Computer-aided detection for computed tomographic colonography is as effective as optical colonoscopy for detection of significant adenomas.
However, the unavoidable interaction between computer-aided detection and the reader has not been addressed.
Dr Steve Halligan and colleagues from England evaluated this interaction.
The research team evaluated 10 readers trained in computed tomography, but without special expertise in colonography.
The readers interpreted computed tomographic colonography images of 107 patients.
Of these patients, 60 had 142 polyps.
|Per-patient sensitivity increased significantly in 70% of readers|
First the readers interpreted the images without computer-aided detection.
The readers then interpreted the images with computer-aided detection after a temporal separation of 2 months.
The researchers determined per-patient and per-polyp detection by comparing responses with known patient status.
With computer-aided detection, 41 of the 60 patients with polyps were identified more frequently by readers.
Per-patient sensitivity increased significantly in 70% of readers, while specificity dropped significantly in only 1 patient.
The researchers noted that polyp detection increased significantly with computer-aided detection.
On average, 12 more polyps were detected by each reader.
The team found that small- and medium-sized polyps were significantly more likely to be detected when prompted correctly by computer-aided detection.
However, overall performance was relatively poor; even with computer-aided detection.
The researchers identified that, on average, readers detected only 10 polyps that were10 mm or more, and 24 polyps 6 mm or more.
The team observed that interpretation time was shortened significantly with computer-aided detection by 2 minutes for patients with polyps.
The research team noted that interpretation time was shortened with computer-aided detection by 3 minutes for patients without.
Overall, 90% of readers benefited significantly from computer-aided detection, either by increased sensitivity and/or by reduced interpretation time.
Dr Halligan's team concludes, “Computer-aided detection for computed tomographic colonography significantly increases per-patient and per-polyp detection.”
“In addition, it significantly reduces interpretation times but cannot substitute for adequate training.”