Interactive method to extract tumor regions from chest X-ray CT images and its application to the benign/malignant discrimination

This paper proposes a new interactive method of extracting lung tumor regions of precisely peripheral shape, using chest X-ray CT images. Although methods have been proposed for extracting the rough shape of the tumor, the general aim of these methods has been to use for screening. Rough shapes of tumors as determined using these methods are not sufficient for use in benign/malignant discrimination, because the precise shapes of the tumors themselves are used as the regions to calculate features in the discrimination. The method proposed in this paper was applied in 78 cases of thin-slice chest X-ray CT images. Results confirmed that appropriate tumor regions were extracted in 67 cases. Furthermore, an increased correct discrimination rate was achieved in benign/malignant discrimination using tumor regions obtained by the proposed method. (author)