Page 210 - Quantitative Imaging of Small Tumours with Positron Emission Tomography

for example pre-treatment (baseline) reads, are validated in pharmacokinetic analysis for prognostic or predictive purposes, PVC may have a substantial impact on validation of the preferred simplified metric. Also, in situations where a variable treatment effect on tumour size is expected (i.e. pseudo-progression during immunotherapy), the impact of PVC could be larger. In dynamic PET acquisitions in brain studies, PVC will be of benefit as the image-derived input function derived from carotid arteries is subject to PVE, but this is not an issue in oncological PET (13). Precision must be balanced against accuracy One of the main focuses of quantitative PET research has been its potential use in response assessment to systemic treatment (14). Use of PET for this purpose is attractive since it may provide for an assessment of the cancer burden as a whole, and for assessment of intrapatient heterogeneity of response. Also, the general hypothesis is that functional changes (e.g. reduction in tumour metabolism or proliferation) precede the anatomical tumour changes that are commonly measured on CT (e.g. using RECIST1.1), allowing for earlier positioning of PET in the clinical follow-up (15,16). As previously stated, knowledge of the intrinsic variability of PET quantification is necessary to be able to discern actual treatment-induced changes versus day-to-day variations (17-21). When aiming to improve the accuracy of small tumour quantification with PVC, or when using novel PSF image reconstructions, a potential negative effect on precision must be acknowledged and investigated as this may negatively affect the response assessment abilities of quantitative PET. In Chapters 2, 4, and 5, we thoroughly investigated how PVC affects the test-retest variability of quantitative PET metrics in patients with NSCLC ([18F] FDG) and metastatic prostate cancer ([18F]FCH, [18F]FDHT, and [18F]DCFPyL). First, we demonstrated that the PVC methods that are prone to yield inaccurate results (i.e. those depending on tumour delineation accuracy) also had a substantial negative impact on the test-retest variability (1). In contrast, methods that performed well in terms of accuracy (e.g. iterative deconvolution) had only a marginal effect on the repeatability as expressed in the intra-class correlation coefficient (ICCs >0.9), and was thus chosen for clinical evaluation in Chapters 4 and 5. [18F]FDHT and [18F]DCFPyL (a PSMA-ligand) both are prostate cancer- targeted radiotracers that may provide for PET-derived biomarkers in metastatic 11 Summarizing discussion 209