Lung Cancer Detection Using Hybrid Methods of Otsu-Based PSO Algorithm Combined with ACO Algorithm

Early identification of lung cancer could lead to a reduction in mortality. Due to the difficult process in the analysis of imaging tests, alternatives such as computer systems that use image processing and design authentication techniques have been widely developed to diagnose the disease for early diagnosis and to give experts a second opinion to develop this process possibly fast. Therefore, this work proposes a method for detecting lung nodules from areas extracted from tomography computed by CNN. First, the nodes are divided into two subdivisions using the Otsu method based on the PSO algorithm. Then, the pieces of nodes and their subdivisions were resized to 28 × 28 dimensions and inserted simultaneously into the networks. The model architecture consisted of three CNNs that eventually shared the same FCL at the end. The ACO method was utilized to optimize some parameters, such as the amount of filters and neurons, because it is a parameter model. The method was tested using the LIDC, resulting in 94.67% sensitivity, 95.16% accuracy, 94.79% accuracy, and an area under the 0.950 ROC curve.