Automatic lung nodules detection in computed tomography images using nodule filtering and neural networks

In this study a new computer-aided detection (CAD) system presented that detect small size nodules (larger 3 mm) in High Resolution CT (HRCT) images. In the first step, the lung region is extracted, then with a type of 3D filtering nodule supposed cases is founded. In the final step, a neural network is used for false positive reduction. For filtering nodule cases from other objects in images, it´s used a cylindrical filter. The detection performance was evaluated experimentally using lung LIDC image database. Suitable results show that the use of the 3D model and the features analysis based FPs reduction can accurately detect nodules in HRCT images.