Abstract:
Long axial field of view (LAFOV) PET/CT provides markedly increased sensitivity and an extended axial field of view. The central research question of this thesis was whether these physical advantages could be translated into the enhancement of clinical quantitative imaging. This question was addressed through three interrelated study components: [18F]FDG parametric imaging using an abbreviated dynamic scan and clinical exploration of this approach in non-small cell lung cancer (NSCLC); quantification of respiratory motion correction using patient cohorts and a motion-free reference; and assessment of robustness against defective detector blocks (DDBs). All investigations were performed using the Biograph Vision Quadra LAFOV PET/CT. First, abbreviated dynamic [18F]FDG protocols were optimised for direct Patlak reconstruction using patient-specific image-derived input functions (IDIFs) and scaled population-based input functions (sPBIFs) in high-sensitivity (HS) and ultra-high-sensitivity (UHS) modes. The optimised protocol was applied to treatment-naive patients with histopathologically confirmed NSCLC to analyse standardised uptake value (SUV), net influx rate constant (Ki), and distribution volume (DV) in relation to tumour subtype and programmed death-ligand 1 (PD-L1) status. Respiratory motion correction methods were then evaluated using clinical datasets and a four-dimensional extended cardiac-torso (XCAT) phantom simulated with a validated digital twin of the Biograph Vision Quadra, thereby providing a motion-free reference. Finally, DDBs were simulated in patient and phantom datasets to assess the influence of the number and spatial distribution of DDBs, count statistics, sensitivity mode, and reconstruction protocol on SUVmean, SUVmax, and image noise.
The protocol-optimisation study showed that a 20-min acquisition (45–65 min p.i.), reconstructed with four Patlak frames and patient-specific sPBIFs, provided a reliable balance between clinical feasibility and quantitative fidelity under the investigated conditions. This protocol preserved Ki estimates relative to patient-specific IDIF-based references and produced higher tumour-to-background contrast than conventional SUV images. Shorter acquisition windows increased deviations and reduced image fidelity. In the NSCLC cohort, lung squamous cell carcinoma lesions showed higher SUVmax and Kimax than lung adenocarcinoma lesions, whereas DVmax did not differ significantly between histological subtypes. PD-L1-positive lesions showed higher Kimax and DVmax than PD-L1-negative lesions, whereas SUVmax showed only a non-significant trend. Thus, Ki and DV should be reported as complementary metrics alongside SUV.
The motion-correction study demonstrated that quantitative accuracy also requires spatial fidelity. Respiratory motion caused lesion blurring, uptake underestimation, and lesion volume overestimation, particularly near the diaphragm. Two correction methods were evaluated, and both improved visual lesion delineation. However, comparison with the motion-free reference showed that greater sharpening and larger SUV recovery did not necessarily correspond to more accurate recovery. The data-driven method produced greater uptake recovery and a more compact lesion appearance but tended to overcorrect selected pulmonary lesions. The image-based method was more conservative relative to the motion-free reference, although it introduced greater image noise. Therefore, motion correction should be evaluated according to lesion location and the quantitative metric being assessed.
The detector-block defect analysis showed that DDBs induced quantitative bias, with the magnitude of bias depending on the number, location, and distribution of DDBs, as well as the count statistics of the examination. The Biograph Vision Quadra maintained quantitative reliability with up to eight sparse or four clustered DDBs under the investigated conditions. EARL2-compliant reconstruction in UHS mode partially mitigated deviations but could not fully compensate for severe detector-block defects.
In conclusion, this thesis shows that the technical advantages of LAFOV PET can be translated into accurate and reliable quantitative imaging when combined with appropriate acquisition protocols, motion-correction strategies, and quality-control procedures.
