- Research groups and areas
- Clinical Research
- Research Ethics Committees
- Scientific annual report
- Calls
- Funded research projects
- Internationalisation
Multimodality Artificial intelligence open-source Tools for Radiation Treatment Optimization in patients with Glioblastoma
- Call
- Joint Transnational Call for Proposals 2022 (JTC 2022)
- Principal investigator
- Luis Martí-Bonmatí
- Role
- Participant
- Year
- 2023
High-grade gliomas (HGG) are very aggressive brain tumors with poor overall survival rates of 16 months on average and 70-80% local recurrence (LR). Treatment concepts have remained almost unchanged for decades. The correlation of tumor imaging hallmarks with treatment response could allow personalized adapted therapies to prevent disease progression. However, the best imaging modality for HGG remains an unanswered question. HGG diagnosis and therapy has been commonly based on magnetic resonance (MR). Positron emission tomography (PET) has been proposed to overcome the MR limitations (50% specificity), when differentiating LR from radiogenic alterations. In this context, the use of hybrid PET/MR systems for simultaneously imaging HGG permits a more accurate comparison of the information provided by both modalities. Our main motivation is the identification of biologically active tumor tissue associated with LR in HGG, in order to replace the homogeneous dose distribution conventionally delivered in radiotherapy treatment, by a dose distribution scaled based on the patient´s specific risk profile of LR. With this purpose, our project aims to identify the best modality for HGG segmentation and for LR prediction. As result of the project, an open-source software is expected, which combines PETand MR- based HGG segmentation and LR prediction to support clinical decisions regarding personalized treatment options. In our proposal, two patient cohorts are involved: 232 patients of the existing prospective GLIAA trial (Freiburg) and small (25) exploratory study part of the current project (Málaga and Valencia). Artificial intelligence (AI) is applied to develop PET- (Freiburg) and MR- (Valencia) based tumor segmentation and radiomics models for LR prediction. Integrated PET/MR models are also evaluated. The expertise of the academia partner (Vienna) focuses on the improvement of the robustness of AI-algorithms and the development of an open-source visualization tool.