Smartbrush

Tumors were manually segmented in 3D by trained raters using an initiation by either a region growing algorithm [26 (link)] (Brainlab SmartBrush, BrainLAB AG, Münich, Germany) or a grow cut algorithm [49 ] (3D Slicer, http://www.slicer.org, accessed on 3 June 2021) and subsequent manual editing. Trained raters were supervised by neuroradiologists and neurosurgeons. The tumor was defined as gadolinium-enhancing tissue on T1-weighted scans, including nonenhancing enclosed necrosis or cysts.

As part of the clinical routine, all patients received MRI scans for radiation treatment planning (median time between PET and MRI, 12 d). In all cases of tumor resection, only postoperative images were used. Axial T1-weighted sequences before and after intravenous injection of 0.1 mmol/kg gadobenate dimeglumine contrast agent (MultiHance; BraccoImaging) were analyzed to measure the total contrast-enhancing tumor, and T2 or fluid-attenuated inversion recovery (FLAIR) sequences were used to measure non–contrast-enhancing tumors. Tumor volumes were manually delineated as defined by the Advisory Committee for Radiation Oncology Practice of the European Society for Radiotherapy and Oncology guidelines using the institutional imaging software (BrainLab Smartbrush; BrainLab) (9 (link)). In cases of multifocal disease, each focus was quantified separately and summed together. In patients undergoing microsurgical tumor resection before radiotherapy, we ensured that postoperative T2 or FLAIR abnormalities were not surgically induced edema or ischemia by reviewing diffusion-weighted imaging sequences.

According to our standard in-house protocol, the magnetic resonance imaging (MRI) (1.5- or 3.0-T scanners: Magnetom Symphony, Siemens, Erlangen; Signa HDxt; GE Healthcare, Little Chalfont, United Kingdom) routinely included axial T2-weighted sequence (with slice thickness of 2 mm), 3-dimensional T1-weighted sequences before and after intravenous administration of gadopentetate dimeglumine (0.1 mmol/kg body weight; Magnevist; Schering Corporation, Kenilworth, NJ), and constructive interference in steady-state sequences (CISS, with slice thickness of 1 mm), with axial, sagittal and coronal reconstructions each. Volumetric tumor analyses of pre- and post-operative MR images were performed by semi-manual segmentation of pre- and post-operative T2/CISS and contrast-enhanced (CE) T1 images and using a commercially available software tool (SmartBrush^®, Elements^®, BRAINLAB AG, Munich, Germany). For study inclusion, only craniopharyngiomas with at least 70% cystic tumor volume were included. This results in a relevant solid tumor fraction of about 10-30% of the tumor volume.