The core issue in computer-aided detection (CAD) of lung nodules in CT (computed tomography) images is the problem of detecting nodules attached to vessels as well as isolated nodules while keeping the number of false positives due to pulmonary blood vessels at a low level. As a solution to this problem, we propose a novel 3D feature termed gdiminution indexh that has a capability to differentiate between pulmonary blood vessels and nodules attached to vessels. The diminution index deals with a 3D region obtained by exploring the nodule candidate at hand and its surrounding in a centrifugal direction with respect to the nodule candidate, and gives a measure of diminution of the 3D region in the centrifugal direction. In this study, a straightforward rule-based nodule detection scheme that employs the diminution index and 5 auxiliary features was evaluated using a data set prepared by placing simulated nodules contiguous to pulmonary blood vessels in clinical CT images. The scheme could detect the majority (78%) of simulated nodules with a moderate rate of false positives (5.3 per scan), indicating that the diminution index has a promising capability in computer-aided detection of lung nodules in CT images.
All Science Journal Classification (ASJC) codes