The applied dose in the SIRT is typically computed for the left and right liver lobe using the tumor burden (tumor to liver lobe volume ratio) as an essential parameter. Being able to determine the tumor burden precisely and efficiently requires automatic methods since in the case of SIRT patients, manual tumor segmentation is a very time-consuming task.
In the SIRTOP project novel segmentation algorithms are being developed using recent deep learning methods for automatic liver and tumor segmentation in CT and MR images. Moreover, an automatic hepatic artery segmentation using deep learning methods is developed to further optimize the planning process by computing the dose locally per liver supply areas. Since liver tumors are predominantly supplied by arterial blood, knowledge of arterial branches is essential for determining the catheter position during intervention. Thus, providing arterial segmentation helps planning the catheter’s path. To be able to accurately split the liver into left and right lobes or even further into its functional territories, we also develop fully automatic methods for segmentation and subdivision of the portal vein and the hepatic veins.