Novel computer‐aided lung cancer detection based on convolutional neural network‐based and feature‐based classifiers using metaheuristics

This study proposes a lung cancer diagnosis system based on computed tomography (CT) scan images for the detection of the disease. The proposed method uses a sequential approach to achieve this goal. Consequently, two well-organized classifiers, the convolutional neural network (CNN) and feature-based methodology, have been used. In the first step, the CNN classifier is optimized using a newly designed optimization method called the improved Harris hawk optimizer. This method is applied to the dataset, and the classification is commenced. If the disease cannot be detected via this method, the results are conveyed to the second classifier, that is, the feature-based method. This classifier, including Haralick and LBP features, is subsequently applied to the received dataset from the CNN classifier. Finally, if the feature-based method also does not detect cancer, the case study is healthy; otherwise, the case study is cancerous.

computer-aided design, convolutional neural network, Haralick texture features, improved Harris Hawks optimizer, independent component analysis, lung cancer diagnosis