Statistical Analysis of Features and Classifiers in Identifying Nodules and Its T Staging in Lung Ct Images

Lung cancer is the most common disease with greater morality rate. Computed Tomography (CT) images are used for early diagnosis of lung cancer with the help of CAD system. Selection of effective feature set and proper classifier for medical images where machine learning techniques are used is a challenging task. Texture analysis of computed tomography (CT) images is one of the important preliminary stages in the detection and classification for lung cancer. The image texture is characterized by Haralick texture with variety of statistical measures. The extracted texture feature values are used by a CAD to differentiate its type as benign or malignant. This paper aims to compare experimental results of 18 features extracted by using Gray Level Co occurrence Matrix (GLCM) and analyses the different classifiers that can be used for classification of nodule as benign or malignant. Measuring the statistical parameters of the nodule is crucial for determining the T stage of the nodule. This paper also analysis the statistical parameters and reported the contribution of minimal feature set for classification and staging. GLCM features are used for classification and Geometric features used are used for T staging. For these analysis 23 images dataset of different types of cancer is used.