Detection of Juxtapleural Nodules in Lung Cancer Cases Using an Optimal Critical Point Selection Algorithm

Detection of lung cancer through image processing is an important tool for diagnosis. In recent years, image processing techniques have become more widely used. Lung segmentation is an essential pre-processing step for most (CAD) schemes. An automated system is proposed in this paper for identifying lung cancer from the analysis of computed tomography images by performing nodule segmentation using an optimal critical point selection algorithm (OCPS) which improves the detection of shape- and size-based juxtapleural nodules located at the lung boundary. A suspect area of nodule is obtained with the help of a bidirectional chain code (BDC) approach and the OCPS This algorithm is used to reduce the time consumption to detect the lung nodule and thereby reduce the computational complexity. Shape and size features are extracted for the area between two critical points to facilitate classification as nodule or non-nodule with the help of a support vector machine and random forest classifiers. This automated method was tested on computed tomography (CT) studies from the lung imaging database consortium (LIDC). The results are compared with the existing techniques using various performance measures such as precision rate, recall rate, accuracy and F-measure. The obtained experimental results indicate that the OCPS combined with a random forest classifier performs better in terms of performance evaluation metrics than existing approaches, with less requirement for computation.