ANN Based Classification of Lung Cancer and TB Images using Texture Features

Lung cancer is one of the most common and lethal types of cancer. As occurs in almost all types of cancer, its cure depends in a critical way on it being detected in the initial stages, when the tumor is still small and localized. Thus an objectively standardized criteria is required for clinically and histologically identification of the individuals suffering from lung cancer. Regular follow-up and professional screening is needed to intercept lung cancer during its curable stage. Death rates due to lung cancer are doubling every decade, no cure disseminated disease is available, so diagnosing and educating individuals at increased risk at an early stage is vital to increase survival rates. With this view, the present study will help the radiologist as well as physicians to identify quickly the deadliest form of lung cancer in the early stage. The texture feature estimation algorithms are applied to various lung cancer chest X-ray images such as small-cell (SC)and non-small-cell (NSC) type, as well as on tuberculosis (TB) images (25 images from each category). Initially, the identifying features are obtained from the X-ray images using image processing and analyzing methods. Then, these features are applied to an expert system to classify the lung cancers into malignant (SC, NSC) and benign (TB).