Ra-223 SPECT for semi-quantitative analysis in comparison with Tc-99m HMDP SPECT: phantom study and initial clinical experience

The energy spectra of Ra-223 with low-energy high-resolution (LEHR), MEGP or HEGP collimators and without any collimator were obtained using a dual-headed SPECT system (Discovery NM630, GE Healthcare, Milwaukee, WI). The thickness of the detector crystal was 3/8 in. Hole diameter, hole length and lead thickness of the collimator were 3.0, 58 and 1.1 mm for MEGP and 4.0, 66 and 1.8 mm for HEGP, respectively. Each of the two vials containing Ra-223 solution (1.1 MBq, 2.0 mL) was carefully placed at a distance of 100 mm from the detector under the different conditions where the vials were set around the air or in the chamber filled with iodine contrast media (CT value, 350 HU), and each energy spectrum was then obtained (Fig. 1). The energy spectrum < 500 keV was obtained for 30 min.

Image quality was assessed using a cylindrical phantom filled with water (diameter, 200 mm; height, 210 mm) in which a cylindrical rod (diameter, 45 mm; height, 200 mm) filled with 2.0 kBq/mL of Ra-223 was embedded along the phantom axis (Additional file 1: Figure S1). The radio concentration of the rod was determined based on the phantom study by Hindorf et al. [12]. SPECT/CT scans were performed with MEGP and HEGP collimators using the same protocol (matrix, 64 × 64; pixel size, 8.8 mm; scan orbit, body contour; frames per detector, 30 (6° steps); acquisition time per frame, 60 s; total acquisition time, 30 min). Total acquisition time was determined based on the balance of count statistics and patient durability in our clinical study described later.

First, the energy window setting around 84 keV was evaluated. We focused on image noise and contrast in image quality of Ra-223 SPECT because of the limited injected dose. As studied in Sr-89 imaging [17], we evaluated the impact of characteristic X-rays of lead (75 keV (Kα) and 85 keV (Kβ) [18]) on image quality of Ra-223 SPECT by comparing the energy window widths between ± 20% and ± 10%. The latter width has been used so far [9, 12], although these characteristic X-rays have not been fully included.

The SPECT data with the two energy window widths were reconstructed using 3-dimensional ordered subset expectation maximisation (3D OSEM) algorithm (five subsets, 10 iterations). Butterworth filter was used for image noise reduction (cutoff frequency, 0.20 cycle/cm; order, 10). Attenuation or scatter correction was not performed. Circular regions-of-interest (ROIs) (diameter, 45 mm) were carefully placed on the hot rod in a transaxial slice (slice thickness, 8.8 mm) with a reference of the corresponding CT (Additional file 1: Figure S1). Background counts (BKG) were also measured using the same 8 ROIs, avoiding high activity around the hot rod. Then, the quality of SPECT images was assessed with the following parameters:

Second, the energy window regarding the other peaks was set at 154 keV ± 10% and 269 keV ± 5% according to previous reports [9, 12]. Similarly, Hr, BKG, HBR and CNR were measured. The procedure of the parameter measurement was repeated 10 times (i.e. 10 arbitrary transaxial slices) obtained with the two collimators.

A modified body phantom simulating bowel activity and lumbar spine metastases was developed and scanned. The configuration of the phantom is shown in Fig. 2a. The phantom was filled with water, and two identical 28 mm spheres (11.5 cm³) with 15.0 kBq/mL of Ra-223 and attenuation medium (350 HU) and a tube-shaped chamber with 15.0 kBq/mL of Ra-223 were embedded. The spheres were set in the same longitudinal axis. One of them (sphere 1) was set in the same anteroposterior axis as a tube-shaped chamber so that uptake of the sphere is overlaid by uptake of the chamber on the planar image. The other sphere (sphere 2) was used as reference. Planar static data were obtained for 30 min in a 128 × 128 matrix. Image acquisition and reconstruction for SPECT were the same as those for a cylindrical phantom.

In this phantom study, the following parameters were evaluated; first, using the data of sphere 2, the threshold for spherical volume of interest (VOI) of bone metastasis for clinical Ra-223 SPECT was determined by comparing the sphere volumes on CT and SPECT. Second, the uptake ratio of sphere 1 to sphere 2 was assessed using a 28-mm circular ROI in the anteroposterior view of the planar image and using the threshold method in SPECT. Third, the sphere-to-background ratio for sphere 2 (SBR) was compared between the planar and SPECT images using 48-mm circular background ROIs for the planar image and 48-mm spherical background VOIs for the SPECT (Fig. 2a). Finally, the phantom was repeatedly scanned to evaluate the linearity between radio-concentration and SPECT count in sphere 2.

Both Ra-223 treatment and image acquisition of Ra-223 in patients with prostate cancer were approved by the institutional review board in Keio University Hospital (approval number 20160203). The study was conducted in accordance with the Declaration of Helsinki. The study protocol was registered at the University Hospital Medical Information Network (UMIN000024274). Written informed consent was obtained from 10 patients with prostate cancer who had a bone metastasis diagnosis confirmed with Tc-99m HMDP SPECT/CT.

Injection dose of Ra-223 dichloride solution was 55 kBq/kg per cycle. Image acquisition was performed after the first injection during six cycles of Ra-223 treatment in all patients. One-bed Ra-223 SPECT scanning was performed in an area where the greatest bone metabolism was revealed in Tc-99m HMDP SPECT. SPECT data were collected 2 h after Ra-223 injection successively with the MEGP and HEGP collimators for 30 min each. Since the difference in uptake time from injection to image acquisition could affect image contrast between the two successive SPECT acquisitions (120 vs 150 min), image acquisition was performed first with the MEGP collimator for five patients and with the HEGP collimator for the remaining five. The energy windows were set at 84 keV ± 20% according to the results of the phantom study. Acquisition and reconstruction parameters were identical to those used for the phantom images.

Tc-99m HMDP SPECT/CT was performed 3 h after injection using the LEHR collimator at 141 keV ± 10%. Tc-99m HMDP SPECT data were also reconstructed on a 128 × 128 matrix (pixel size, 4.4 mm) with 3D OSEM (five subsets; 10 iterations) and Butterworth filter (cutoff frequency, 0.33 cycle/cm; order, 10) using attenuation and scatter correction (dual-energy window method) and resolution recovery with Evolution for Bone (GE Healthcare).

With a reference of CT images, the LBR was calculated for metastatic lesions showing increased Ra-223 uptake at least two times greater than uptake in the normal spine with both the MEGP (LBRme) and HEGP (LBRhe) collimators. The LBR for Tc-99m HMDP SPECT was also calculated (LBRtc). VOIs were placed on the metastatic lesions to encompass the selected metastasis at a contour level of 60% of the highest count. For background VOIs, spherical VOIs (50 cm³) were carefully drawn within the liver parenchyma with reference to the corresponding CT images, avoiding the large hepatic vein. Then, the LBR was measured as mean counts in the lesion ROI divided by mean counts in the background ROI.

All statistical analyses were performed using JMP 12.0.1 software (SAS Institute Inc., Cary, NC). All data were expressed as median and range (minimum to maximum) and analysed with non-parametric methods. Comparisons of Hr, BKG, HBR and CNR for the cylindrical phantom between the energy window of 84 keV ± 20% and ± 10% were performed using the Wilcoxon signed-rank test. After selecting the optimal energy window, around 84 keV, comparisons of Hr, BKG, HBR and CNR among the three energy windows were performed using Friedman’s test, and when significant differences were found, the Steel-Dwass test was performed on each pair. Comparisons of SBR for the modified body phantom between planar vs SPECT or MEGP vs HEGP were also performed using the Wilcoxon signed-rank test.

Regarding the clinical study, detection of bone metastasis was compared between Ra-223 SPECT and Tc-99m HMDP SPECT using the Wilcoxon signed-rank test. The relationships between LBRme or LBRhe vs LBRtc were assessed using linear regression analysis. The comparison between LBRme and LBRhe was performed using paired t test. The differences at the 95% confidence level (p < 0.05) were considered to be statistically significant.

The energy spectra of Ra-223 with the MEGP or HEGP collimators and without any collimator are shown in Fig. 1. The peak at 84 keV when using HEGP has a wide width probably because of the characteristic lead X-rays occurring by interaction of primary photons and lead. Detector sensitivity under the use of the HEGP collimator relative to the use of the MEGP collimator and the photopeak height at 154, 269, 351 and 402 keV relative to that at 84 keV under the use of the HEGP collimator are shown in Table 1. The existence of some attenuation medium surrounding the vial had limited impact on the energy spectrum, suggesting that patient body is unlikely to affect the energy spectrum.

Figure 3 shows the comparison of SPECT image quality of the cylindrical phantom between the energy window settings of 84 keV ± 20% and 84 keV ± 10%. Hr, BKG and CNR were significantly higher at 84 keV ± 20% than at 84 keV ± 10%, irrespective of collimators. It should be noted that HBR was similar between the two window settings whereas CNR at 84 keV ± 20% was approximately 50% higher than CNR at 84 keV ± 10% (p < 0.01). Based on these results, the optimal window width around 84 keV for Ra-223 SPECT was determined to be 84 keV ± 20%, and then, further analyses were performed using this window setting.

Additional file 1: Figure S2 shows Hr, BKG, HBR and CNR at 84 keV ± 20%, 154 keV ± 10% and 269 keV ± 5%. Hr at 84 keV was significantly higher with the HEGP than with the MEGP collimator (p < 0.05) whereas Hrs at 154 keV and 269 keV were significantly lower with the HEGP than with the MEGP collimator (p < 0.05). BKG was significantly lower with the HEGP collimator than with the MEGP collimator at all the energy windows (p < 0.01), as visually shown in Additional file 1: Figure S1B and C. HBR at 84 keV was significantly higher with the HEGP than with the MEGP collimator (p < 0.01). Based on these results, further analyses were performed at 84 keV ± 20%.

On SPECT images of a modified body phantom, the volumes of sphere 2 (11.5 cm³) with thresholding at 50, 60, 70 and 80% of the maximum sphere value were 15.8, 9.0, 6.2 and 1.4 cm³ when using the MEGP collimator and 17.9, 12.4, 11.0 and 9.0 cm³ when using the HEGP collimator, respectively. In order to match VOI size with true lesion size, thresholding at 60% of the maximum was used for further phantom and clinical SPECT studies.

Figure 2b, c shows a planar image of the phantom in the anteroposterior view and the mid-sagittal plane of the SPECT image in which the spheres are most clearly shown, respectively. Irrespective of the collimator used, the uptake value of sphere 1 was overestimated by approximately 45% on the planar image whereas it was almost identical to the uptake value of sphere 2 on the SPECT image (Table 2). On the other hand, SBR in SPECT was more than twice as high as in planar imaging (p < 0.001). In both planar and SPECT imaging, SBR was also significantly higher with HEGP than with MEGP (p < 0.01). Figure 4 shows a good linearity between the radio-concentration of Ra-223 and SPECT counts at 84 keV ± 20% with both the MEGP and HEGP collimators.

In the evaluated 10 patients, Tc-99m HMDP SPECT depicted more lesions than Ra-223 SPECT with each collimator (84 vs 51 for HEGP, p < 0.01; 84 vs 36 for MEGP, p < 0.01). In addition, Ra-223 SPECT depicted more lesions with the HEGP than with the MEGP collimator (51 vs 36, p = 0.013), and no lesion was only detected with the MEGP collimator. Thus, semi-quantitative analysis was performed for the 36 metastatic lesions to directly compare the MEGP and HEGP collimators (Table 3). The comparison of background count (patient-based), lesion count, and LBR (lesion-based) between these collimators is shown in Fig. 5. The lesion count was similar between the HEGP and MEGP collimators whereas the background count was significantly lower with the HEGP than with the MEGP collimator (p < 0.01). The relationships between LBRme or LBRhe and LBRtc are shown in Fig. 6. There were positive correlations between LBRtc and LBRme (r = 0.67, p < 0.01) or LBRhe (r = 0.69, p < 0.01). LBRhe was significantly higher than LBRme (p < 0.01). An example of Ra-223 SPECT with the MEGP and HEGP collimators compared to Tc-99m HMDP SPECT in a patient with bone metastasis is presented in Fig. 7. Bowel uptake was seen in nine patients; six of which showed greater bowel activity than uptake in metastatic sites (Table 3). A representative case in which bowel uptake was successfully separated from disease uptake using SPECT is shown in Fig. 8.