Kidney is a fist-sized organ which performs complex and vital functions that keep the body in balance (Figure 1). Kidney tumors are often incidentally found during non-contrast CT (NCCT) scans. Segmentation of kidney tumors on NCCT images is challenging since they have similar intensity values compared to the normal tissues (Figure 2). In this paper , the authors proposed a framework that explicitly captures protuberances to enable a better segmentation for isodensity tumors (Figure 3). The proposed framework consists of three networks. First, Base Network predicts initial kidney and tumor region masks. Second, Protuberance Detection Network, which is trained using a synthetic dataset, predicts protruded regions from the kidney mask. Finally, Fusion Network fuses the outputs from the base network and protuberance detection network. The author conducted experiments using a publicly available dataset and showed that the proposed method can accurately segment tumors, even with only the slightest clue of protruded areas (Table 1). In the future, it is expected that the proposed method will be extended to other organs (e.g., adrenal gland, liver, pancreas), allowing for more accurate tumor detection.
 Hatsutani, T., Ichinose, A., Nakamura, K., Kitamura, Y. (2023). Segmentation of Kidney Tumors on Non-Contrast CT Images Using Protuberance Detection Network. In: Greenspan, H., et al. Medical Image Computing and Computer Assisted Intervention – MICCAI 2023. MICCAI 2023. Lecture Notes in Computer Science, vol 14226. Springer, Cham.
CAUTION：This is Fujifilm Global Website. Fujifilm makes no representation that products on this website are commercially available in all countries. Approved uses of products vary by country and region. Specifications and appearance of products are subject to change without notice.