Logarithmic Texture Analysis for Early Lung Cancer Screening on Contrast Enhancement CT Images

This study aims to investigate the effects of contrast enhancement computed tomography (CECT) images on the malignant distinguishes of solid lung nodules. As a representational describer of tumor heterogeneity on images, texture features are used as the classification basis for lung cancers and non-cancers in this preliminary study. 49 patients with malignant bias suspicious lung nodules (all about 3cm) underwent non-contrast and contrast enhancement CT in arterial and venous phases were collected. And the biopsy reports show that there are 28 malignant (or cancer) and 21 benign (or non-cancer) nodules. To enlarge the feature difference between malignant and benign nodules, we proposed an improved logarithmic texture feature analysis scheme. Support vector machine classification and leave-one-out validation were implemented. The areas under the receiver operating characteristic curves of the non-contrast and contrast enhancement CT images in arterial and venous phases are 0.67, 0.73 and 0.77 respectively. Therefore the texture features presented by contrast enhancement CT images show better distinguishing capacity on solid lung nodules than the non-contrast enhancement CT images.