Computerized Detection of Lung Nodules with an Enhanced False Positive Reduction Scheme

Computed tomography (CT) scan of lungs produces high volume of data, which is difficult to assess manually. Hence, computer-aided detection (CAD) of pulmonary nodules has become a major area of interest in biomedical imaging. Reducing the number of false positives (FPs) is considered a high priority for enhancement of any CAD system. Here we report a novel hybrid learning scheme for reducing the number of FPs in a computerized lung nodule detection system. This novel scheme consists of two main stages, namely fuzzy c-means clustering and iterative linear discriminant analysis. The main advantage of the proposed iterative linear discriminant analysis is its case adaptive nature designed to maintain a good level of sensitivity. We compare the results obtained from this hybrid scheme with a rule-based FP reduction approach and show the superiority of our novel scheme.