Introduction

High sensitivity and high energy resolution single-photon emission computed tomography (SPECT), based on semiconductor detectors, presents unique possibilities for nuclear cardiac imaging [1,2,3]. Importantly, the high energy resolution provides a new opportunity for simultaneous dual-isotope (SDI) SPECT imaging using nuclide combinations such as Tc-99m and I-123 (Tc-99m/I-123) [4,5,6,7]. SDI-SPECT studies were first reported around 1978 [8], although few studies of combinations of new nuclides appeared until the cadmium-zinc-telluride (CZT) detector was developed recently. However, SDI-SPECT offers many advantages, such as examinations under the same physiological conditions, reduction of acquisition times, and improvements in patient throughput [9]. In the Anger SPECT (A-SPECT) system, we used a combination of Tl-201 and I-123 (Tl-201/I-123) to diagnose vasospastic angina in patients with acute myocardial infarction [10, 11]. We were hesitant to conduct that investigation because the combination requires a high dose of Tl-201, but recent remarkable improvements in D-SPECT indicate that it may now be feasible to carry out Tc-99m/I-123 SDI imaging instead [12, 13], which involves lower doses of radiation but is difficult to perform with conventional A-SPECT. In this study, we quantitatively assessed Tc-99m/I-123 SDI imaging with D-SPECT.

Materials and Methods

Semiconductor-Based SPECT System

Our D-SPECT system (Spectrum Dynamics, Israel) is equipped with nine arrays of CZT detectors, with each detector block consisting of 16 × 64 individual pixels with a spacing of 2.46 mm in each dimension, resulting in a total detector surface of 39.4 mm × 157.6 mm [14,15,16]. Each detector can be rotated a maximum of 110° without moving parts and is equipped with a square tungsten parallel-hole collimator (septal thickness, 0.2 mm; pitch, 2.46 mm; and length, 21.7 mm) [14, 17]. Data acquisition was completed in approximately 10 min in all cases. Tc-99m/I-123 data were simultaneously acquired in energy windows centered at 140 keV (window width, 133–145 keV) for Tc-99m, and 159 keV (window width, 152–168 keV) for I-123 emission. Similarly, Tl-201/I-123 data were acquired at 70 keV (window width, 64–77 keV) for Tl-201, and 159 keV (window width, 152–168 keV) for I-123. The collected data were reformatted (matrix size, 64 × 64; voxel size, 4.92 × 4.92 × 4.92 mm3) by reconstruction with and without scatter correction and reorientation. The scatter correction was processed using a scatter correction method, designed for D-SPECT by Kacperski et al. [18].

NaI-Based SPECT System

The NaI-based A-SPECT system (Bright-View; Philips, Japan) used for comparison with D-SPECT consisted of dual-head rotating detectors, and was equipped with a general-purpose low-energy collimator (septal thickness, 0.2 mm; pitch, 1.40 mm; length, 24.7 mm). A set of 32 projection images was obtained (step-and-shoot method, 25 s/projection, 64 × 64 matrices) over a 90° arc. Energy peaks were set to Tc-99m (140 keV), I-123 (159 keV), and Tl-201 (70 keV) and energy window widths were set to 20%. The energy resolution (as a percentage of the full width at half maximum (FWHM)), measured in accordance with the National Electrical Manufacturers Association (NEMA) standard, is Tc-99m (9.9%), I-123 (10.1%), and Tl-201 (13.4%) [19]. Data acquisition was completed in approximately 15 min in all cases. The collected data were reformatted (matrix size, 64 × 64; voxel size, 6.39 × 6.39 × 6.39 mm3) by reconstruction with a filtered back projection and reorientation. Scatter correction could not be used with A-SPECT.

Evaluation of the Energy Resolution

In D-SPECT, energy resolution was measured for Tc-99m (4.34 MBq/ml), I-123 (4.34 MBq/ml), and Tl-201 (4.34 MBq/ml) using tube phantoms (inner diameter, 9.5 mm; height, 300 mm; volume, 21.3 ml). Each nuclide was mixed in a tube phantom using a combination of Tc-99m/I-123 or Tl-201/I-123, and the percentage of FWHM, an index of energy resolution, was measured. In addition, we evaluated the energy resolution of the D-SPECT and A-SPECT systems for a combination of Tc-99m/I-123 or Tl-201/I-123 using an RH-2 phantom (Kyoto-Kagaku, Japan) (Fig. 1a). In the RH-2 phantom, the simulated cardiac phantom (Fig. 1b) was filled with water and fixed. The left ventricle (LV) in the cardiac phantom was divided into six regions (Fig. 1c), each filled with a solution of Tc-99m (0.25 MBq/ml) alone, a solution of I-123 (0.13 MBq/ml) alone, a mixture of Tc-99m (0.50 MBq/ml) and I-123 (0.26 MBq/ml), a solution of Tl-201 (0.19 MBq/ml) alone, a solution of I-123 (0.19 MBq/ml) alone, or a mixture of Tl-201 (0.38 MBq/ml) and I-123 (0.38 MBq/ml). The accuracy of the energy resolution in the system and reconstructed image contrast were evaluated in terms of a simulated transmural defect induced by a difference in energy. In addition, D-SPECT without detector rotation can cause image degradation on the inferior wall due to energy attenuation, as in the case of 180° collection of A-SPECT [20]. To verify this, we also evaluated the case where the cardiac phantom was rotated 180° as shown in Fig. 2a, and the position of the transmural defect was replaced from the anterior wall to the inferior wall (or vice versa).

Fig. 1
figure 1

a Structure of the RH-2 phantom. The phantom was made of acrylic and packed with an acrylic heart, Teflon spine, and wooden lungs. The mediastinal area was filled with water, and the size of the container simulated the human body. b Simulated cardiac phantom included in the RH-2 phantom. c Outline of the left ventricle of the cardiac phantom divided into six regions

Full size image
Fig. 2
figure 2

a Outline of the image obtained by slicing the cardiac phantom in the short-axis direction. Right panel depicts the image where the cardiac phantom is rotated 180°. b Region of interest dimensions and location used for quantitative analysis, created using the Image J software for a combination of Tc-99m/I-123 or Tl-201/I-123. Lower panel depicts the image where the cardiac phantom is rotated 180°

Full size image

Quantification of SDI Imaging

The accuracy of energy resolution and the impact of cross-talk were quantified by measuring the contrast-to-noise ratio (CNR) and the transmural defect contrast in the LV wall (CTD) from the count values of a ventricular short-axis image. Region of interest (ROI) dimensions and locations of the myocardium and the transmural defect, used for quantitative analysis, were set using the Image J software. By rotating the cardiac phantom 180°, the ROI was set in a simulated transmural defect on the anterior wall side and the inferior wall side for each nuclide (Fig. 2b). The area of the background region used for the measurement of the CNR was set to be almost the same as the area of the myocardial region. The same ROI was set for the 10-slice ventricular short-axis image simultaneously obtained from one phantom experiment, and the average and standard deviation of CNR and CTD obtained from each image were calculated. CNR and CTD were calculated using the following equations [21]:

$$ \mathrm{CNR}=\left({M}_{my}-{M}_{BG}\right)/\sqrt{S{D}_{my}^2-S{D}_{BG}^2,} $$
(1)

where Mmy is mean myocardial counts, MBG is mean background counts, SDmy is mean myocardial standard deviation (SD), and SDBG is mean background SD.

CTD was defined as the percentage difference between Mmy and MTD:

$$ {C}_{\mathrm{TD}}\ \left(\%\right)=\left[\left({M}_{\mathrm{my}}-{M}_{\mathrm{TD}}\right)/{M}_{\mathrm{my}}\right]\times 100 $$
(2)

where Mmy is the mean myocardial count and MTD is the transmural defect count. The ideal value for CTD is 100%.

To clarify the correlation between quantitative evaluation and visual evaluation, two cardiologists (experience, 5 and 26 years) and two radiological technologists (clinical experience, 26 and 33 years) participated in a phantom observation test, in which they compared images subjectively in qualitative terms [4].

Results

Figure 3 shows the energy spectrum of the D-SPECT and A-SPECT systems for (a) Tc-99m/I-123 and (b) Tl-201/I-123. The energy resolution (percentage of FWHM) of the D-SPECT system obtained from the data shown in the figures was 5.4%/5.1% for Tc-99m/I-123 and 5.4%/5.3% for Tl-201/I-123, which was approximately two times higher than that of A-SPECT. Figure 4 shows the comparison of the D-SPECT and A-SPECT systems on the short axis images without scatter correction, for Tc-99m/I-123 or Tl-201/I-123. Table 1 shows the CNR and CTD calculated from the data shown in those figures. No notable difference was confirmed in the CNRs of the two systems; however, the combination of Tc-99m/I-123 in D-SPECT increased CTD by 28.7% for Tc-99m and 51.4% for I-123 in the anterior wall compared to A-SPECT without scatter correction. Similarly, the combination of Tl-201/I-123 increased 21.3% for Tl-201 and 21.4% for I-123. Regarding the inferior wall, CTD was 15 to 30% lower than that of the anterior wall in both systems, although a similar improvement was confirmed. In DSPECT, the combination of Tc-99m/I-123 had a slightly better CTD than T1-201/I-123.

Fig. 3
figure 3

The energy spectrum of the D-SPECT and A-SPECT systems at a Tc-99m/I-123 and b Tl-201/I-123. The dotted line shows the energy window width of D-SPECT

Full size image
Fig. 4
figure 4

Comparison of the D-SPECT and A-SPECT systems on the cardiac phantom images (short axis slices) without scatter correction, for Tc-99m/I-123 or Tl-201/I-123

Full size image
Table 1 Comparison of the image quality of D-SPECT and A-SPECT images without scatter correction
Full size table

Figure 5 shows short axis images of D-SPECT for (a) Tc-99m/I-123 and (b) Tl-201/I-123, with and without scatter correction, and Table 2 shows the CNR and CTD calculated from the corresponding data. CTD of Tc-99m/I-123 was improved with scatter correction at both nuclides (p < 0.05), but in Tl-201/I-123, no significant improvement was confirmed in I-123 (p > 0.05). In the inferior wall, CNR improved 0.3 in both nuclides, CTD improved by 10% at Tc-99m and 11.1% at Tl-201.

Fig. 5
figure 5

Short-axis images from the D-SPECT system of the cardiac phantom, with and without scatter correction, for a Tc-99m/I-123 and b Tl-201/I-123. Right panels depict the case where the cardiac phantom is rotated 180° for each nuclide. In a combination of Tc-99m/I-123, the scattering correction reduced the collection counts by approximately 31% for Tc-99m and by approximately 24% for I-123. Similarly, the combination of Tl-201/I-123 showed counts reduced by approximately 53% for Tl-201 and by approximately 27% for I-123

Full size image
Table 2 Image quality of Tc-99m/I-123 and Tl-201/I-123 SDI imaging for D-SPECT with and without scatter correction
Full size table

Discussion

We carried out a phantom experiment to quantitatively assess the merits of Tc-99m/I-123 SDI imaging with D-SPECT. The semiconductor-based D-SPECT system provided approximately two times higher energy resolution than the conventional A-SPECT system for Tc-99m and I-123 [4, 19], implying better image quality and less cross-talk with the simultaneous use of dual isotopes. We carried out image reconstruction with and without scatter correction to examine the influence of cross-talk on SDI imaging [22]. The effects of cross-talk for Tc-99m/I-123 and Tl-201/I-123 SDI imaging could be detected by quantitative analysis such as CNR or CTD, and visual inspection of the RH-2 phantom images. As shown in Table 2, when scatter correction was used, the image quality for Tc-99m and Tl-201 were both improved in comparison with the results when no scatter correction was applied (p < 0.05). D-SPECT without detector rotation also proved to cause image degradation of the inferior wall due to energy attenuation, as in the case of 180° collection of A-SPECT. As a countermeasure against this, scatter correction was considered to be one effective means. By using scatter correction, the image quality of the inferior wall, CNR improved by 0.3, CTD by 10% at Tc-99m and 11.1% at Tl-201. Therefore, it was considered effective to use scatter correction for Tc-99m and Tl-201 in the case of SDI imaging. In a combination of Tc-99m/I-123, the collection counts decreased by approximately 31% for Tc-99m and by approximately 24% for I-123. Similarly, the combination of Tl-201/I-123 showed counts reduced by approximately 53% for Tl-201 and by approximately 27% for I-123. When using scatter correction clinically, it seems necessary to set the collection time so that the LV count will be 1 M counts or more [23]. I-123 was poorer in system sensitivity than Tc-99m or Tl-201 [19], and the improvement in image quality produced by scatter correction was also small (less than 5%), so we conclude that it is better to prioritize count acquisition over scatter correction. Conventionally, we had used a combination of Tl-201-Cl (111 MBq)/I-123-15-odophenyl 3-methyl pentadecanoic acid (111 MBq) to diagnose vasospastic angina with A-SPECT [11]. However, Einstein et al. [12] have reported avoidance of the use of thallium as a dual-isotope cardiac tracer. Tc-99m/I-123 SDI imaging with D-SPECT has benefits such as reduction of the injection dose by changing the nuclide from Tl-201 to Tc-99m. D-SPECT was considered to be capable of performing high-quality SDI imaging using Tc-99m/I–123 as with Tl-201/I-123.

Conclusion

D-SPECT was considered to be capable of performing high-quality SDI imaging using Tc-99m/I-123, and further improvement of image quality can be expected by using scatter correction.