Modified Locust Swarm optimizer for oral cancer diagnosis

Abstract

Early diagnosis of Oral Cancer is too significant to prevent death from fatal oral cancer. Because of the high complexity of the oral cancer diagnosis by the physicians at the early stages, in this study, a new intelligent method for this purpose is an auxiliary tool to reduce human medical errors. The proposed technique includes different parts including image segmentation based on the Reinforcement Learning method, image feature extraction, by Gabor wavelet transform, and the final classification based on an RBF-kernel-based SVM. For decreasing the system complexity, the optimum features were selected based on a modified metaheuristic, called Modified Locust Swarm Optimization (MLSO) algorithm. This algorithm is also used in the classification step to provide optimal configuration for SVM based on its kernel. For validation of the efficiency of the suggested method, it is carried out to the “Oral Cancer images” dataset. A comparison of final results with several other latest techniques to indicate the higher efficiency of the system. Simulation results show that the proposed method with 96.94% provided the minimum error ratio against the other comparative methods. Also, the results indicate that the proposed method with 93.89% sensitivity, 92.37% specificity, 92.37% PPV, and 96.94% NPV presents the best efficiency among the others.