Texture Feature Extraction to Improve Accuracy of Malignant and Benign Cancer Detection on CT-Scan Images

Lung cancer is a type of lung disease characterized by uncontrolled cell growth in lung tissue. Whereas nodules (benign cancer) are small, round or egg-shaped lesions in the lungs. The current method used to diagnose lung cancer from CT scan images is by observing a data set of 2-D CT Scan images using naked eye, then interpreting data one by one. This procedure is certainly not effective. Research conducted aims to extract texture features to improve accuracy of malignant and benign cancers detection in CT scans. This research covers 5 (five) main points. First is pre-processing CT-Scan images. The second is automatic segmentation of lung area using Active Appearance Model (AAM) method. Third is segmentation of candidates who are considered cancer using morphological mathematics. Fourth, process of detecting benign and malignant lung cancer using Support Vector Machine (SVM). Fifth is visualization of malignant and benign lung cancer using Volume Rendering. Accuracy of malignant and benign cancers detection is 79.7%