Detection of lung tumor using SVM and Bayesian classification

Abstract

Lungs being an important organ in the respiratory system, it is prone to many chronic diseases involving tumor cells. These lung tumors are treatable, if diagnosed at early stage. Among lung tumors, the non-small cell category is irresponsive even for chemotherapy treatment when diagnosed at later stage. This work concentrates on improving the diagnosis of non-small tumor cells at early stage through image processing techniques. The CT image of lungs is used for discriminating the tumor cells from healthy non-tumor cells. Upon using computer aided image processing techniques, the level of accuracy in assessing the tumor cells can be improved. Initially, the noise present in the CT image is removed using Wiener filter by improving the signal to noise ratio. The vascular structures in the image are removed and possible tumor cells are segmented from other healthy cells using region growing technique. After extracting the features, the Support Vector Machine and Naïve Bayesian techniques are used for classifying the tumor cells and healthy cells.