(1) Patient Study 1
Of the 3,633 whole-body 18F-FDG PET/CT cases at our institute, we identified 61 patients with urinary catheters. Table 1 summarizes the demographics of these patients. Of the 61 patients, five (8.2%) had halo artifacts caused by the urinary catheter in TFS 4-mm PET images.
Figures 2 and 3 illustrate two representative cases. The sinogram images of μ-maps, a mask for TFS 4-mm, and emission (prompt-minus-random) are shown in Figures 4A–E and 5A–E. In both cases, the urinary catheters were absent on the μ-map sinograms due to the masking process for TFS 4-mm (yellow arrows in Figs. 4B,C and 5B,C). Figure 6 shows the sinogram profiles corresponding to the horizontal-axis direction in Figures 4 and 5. The scatter was overestimated (red lines in Fig. 6A,B), and thus a halo artifact was identified on the TFS 4-mm PET images, but no halo artifacts were present on the NSC images (Figs. 2C,3C).
Our visual observation of the μ-maps of the two cases revealed that urinary catheters were positioned inside the masked regions and were absent on the 4-mm voxel μ-maps (Figs. 4B,C and 5B,C). The urine in the catheter was not visible in the CT image (yellow arrow in Fig. 2A), and not present on the μ-maps (Figs. 2B and 4B). However, the PET images showed the urine (red arrows in Fig. 2C). This suggests that there is a mismatch between the μ-maps and the PET images, and this mismatch may be ascribed to the urine in the urinary catheter having shifted in the interval between the CT imaging and the PET imaging (urine shift pattern).
Figure 3 shows the presence of a urinary catheter and urine in both the CT and PET images (red arrows in Fig. 3A,C), but the urinary catheter and urine are absent in the 4-mm voxel μ-map (Figs. 3B and 5B), leading to the mismatch between the μ-maps and the PET images. This mismatch occurred for a patient's case with the urinary catheter placed in an arc (blue arrow in Fig. 3C). Four cases including the one illustrated in Figure 3 showed a similar phenomenon, i.e., the tube curve pattern.
(2) Phantom Study
We developed an experimental approach to reproduce the phenomenon that occurred in Patient Study 1.
Figure 7 provides phantom images that simulate the urine shift pattern. The region with the mismatch between the CT images (μ-maps) and the PET images identified in Patient Study 1 corresponds to the second urinary catheter region from the left in Figure 1A. No 18F-FDG solution was added to the urinary catheter that was showing in the CT image, and this urinary catheter was not visualized in the μ-maps. However, the PET images showed the 18F-FDG solution, suggesting that there is a mismatch between the μ-maps and the PET images, as in Patient Study 1.
The TFS 4-mm in the urine shift pattern overestimated the scatter (red line in Fig. 6C), resulting in the presence of halo artifacts. The SUV was <0.5 and was underestimated (slices 30–100 in Fig. 8A). Halo artifacts also appeared in the TFS 2-mm PET image. The urinary catheter was not shown in the 2-mm voxel μ-map, causing the mismatch and the scatter to be overestimated (green line in Fig. 6C). By contrast, in the MCS 4-mm PET image, the scatter was correctly estimated (blue line in Fig. 6C) and no artifact appeared, resulting in a correct evaluation of the SUV (Fig. 8A). No artifact appeared in other regions (CT–/PET–, CT+/PET–, CT+/PET+) with the scatter accurately estimated, and the SUV at these regions was 1.0 (Fig. 8A, Suppl. Fig. S1).
Figure 9 provides phantom images that simulate the tube curve pattern. The region of the arc-shaped urinary catheter appears in the CT image but not on the 4-mm voxel μ-map. This resulted in mismatch with the PET image. The TFS 4-mm overestimated the scatter (red line in Fig. 6D), resulting in the presence of halo artifacts. The SUV was close to zero and was underestimated when the tube angle was large (slices 110–180 in Fig. 8B). By contrast, no halo artifact appeared in the TFS 2-mm PET image. Here the urinary catheter was clearly depicted in the 2-mm voxel μ-map, and the scatter was correctly evaluated (green line in Fig. 6D). In the MCS 4-mm PET image, the scatter was also correctly evaluated (blue line in Fig. 6D), and no halo artifact appeared. Since there was no halo artifact, the SUVs for both TFS 2-mm and MCS 4-mm PET images were 1.00 ± 0.05 (mean ± standard deviation) in all catheter regions (Fig. 8B).
(3) Patient Study 2
We investigated whether the artifacts could be improved by using TFS 2-mm or MCS 4-mm in cases in which halo artifacts had appeared in the TFS 4-mm PET images.
In the urine shift pattern, the urinary catheter was not depicted in the 2-mm voxel μ-map (Figs. 2B and 4G,H), and there was a mismatch between the μ-maps and the PET images as with the TFS 4-mm PET images. The result was that the scatter was overestimated, and the halo artifact remained (Figs. 2C, 6A).
In the tube curve pattern, the urinary catheter was depicted in the 2-mm voxel μ-map, and this eliminated the mismatch (Fig. 3B and 5G,H). The result was that the scatter was correctly evaluated, and no halo artifact appeared here (Figs. 3C, 6B).
The MCS 4-mm image also correctly evaluated the scatter in both patterns, and there were no halo artifacts (Figs. 2C, 3C, 6A,B).
Table 2 summarizes the SUV data for the five patient cases with halo artifacts. In the cases with the urine shift pattern, the SUV of the TFS 4-mm PET image was similar to the TFS 2-mm PET image, but that of the MCS 4-mm PET image was larger than both TF-SSS PET images. In the cases with the tube curve pattern, the average SUVs of the TFS 2-mm and MCS-4mm PET images were larger than those of the TFS 4-mm PET images.
The halo artifacts disappeared in the TFS 2-mm PET images in four of the five patients but not in the remaining patient, whereas the halo artifacts were completely absent in the MCS 4-mm PET images in all five patients.