Page 88 - Quantitative Imaging of Small Tumours with Positron Emission Tomography

PET imaging protocol Patients were scanned on two consecutive days on a whole body time-of-flight Gemini TF64 PET-CT scanner (Philips Healthcare, Netherlands) with EARL accreditation (16). A 4h fasting period was included to minimize intra-intestinal bile activity. Intravenous injection of ±200 MBq [18F]FDHT was followed by a 30 min dynamic scan of the chest (with aorta in field of view) to acquire an image derived input function. Venous blood samples were drawn at 5, 10 and 30 min. Analysis of venous samples included measurements of whole blood and plasma activity concentrations, parent fraction, and metabolites (details in (19)). A whole body scan (3min/bed position) was made from mid-thigh to skull vertex at 45 4 min post-injection. Complying with the EARL1 guideline (16), whole body PET images were reconstructed with standard iterative time-of-flight reconstruction algorithm (BLOB-OS-TF) with 3 iterations and 33 subsets, with a matrix size of 144x144 and voxels 4x4x4mm. Images were corrected for scatter, random coincidences, decay, and attenuation (low-dose CT; 80 mA at 120-140 kV, 5 mm slice thickness). We additionally reconstructed images with the PSF algorithm as provided by the vendor (Philips Healthcare) to conform with EARL2 guidelines. This comprises post-reconstruction image processing using the Richardson-Lucy iterative deconvolution algorithm with sieve noise regularization (PSF option: 1 iteration, regularization full-width-at-half-max at 6 mm) as resolution recovery method (20). This algorithm uses a scanner-specific spatially variant PSF to improve image resolution, and is described as follows (20): Eq. 1 where Ii+1 is the current image estimate; Ii is the image estimate from the ith iteration; f is the system Gaussian PSF; s is the sieve kernel; and Io is the original measured image. To evaluate the impact of count statistics (i.e., acquisition time), we split the original list-mode data of each whole body PET scan on an alternating count-wise basis into two new datasets, which were subsequently reconstructed into whole body images (as proposed in (21,22)) using both EARL1 (4 mm) and EARL2 (4 mm+PSF) reconstructions. This generated two statistically equivalent but count- independent PET-images each containing 50% of the original counts (referred to as split 1 and split 2). Due to the linear relationship between (decay corrected) [18F]FDHT PET repeatability 87