Although CT colonography has become a technically feasible screening tool for colorectal cancer, few radiologists have implemented the exam as a service line in their practice, perhaps because exam interpretation is time consuming, requiring analysis of 2D and 3D data sets. A training method for virtual dissection reading, along with concurrent use of computer-aided detection (CAD) technology, may shorten the learning period and significantly increase diagnostic values achieved by inexperienced interpreters.
“CT colonographic reading is a complicated process encompassing two problems: detection of a possible lesion and then its characterization,” wrote the authors of research published this month in Radiology.
In a study undertaken in the department of medical imaging at the Clinique Saint-Joseph in Liege, Belgium, a team evaluated both the learning curves and reading times of inexperienced readers who used a virtual dissection reading method for CT colonography data sets, with and without concurrent CAD.
A retrospective data set of 100 CT colonography exams on 49 men and 51 women were assembled by the researchers. Four radiologists with no prior exposure to the CT colonographic software (Autodissection, version 7.4.63; GE Healthcare) and no experience in CT colonographic reading, were asked to review the data sets by using the dissection views for detection of possible findings and by using the 2D and 3D views for characterization.
Two of the readers had use of CAD software (Colon VCAR, beta version, GE) and two did not. Reading reports by the interpreters included reading times, measurement and localization of the abnormalities, and characterization as a polyp, cancer, or nonpolypous finding.
“The 100 blinded CT colonographic data sets were divided into five groups in which they were randomly mixed and presented in the same reading order to the four readers,” the authors noted.
As expected, average reading times decreased from the first to the fifth set of studies, from approximately 19 minutes to nearly 12 minutes in both the CAD and non-CAD user group. However, the use of CAD increased the detection rate.
“In our study, concurrent CAD reading appeared beneficial because average detection rate was high, 28 (70 percent) of 40, right from the first session,” the researchers wrote.
In addition, they noted that the diagnostic values with the use of CAD stayed above the unassisted values for the entire study.
The group found that the virtual dissection reading method enables a short learning period for CT colonography, and their results indicate that concurrent CAD reading can further reduce the training time for inexperienced readers.