Classification of Lung Nodules by Artificial Neural Network with Pixel-Based Learning
In medical image processing, classification of medical images using conventional classifiers is challenging as they involve different procedures such as feature selection, feature extraction and segmentation. These processes may cause inaccurate feature calculation and segmentation and cause errors in classification. However, these kind of errors can be avoided by implementing pixel-based learning in image classifiers where feature calculation or segmentation is not required. In this paper, a process of pixel-based machine learning in Artificial Neural Network (ANN) to classify lung nodules on Computed Tomography (CT) image is discussed. In the experiments, we use Waikato Environment for Knowledge Analysis (WEKA) to perform the classification. The training and testing data are collected from the Lung Image Database Consortium (LIDC) and Image Database Resource Initiative (IDRI) as well as The Reference Image Database to Evaluate Therapy Response (RIDER) database is used in the experiment. By using original pixel values in the training of ANN classifier, lung nodules are able to be detected in the output images.