Optimized Lung Nodule Prediction Model for Lung Cancer Using Contour Features Extraction

Abstract

Lung cancer is one of the most complicated diseases in human assessment. It affects the lives of humans critically. For the treatment of the lungs and prevention of the disease, it is important for an accurate and timely diagnosis of the disease. In most aspects, it was not as successful as the traditional method, which uses past medical history. Classification techniques are effective and reliable in categorizing affected lungs and people with normal lungs. But unfortunately failed to provide proper nodule location as this became a tedious task to find it. In the proposed method, Hybrid Particle Swarm Optimization-Lung Nodule Candidate (HPSO-LNDC) Detection depends on the diagnostic system using disease image data set for lung segmentation assessment. Initially, the input Data images are pre-processed to reduce the computational complexity of the system. Then, those segmented images are subject to the proposed HPSO-LNDC model. To solve the problem of accuracy, so many researchers have used slicing techniques, but most of these techniques are stuck in the local minima and suffer from premature convergence. Therefore the HPSO-LNDC model is integrated with the Hybrid PSO to reduce loss function occurring in the lung nodule detection architecture. Such a combination helps the HPSO-LNDC to avoid low similarity values. The proposed algorithm is simulated using the MATLAB tool and tested experimentally. It is defined as accuracy, detection level and Dice Similarity Coefficient (DSC). Results show that HPSO-LNDC with 85% accuracy and DSC of 90% was better than conventional methods.