Fuzzy information granulation towards benign and malignant lung nodules classification

Abstract

Lung cancer is the second common cancer with the highest death rate in the world. Cancer diagnosis in the early stages is a critical factor for increasing the treatment speed. This paper proposes a new machine learning method based on a fuzzy approach to detect benign and malignant lung nodules to early-diagnose lung cancer by investigating the computed tomography (CT) images. First, the lung nodule images are pre-processed via the Gabor wavelet transform. Then, some of the texture features are extracted from the transformed domain based on the statistical characteristics and histogram of the local patterns of images. Finally, based on the fuzzy information granulation (FIG) method, which is widely recognized as being able to distinguish between similar textures, a FIG-based classifier is introduced to classify the benign and malignant lung nodules. The clinical data set used for this research are a combination of 150 CT scans of LIDC and SPIE-APPM data sets. Also the LIDC data set is analyzed alone. The results show that the proposed method can be an innovative alternative to classify the benign and malignant nodules in the CT images.