Wavelet Analysis for Identification of Lung Abnormalities using Artificial Neural Network Amil Ahmad Ilham 1 , Indrabayu 1 , Rezkiana Hasanuddin 1 , Deasy Mutiara Putri 1 1 Informatics Engineering Study Program Electrical Engineering Department Hasanuddin University, Indonesia amil@unhas.ac.id, indrabayu16@gmail.com, rezkianaaa@gmail.com, deasymutiaraputri@gmail.com Abstract—This research analyzed the use of daubechies wavelet as a feature extraction and confusion matrix as the principal parameter of accuracy percentage level in neural network. Detection process began with image pre-processing, lung area segmentation, feature extraction, and training phase. Classifications of the system output consisted of normal lung, pleural effusion, and pulmonary tuberculosis. Seventy five amounts of thorax samples were used as training data and thirty five thoraxes were used as test data. The experiment results showed that the decomposition at level 7 with order db6 was the best configuration for feature extraction which attained up to 91.65% of accuracy. Keywords—thorax, image processing, daubechies wavelet, feature extraction, confusion matrix, artificial neural network. I. INTRODUCTION A lot of computer based methods have been presented to help the work of radiologist. Examination of thorax (chest x- ray) is one of the method that usually done by the hospital for patient in many cases. Chest x-ray shows the image of the heart, lungs, respiratory, blood vessels, spine, and ribs. Lung is an organ located in the thorax that is most subject to diseases. Some of them are tuberculosis and pleural effusion. Chest x-ray are used to diagnose many conditions involving chest wall, thorax and bone structure inside the thoracic cavity including the lungs [1]. Pulmonary tuberculosis is an infectious disease caused by bacillus tuberculosis microbacterium. Pleural effusion is a condition where there is excessive fluid in pleural cavity, which if left unchecked this condition will endanger the sufferer’s life. Pleura is a thin layer of tissue that covered the lungs and lining the inner wall of the chest cavity [2]. In Indonesia, Pulmonary Tuberculosis is the 2nd leading cause of death after heart disease and other blood vessels. Aside from that, Indonesia is the 3rd country in the world that has most Pulmonary Tuberculosis patient after China and India. Pulmonary Tuberculosis could be found in numerous population with low socio-economic conditions and attack the productive age group (15-54 years old) [3]. II. RELATED WORK Reference [4], [5] analyzed the use of image processing to identify an x-ray image. Reference [4] focused on texture analysis and [5] used image enhancement for x-ray images using histogram equalization. Reference [6], [7] used wavelet for iris recognition. In [6], wavelet is used as edge detection technique in Matlab. in [7], used wavelet transform and mahalanobis distance, he compared the influence of wavelet decomposition level and order towards accuracy level. [8] Identify cancer existence in mammography using haar wavelet. Performance validation of neural network is used in [9] to build detection system for lung cancer. III. METHODOLOGY The proposed Lung Abnormalities Detection System can identify three output classifications, normal lung, pleural effusions, and pulmonary tuberculosis. Figure 1. System Flow of Lung Abnormalities Detection System design began with data classification based on diagnosis, normal lung, pleural effusion, and pulmonary tuberculosis. The data was taken in the Radiology Emergency Installation section of Wahidin Sudirohusodo Hospital with the amounts of 90 images. The images consist of 52 normal lung samples, 23 pleural effusion samples, and 15 pulmonary tuberculosis samples. These sample images are subdivided into two, samples for training data and samples for test data. The amount of samples for the training data are 43 normal lung 2014 Makassar International Conference on Electrical Engineering and Informatics (MICEEI) Makassar Golden Hotel, Makassar, South Sulawesi, Indonesia 26-30 November 2014 281  