Detection of lung tumor in CE CT images by using weighted Support Vector Machines

Lung tumor detection using Contrast Enhanced (CE) Computed Tomography (CT) images plays a key role in computer aided diagnosis and medical practice. Detection of a lung tumor and accurate segmentation is a very challenging task. One major task is to perform classification between a normal (healthy) lung tissue and abnormal (tumor) tissue. However this distribution of data is nonlinear and training a classifier on this kind of data is a difficult process. Limitation of existing approaches is that they assign equal importance to each input feature; this weight assessment is not true for all problems. In this paper we propose a novel method for assigning optimal weights for the calculated features. This proposed technique is tested on CE CT Lung images. Simulation results and analysis showed that our proposed system has shown better classification accuracy than the conventional SVM.