Footnote:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Description:
Glioblastoma is the most common malignant tumor of the central nervous system. In particular, the highly invasive nature and irregular growth of glioblastoma contribute to a median patient survival of only 15 months despite a multidisciplinary therapeutic approach. The aim of this work is to quantitatively evaluate the clinical applicability of tumor spread prediction using mathematical modelling. To this end, tumor spread simulations made using the modelling approach of Engwer et al. are compared to tumor structures made by experienced radiation oncologists. In order to be able to perform the simulations as well as the evaluations, a comprehensive workflow will be established which allows a reproducible and largely automated preparation of the basic data. Initially, an automatic tumor segmentation will be made on 4 MRI sequences which will serve as a starting value for the growth simulation. Following the multi-scale modelling approach according to Engwer et al., the future tumor spread is simulated including DTI data. Prior to the final tumor simulation, the influences of the individual simulation parameters on the simulation result are investigated. Parameters that are already defined in the literature or having only a minor impact on the simulation are assumed to be constant, while influential parameters are varied in the optimization process. For the evaluation of the simulation results, reference tumor structures are created on regular follow-up MRIs of the patient by experienced radiotherapists and then compared volumetrically and geometrically with the simulated tumor. Using three patient datasets, agreements between simulation and reference of 0.62, 0.65, and 0.66, expressed by the Dice index can be achieved. If the simulated tumor is additionally compared with tumor structures created by three other radiotherapists, the results improve to 0.62, 0.68 and 0.7. If one also takes into account that the variability of the tumor contouring within a group of four experienced radiotherapists lies at an average ...