New computer-assisted diagnosis (CAD) techniques with increased accuracy for detection of lung nodules and improved algorithms to eliminate false-positive findings were among the developments reported today by researchers from New York and California.A team from New York University presented their work on an interactive CAD system that aids radiologists using low-dose, multislice CT for detection of pulmonary nodules.
The researchers used a visual display of their CAD scheme's 3D-rendering capabilities to demonstrate how the system detects small nodules and shows connectivity between nodules and blood vessels. The system, they said, can also increase a study's sensitivity with fewer than two false positives per case.
In another presentation, David S. Paik, Ph.D., of Stanford University's department of radiology, discussed his development of a CAD algorithm capable of detecting small nodules and eliminating false positives in a two-phase procedure.
Paik's study consisted of 304 CT images of a 54-year-old man, which showed 60 native nodules confirmed by three radiologists, and 60 software-simulated nodules ranging from 1.5 to 13.3 mm. The system detected nodules 6 mm and larger with 100% sensitivity and only one false positive, and nodules larger than 3 mm with 92.5% sensitivity and 10 false positives. There was no statistical difference in CAD performance on native versus simulated nodules.
"For any CAD system's algorithms, the contribution is two-fold: increasing accuracy and sensitivity and the potential to increase efficiency for radiologists," Paik said.
CAD schemes can reduce false positives at a better rate than the radiologist's eye alone, avoiding unnecessary workups, and potentially reducing the need for biopsies.
Many radiologists involved in lung cancer screening are interested in CAD but have not felt the need to incorporate it into their practice. They are comfortable with the way they do screenings and do not need extra help to handle case workloads, said Dr. K. Ty Bae, an assistant professor of radiology at the Mallinckrodt Institute. But the fact that CAD systems are helping to increase sensitivity and minimize the number of false positives, combined with the increasing demand for diagnostic services and a shortage of specialists, may soon cause the hold-outs to reconsider, he said.
