Intelligent Computing Approach for Lung Nodule Detection

This chapter focuses on automating the diagnosis of lung cancer by identifying and classifying lung nodules in lung CT images. This approach involves formulating a weighted dual goal optimization problem that selects an optimal feature subset for accurate classification of lung nodules. In order to guarantee that a compact feature subset enhances nodule classification accuracy, the weights allocated to the discriminating feature subset and classification accuracy are strategically determined. Here, an adaptive weighted aggregation strategy based on the Group Improvised Harmony Search (GrIHS) evolutionary optimization technique has been proposed to tackle the optimization problem. To further improve nodule classification accuracy, an adaptive k-NN classifier has been integrated into the GrIHS technique to determine the optimal number of neighborhoods in the search space. The proposed methodology has been successfully applied to the Lung Image Database Consortium (LIDC) data set. The experimental findings demonstrate the efficacy of the proposed method, with only 12 differentiating features required to achieve a remarkable sensitivity of 97.59% and blind testing accuracy of 97.78%. Consequently, the proposed methodology has the potential to assist radiologists in diagnosing lung cancer by providing reliable support in the interpretation of lung computed tomography images.