HD-GLIO 2.0

Delineation of the different tumor compartments is an essential procedure for image data analysis as well as quantitative follow-up of tumor volume for response assessment. Artificial neural networks allow us to automate this process by providing automated tumor segmentations with low effort and improved precision compared to manual segmentation by human raters. In this project, we will develop the 2.0 Version of our previously established tumor segmentation algorithm HD-GLIO, which focused on adult tumor patients and enabled automated tumor response assessment in adult gliomas (https://github.com/CCI-Bonn/HD-GLIO). Specifically, we will re-train the HD-GLIO network including pediatric patients, enabling it to segment a more diverse set of tumor types. Furthermore, we will improve the flexibility of the algorithm by developing a model capable of handling missing sequences in the segmentation process.