Tetsuya Tozaki and Michio Senda
medical image processing, FDG-PET, structural information, curvatures
In this paper, we investigate the methods to express the structural information such as tube-like shape or isolated region based on fluoro-deoxy glucose(FDG) PET images and its visualization. To extract the human structural information, we calculate the curvatures of two types. One is calculated by Hessian matrix, the other is based on the four dimensional hyper-surface of FDG-PET images. There are three curvatures through each calculations, and these curvatures express original structures such as the linear shape and isolation degree. Using by these features, we extract the abnormal regions which include doubtful cancer. For the visualization, we propose to construct the newly image, that is the extracted abnormal regions are superimposed on the ordinary image which used for diagnosis of cancer. Moreover, we evaluate each abnormal regions from the view point such as the statistical FDG accumulation or the shape information, and we certify the effectiveness of out methods in the diagnosis of cancer.