Multicentre study for the harmonization of semi-quantitative analysis of 123I brain SPECT examinations

Nuclear medicine imaging can detect biochemical and physiological processes within the human body. This technique involves the administration of a radiopharmaceutical compound, either intravenously or through inhalation. The compound emits radiation, which allows for the precise localisation of its accumulation within the body. Since the resulting image represents the distribution of the radiopharmaceutical within the human body, it is possible to determine where a specific metabolic process occurs. Using cerebral SPECT (Single Photon Emission Computed Tomography) with the radioisotope 123I, it is possible to determine whether a subject is affected by Parkinson’s disease, even in its early stages. Parkinson’s disease is the second most prevalent neurodegenerative disorder after Alzheimer’s disease, affecting 1% of individuals over 60 years and 4% of those over 85 years. This disorder is characterised by a progressive loss of specific neurons in the substantia nigra, which are responsible for synthesizing dopamine, a neurotransmitter that regulates fine movements. The reduction of these neurons leads to decreased dopamine levels and to a consequent reduction in dopamine receptors number, especially in the basal ganglia, a group of four structures located in the central region of the brain. The radiopharmaceutical 123I Ioflupane specifically localises in dopamine receptors and non-specifically in the cerebral cortex; its concentration within the basal ganglia is proportional to the number of healthy neurons in the substantia nigra. In healthy subjects, the radiopharmaceutical uptake in the basal ganglia is higher than in the cerebral cortex, whereas in subjects with Parkinson’s disease, the uptake is reduced in the basal ganglia either symmetrically or asymmetrically. Radiopharmaceutical biodistribution is assessed through a SPECT examination based on gamma photon emission at 159 keV by 123I. Images obtained must be corrected for photon attenuation and scatter. Raw data from SPECT acquisition can be reconstructed using FBP (Filtered Back Projection) or iteratively, for example, with OSEM (Ordered Subset Expectation Maximisation) algorithm. During reconstruction, filters such as Butterworth or Gaussian filters can also be applied. The images obtained can be analysed using different software. In this thesis, the software used include BasGan, developed by Università degli Studi di Genova, DaTQUANT (GE Heathcare) and Striatal (Siemens). These three software provide the concentration ratio between the basal ganglia (caudate nucleus and putamen) and the cerebral cortex, which is used as the background. They also include a database of healthy subjects to provide age-matched reference values, facilitating disease assessment. To perform a semi-quantitative analysis, two VOIs (volumes of interest) are defined: one in the basal ganglia and one in the cortex. The evaluated parameter is the SBR (Specific Binding Ratio), which is the ratio between the difference in terms of mean pixel value of the two VOIs and the mean pixel value in the background VOI. To compare exams from different hospitals, it is necessary to harmonise image acquisition and reconstruction to ensure reproducibility of the measured SBR. Data from an anthropomorphic phantom, called striatal phantom, were collected from 12 centres in the RIN (Rete IRCCS delle Neuroscienze e della Neuroriabilitazione). Phantom acquisitions were performed for four ratios of 123I concentration in the basal ganglia compared with the background: 10:1, 7.5:1, 5:1, and 2.5:1. For each ratio, a SPECT (or SPECT-CT) acquisition with LEHR (Low Energy High Resolution) collimators was executed. In the first phase of this study (Phase A), the reconstruction was performed by the same hospital that conducted the acquisition, following a standardised procedure. In the second phase (Phase B), raw data collected from the centres were reconstructed centrally at ASST Papa Giovanni XXIII (Bergamo) using Xeleris (GE) and Syngo (Siemens) software for GE and Siemens acquisitions, respectively. The reconstructions were carried out using both FBP and OSEM algorithms. A Butterworth filter was applied to the FBP reconstructions, while a Gaussian filter was used for the OSEM reconstructions. The optimal parameters for the Butterworth filter were determined before starting the harmonisation process. Since the definition of the Butterworth filter differs between Xeleris and Syngo systems, it was necessary to identify parameters for the two filters that would make them similar. Additionally, the optimal number of equivalent iterations (EM) for the iterative OSEM reconstructions was also established. A comparative analysis of images reconstructed using Syngo and Xeleris systems revealed significant differences in the measured SBR values, particularly at higher EM levels. It was observed that the variability in the recovery coefficients from reconstructions of Phase A was greater than that seen in centrally reconstructed images (Phase B). This underscores the importance of a systematic approach to image reconstruction, which can significantly reduce variability and enhance the comparability of images acquired from different centres. The introduction of attenuation correction using Chang’s method was shown to increase the recovery coefficient. However, this correction did not significantly alter the variability of the fitting lines, likely due to the consistent application of corrections in a centralised setting. If attenuation correction were applied independently by different centres, inconsistencies in the region of interest (ROI) placement could introduce additional variability. In comparing FBP and OSEM reconstructions, no significant change in the dispersion of the angular coefficients of the fitting lines was found. However, the differences in OSEM implementations between Syngo and Xeleris systems highlight the need for careful consideration when selecting a reconstruction method. Given the potential for systematic differences, FBP reconstructions are recommended for achieving consistent and comparable results across different imaging systems. Overall, the results of this thesis emphasise the importance of standardised protocols to minimise variability and ensure the reliability of quantitative imaging in SPECT.