Identification of Lung Cancer Using a Back Propagation Neural Network

Abstract

Reading image of lung cancer screening well-known as X-ray by practitioners are sometimes subjective. This research tried to create software that can detect lung cancer as a comparison of the work of medical practitioners using artificial neural networks (ANN), with X-ray movies taken from the tool diagnostic radiography (DR) stored in the compact disc. The dependent variable observation in this study is the identification of DR X-ray image size of 1024 x 1024 pixels. A total of 10 images X-ray which has been observed by the physician radiology. With 5 images X-ray normal and 5 images lung cancer. In this study, the image processing is done through three stages: neighborhood averaging, median filter and histogram equalization. The result of these features are grouped in normal categories. From test results stating the truth 80%. To facilitate the user in the lung disease pattern recognition. GUI applications design using MATLAB. We use some form of image processing which includes form training andtesting. The best parameters obtained from this research include learning rate=0.3, the number of hidden layer=30 and tolerance error=10-8. From the results obtained by the level of accuracy of the training image of normal lung, lung cancer in a row is 80%. Overall the level of accuracy of the results is 80%.