A Computer Aided Diagnosis System for Lung Cancer based on Statistical and Machine Learning Techniques

lung Cancer is believed to be among the primary factors for death across the world. Within this paper, statistical and machine learning techniques are employed to build a computer aided diagnosis system for the purpose of classifying lung cancer. The system includes preprocessing phase, feature extraction phase, feature selection phase and classification phase. For feature extraction, wavelet transform is used and for feature selection, two-step statistical techniques are applied. Clustering-K-nearestneighbor classifier is employed for classification. The Japanese Society of Radiological Technology’s standard dataset of lung cancer has been utilized to evaluate the system. The dataset has 154 nodule regions (abnormal) where 100 are malignant and 54 are benign and 92 nonnodule regions (normal). An Accuracy of 99.15% and 98.70 % for classification have been achieved for normal versus abnormal and benign versus malignant respectively, this substantiate the capabilities of the approach presented in this paper. Indexed Terms — Computer Aided Diagnosis, Lung Cancer, Statistical Feature Selection, Cluster k-Nearest Neighbor term