2015 年 53 巻 Supplement 号 p. S418-S419
As evaluation of disease progression or prediction of prognosis of patients with polycystic kidney disease (PKD), measurement of total kidney volume (TKV) has widely been accepted clinically. To establish automatic TKV measuring method with accuracy in PKD, we developed a segmentation algorithm of renal region. The coronal section of MRI T1 weighted images were used for the present study to capture the irregular boundary of polycystic kidney which often contacts with liver due to nephromegaly. Since biphasic luminance value of cysts and renal parenchyma was distributed, the GraphCut algorithm was utilized to discriminate renal region from the other organs. Then, the likelihood of pixels in foreground and background were calculated using Gaussian Mixture Model (GMM), and the GraphCut was performed. Segmentation results were evaluated by precision and recall, and compared with the method of the previous studies applying the region growing method and the level set method. The proposed procedure indicated high recall and precision compared to the previous studies.