DYNAMIC FDG-PET ENABLES RELIABLE SINGLE-SUBJECT METABOLIC BRAIN NETWORK ANALYSIS

Positron Emission Tomography (PET) using 2-deoxy-2-[¹⁸F]fluoro-D-glucose (FDG) serves as a proxy for neuronal activity and can be used to investigate metabolic brain connectivity. Recently, interest has grown in using dynamic FDG-PET to compute single-subject metabolic connectivity and applying graph theory for diagnostic and staging purposes. However, there is a lack of methodological consensus. To address this gap, we conducted a test-retest study to identify a robust analysis approach. Two methodologically identical 1-hour dynamic FDG-PET scans (20 MBq FDG intravenously) were acquired from 15 healthy Sprague-Dawley rats (18-20 weeks old) within one week. Data were reconstructed into 4D images with 0.4x0.4x0.4 mm³ voxel size using 3 different time framing schemes: 60x 1-min frames, 30x 2-min frames, and a mixed scheme of 12x 30-sec, 19x 1-min, and 7x 5-min frames. Images were smoothed, spatially aligned, and intensity-normalized. Time-activity curves were extracted for 56 volumes of interest (VOIs) of the Schiffer atlas, and Pearson correlation coefficients were calculated between all VOI pairs. Correlations were Fisher z-transformed and used as edge weights. Graphs were generated at densities of 10%-60% of strongest edges. Reliability was quantified using the intraclass correlation coefficient ICC2. Thirty 2-min frames yielded the highest test-retest reliability for both graph edges (Fig. 1a; average ICC2: 0.62), and graph metrics across densities (Fig. 1b). These results demonstrate that dynamic FDG-PET-based metabolic connectivity can be quantified with good reliability, supporting its use for single-subject network analysis. To assess its disease relevance, we will apply this technique in a model of intracerebral hemorrhage.