Automatic detection of solitary pulmonary nodules using swarm intelligence optimized neural networks on CT images

Lung Cancer is one of the most dangerous diseases that cause a large number of deaths. Early detection and analysis will be the only remedy. Computer-Aided Diagnosis (CAD) plays a key role in the early detection and diagnosis of lung cancer. This paper develops a CAD system that focus on new heuristic search algorithm to optimize the Back Propagation Neural Network (BPNN) in characterizing nodule from non-nodules. The proposed CAD system consists of four main stages: (i) image acquisition (ii) lesion detection, (iii) texture feature extraction and (iv) tumor characterization using a classifier. The optimization mechanism employs Particle Swarm Optimization (PSO) with new inertia weight for NN in order to investigate the classification rate of these algorithms in reducing the problems of trapping in local minima and the slow convergence rate of current evolutionary learning algorithms. The experiments were conducted on CT images to classify into nodule and non-nodule from the tumor region of interest. The performance of the CAD system was evaluated for the texture characterized images taken from LIDC-IDRI and SPIE-AAPM databases. Due to improved inertia weight used in Particle Swarm (PS) the CAD achieves highest classification accuracy of 98% for solid nodules, 99.5% for part solid nodules and 97.2% for non solid nodules respectively. The experimental results suggest that the developed CAD system has great potential and promise in the automatic diagnosis of tumors of lung.