Ultra-low-dose total-body ¹⁸F-FDG PET/CT in patients with autoimmune inflammatory arthritis: evaluation of image quality with shorter scan time

Purpose: Total-body PET/CT scanners offer the unique ability to evaluate the systemic inflammatory burden of autoimmune inflammatory arthritic (AIA) conditions using ultra-low dose protocols. However, AIA patients are unable to tolerate long scan times. We prospectively evaluated the influence of shortening the scan time on image quality in participants with AIA and non-AIA controls using a total-body PET/CT scanner.

Methods: Participants with an established AIA condition (Rheumatoid Arthritis, n=3; Psoriatic Arthritis, n=8) and non-AIA controls (Osteoarthritis, n=3) were prospectively enrolled. All participants underwent a total-body PET/CT scan (uEXPLORER, United Imaging Healthcare) for 20 minutes starting at 40 minutes after the intravenous injection of 75.5±4.4 MBq of ¹⁸F-FDG at a single-timepoint. List-mode data were acquired. Data were reconstructed as a single frame (20-min long), four 5-min frames, and twenty 1-min frames using the ordered subset expectation maximization method (OSEM) with 20 subsets and 4 iterations, into a 150x150 matrix producing 4 mm isotropic voxels. All reconstructions were performed with and without point spread function (PSF) modeling. No post-reconstruction smoothing filters were employed. Volumes of interest (VOIs) of 3.7 ml and 6.6 ml were manually drawn on the ascending aorta blood pool (BP) and liver. The coefficient of variation (COV) from these VOIs was calculated. Another VOI (size adjusted according to the anatomical site) was placed on the most active arthritic lesion to measure the lesion SUV_max and calculate lesion-to-background ratio (LBR) in relation to BP SUV_mean. LBR values from the 1-min frames were used to quantify the temporal changes in lesion ¹⁸F-FDG uptake over the 20-min of acquisition. The ROIs were manually manipulated between frames if motion was visualized on the images.

Results: The 14 study participants were all males (age: 59±13 yrs). The average COV (%) for the 20-min, 5-min, and 1-min scans without PSF modeling were, respectively: 8±2, 15±4, 32±8 for the BP and 8±2, 15±3, 31±7 for the liver. The COV (%) from the corresponding images with PSF modeling were marginally improved across both BP and the liver (by an absolute average of 2%). The average LBR, for the 20-, 5- and 1-min frames, were, respectively:1.60±0.49, 1.78±0.61, 1.95±0.69 without PSF and 1.73±0.53, 1.95±0.69, 2.10±0.75 with PSF. The LBR increased gradually over the 20-min acquisition and was slightly higher for AIA (range: 1.81-2.35) compared to non-AIA (range: 1.18-1.62). On an average, LBR increased 47% in non-AIA lesions compared to 39% for AIA lesions.

Conclusions: Shortening of scan time is crucial in the arthritic population to enhance patient comfort and reduce intrascan movement. These initial results indicate that scan times as short as 5-min, with ~75.5 MBq ¹⁸F-FDG injected dose, appeared to provide a COV below 15%. Analysis of the temporal ¹⁸F-FDG uptake characteristics in lesions of the different arthritides may provide insights regarding the underlying pathological processes. However, the COV is quite high for 1-min frames. At 4 iterations and 20 subsets, PSF modeling increased LBR by ~8% and marginally improved COV. The effect of different reconstruction schemes and the potential for deep learning-based noise reduction warrant further investigation.

• Download figure
• Open in new tab
• Download powerpoint