Recognition of boundary voxels of tumors in chest CT images for extraction of tumor regions

In this paper, the properties of boundaries of lung tumors are considered in terms of extraction of tumor regions from chest X-ray CT images. We calculated features on boundary voxels those were determined by a medical doctor. We also calculated them in the adjacent region of boundary to investigate the difference between those regions. Lung tumors can be classified into four types in respect of CT values around their boundaries. They are solid type, the non-GGA, the mixed-GGA and the pure-GGA. It is thought that medical doctors determine boundaries on each criterion for each type of tumors. The voxels were classified into the boundary region and the adjacent region for each type using SVM [1].