One of the current challenges for CAD systems is detecting large nodules, which are rare, said presenter Arnaud Arindra Adiyoso Setio of Radboud University Medical Center. As these large nodules have a high degree of suspicion for being malignant, it's important that they be caught during screening, he said.
As a result, the group developed a dedicated CAD system for large-nodule detection on chest CT scans. The proposed CAD algorithm can be combined with existing CAD systems to improve their performance, Setio said.
In their study, the researchers found that their software could identify the vast majority of highly suspicious lesions on thoracic CT scans, with only a small number of false positives, he said.
"Such an optimized CAD system could be proved useful in improving the performance of general CAD systems when combined," Setio told AuntMinnie.com.