Personalized cancer therapies are one of the main promises of modern medicine. The analysis of patient genetics on the one hand and tumor phenotype extracted from medical images on the other hand could provide important information on treatment response or resistancy of a tumor and could help to guide physicians in selecting the most adequate therapy. The rapidly emerging field radiomics aims to describe the tumor phenotype by calculating a large number of image biomarkers (radiomic features) describing the tumor phenotype in a medical image. Radiomics could therefore play an important role in future clinical decision making. The imaging modality PET (Positron emission tomography) is visualizing underlying biological processes and is frequently used for cancer diagnosis, cancer staging, and prognosis. Several studies reported on the promising value of radiomic features extracted from PET images regarding diagnosis or treatment response assessment. However, to date, radiomic features are only used for scientific purposes but are not yet implemented in the clinic. This is due to several challenges coming with radiomics. One of them is the low reproducibilty of radiomic features. I.e. if the same patient would be scanned in two different hospitals, a large number of radiomic features would result in large differences across hospitals. Therefore, in each hospital the physician would draw a different conclusion. By standardizing image acquisition, reconstruction and, pre-processing, a large number of radiomic features can be harmonized so that images from different center become comparable. To standardize the steps in the radiomic workflow and to get one step closer to the clinical application of radiomics was the aim of this thesis.
|Qualification||Doctor of Philosophy|
|Place of Publication||[Groningen]|
|Publication status||Published - 2021|