
Three radiologists tested the system using two separate reading methods. The first did not include automated CAD information, while the second utilized the integrated platform.
Results showed the CAD program cut reading times by 7 to 44 percent compared to the traditional manual method. According to the study, Reader 1 experienced a 44 percent reduction in reading time, Reader 2 a total of 7 percent, and Reader 3 saved 19 percent.
The total number of nodules recorded through each method was similar and readers “tended” to accept the automated measurements during their readings, according to authors.
Unlike typical CAD research which focuses on sensitivity and specificity or inter-reader agreement, Brown et al. focused exclusively on the speed of report generation. Although their study provided “important insights” they maintained that further efforts are needed to examine efficiency improvements.
“Overall the total number of nodules reported under the two reading methods is similar, so the data suggests that in general for this CAD system it produces comparable levels of sensitivity while increasing efficiency for the type and size of nodules included,” authors wrote. “However, it is an area requiring further study…”
Matt joined Chicago’s TriMed team in 2018 covering all areas of health imaging after two years reporting on the hospital field. He holds a bachelor’s in English from UIC, and enjoys a good cup of coffee and an interesting documentary.