Page 247 - 18F-FDG PET as biomarker in aggressive lymphoma; technical and clinical validation

Summary, discussion and future perspectives
The project was financially supported by a KWF/Alpe d’Huzes grant (VU 2012- 5848). In this consortium both the clinical individual patient data as well as imaging data from the included international studies were collected, nowadays including about 2300 DLBCL patients [39]. Studies that are currently included in the PETRA database are from Germany [21], United Kingdom [40,41], Switzerland [42], Hungary [43] (together with patients from other countries enrolled in the IAEA study), Italy [44], Nordic/US intergroup study [45] and the Netherlands/Belgium [14,16]. Not only [18F]FDG PET, but also baseline [18F]FDG PET and end-of-treatment [18F]FDG PET data were collected. By building a sustainable database, this PETRA database can also be used to address future research questions, other technical or clinical validations, designing more efficient interim [18F]FDG PET trials and extensions to e.g. other lymphoma subtypes can be made. Furthermore, the PETRA database with qualitatively good studies could also help in the design of future high quality studies. This makes the PETRA database of interest both for clinical researchers as well as for imaging or pharmaceutical companies.
After the original interim [18F]FDG PET project, resulting in this thesis, The PETRA consortium continued with a project focusing on baseline [18F]FDG PET. A consortium grant from KWF was obtained for the investigation of radiomics and this project has started in January 2019. Radiomics analysis of baseline [18F]FDG PET provides quantitative features of tumor characteristics such as intensity, shape, volume, localization and texture and also information about intra- and intertumor heterogeneity. The aim of that project is to identify and validate those features that are robust to variability in image quality and contain prognostic information in addition to classical baseline prognostic factors of the international prognostic index (IPI) [46,47].
One of the features that is included in the radiomics analysis is metabolic tumor volume. The methods used in the pilot study from Chapter 5 were recently also tested in a case-control study with 138 patients from the PETRA database [48]. In that study it was concluded that an automated estimation of metabolic tumor volume is feasible. Both the SUV4.0 and a majority vote strategy (MV2) were recommended to be evaluated further. The association with clinical outcome is currently being explored in a larger database within the PETRA consortium [49]. Furthermore, a project on optimization of operational characteristics of end-of-
245
9