On the optimization of bone SPECT/CT in terms of image quality and radiation dose

Abstract Introduction The purpose of this study was to present the optimization process of CT parameters to reduce patient exposure during bone SPECT/CT without affecting the quality of SPECT images with attenuation correction (AC). Material and methods A fillable phantom reflecting realistic bone scintigraphy conditions was developed and acquired on an AnyScan SC. SPECT/CT scans were carried out with different x‐ray tube current values (10, 20, 30, 40, 50, 60, 70, 90, 110, 130, 150, and 200 mA) at three different high‐voltage values (80, 100, and 120 kV). The contrast (C) and coefficients of variation (CV) in the SPECT images as well as the signal‐to‐noise ratio (SNR) and noise (SDCT) in the CT images with CTDIvol were measured. An optimal acquisition protocol that obtained SPECT/CT images with no artifacts on both CT and SPECT images, acceptable C, SNR, CV, and SDCT values, and the largest reduction in patient exposure compared to the reference acquisition procedure was sought. Results The optimal set of parameters for bone SPECT/CT was determined based on a phantom study. It has been implemented in clinical practice. Two groups of patients were examined according to the baseline and optimized protocols, respectively. The new SPECT/CT protocol substantially reduced patients’ radiation exposure compared to the old protocol while maintaining the required diagnostic quality of SPECT and CT images. Conclusions In the study, we present a methodology that finds a compromise between diagnostic information and patient exposure during bone SPECT/CT procedures.

3-4 mSv, an effective CT dose can vary substantially depending on the device used, exposure parameters, and diagnostic center practice. 4 For higher dose diagnostic quality CT studies, an effective dose may range between 4 and 14 mSv. [4][5][6][7] In CT scans performed to localize pathological tracer uptake found in SPECT and attenuation correction (AC) (low-dose CT), effective doses in the range of 0.6 mSv-4 mSv were reported. 8,9 It is estimated that for 99m Tc-MDP bone scintigraphy with SPECT/CT examination, even reduced exposure parameters may increase the total effective dose in the range of approximately 60-85% compared to SPECT without CT. 10,11 That is why many nuclear medicine specialists are interested in reducing patient exposure to ionizing radiation during SPECT/CT, especially emphasizing CT. 3,[12][13][14][15][16] However, patient exposure during CT is directly related to CT image quality, which cannot lose its diagnostic value due to different exposure parameters. The selection of exposure parameters, mainly x-ray tube current and high voltage, has a major impact on both aspects of CT examination. Excessively high values of these parameters do not necessarily lead to additional diagnostic information, but result in greater exposure of patients and staff. Lower values in turn minimize exposure, but can also lead to lower diagnostic CT image value. Two publications presenting studies on the optimization of the patient dose as well as the CT image quality as a part of bone SPECT/CT examination can be found in the literature. 17,18 In both articles, the authors compared only two sets of CT exposure parameters. In addition, both papers did not justify on what basis the CT parameter sets were selected for comparison.
The relationship between CT image quality and SPECT image quality through the AC procedure is also an important issue. AC of SPECT data using CT data must overcome fundamental difficulties: the polyenergetic characteristics of the continuous x-ray energy spectrum in computed tomography and differences in the radiation energy used to obtain SPECT and CT images. It is therefore necessary to convert x-ray linear attenuation coefficient µ (associated with Hounsfield units [HU]) to the µ coefficient of gamma radiation (140 keV for 99m Tc). AC methods have their limitations, especially in dense materials. [19][20][21] Bone tissue is a mineralized structure with two tissue types: cortical bone and cancellous bone. Cortical bone has a HU value of approximately 1,700-2,000, whereas cancellous bone has a HU value of approximately 150-300. 22 In addition, beam hardening from a dense CT target may affect the determination of the μ coefficient for an object's given element depending on its location in that object. Most CT scanners have implemented procedures to correct beam hardening, working effectively for tissues similar in density to water. However, algorithms may behave less accurately for high-density structures (that is, bones or implants). 23 It was also shown that the noise level in SPECT images reconstructed using the MLEM iterative method is proportional to the noise level in corresponding μ maps. 24 The noise level in CT images can also affect SPECT image quality by lowering the local µ coefficient values. 25,26 Noise and artifacts in CT images can potentially decrease SPECT image quality during the AC procedure and may also have a negative impact on its anatomical assessment. 26 All of these issues may have an impact on SPECT/CT image quality and inaccurate determination of µ maps used for AC. They can lead to artifacts or other errors in SPECT images, especially in bone imaging using 99m Tc-MDP, which binds to high-density tissue. It can be postulated for bone SPECT/CT that the lowest effective CT dose (and the corresponding lowest values of exposure parameters) will be achieved when the impact of low-dose CT on reconstructed SPECT becomes practically unacceptable. 26 The purpose of this study was to optimize CT parameters to reduce patient exposure during bone SPECT/CT without affecting the quality of SPECT images with AC. The phantom body was filled with water mixed with sodium pertechnetate solution Na 99m TcO 4 (a warm background simulating residual activity in a patient's body outside the skeleton). Cylinders (hot sources simulating accumulation of the main pool of activity in the patient's skeleton) were filled with a solution of di-potassium hydrogen phosphate (K 2 HPO 4 ) mixed with water and Na 99m TcO 4 (100 g of salt was dissolved in 67 g of water). 27 It was shown in Ref.

2.A.2 | Image acquisition and reconstruction
All of the measurements were conducted using an AnyScan SC First, a phantom image was acquired with the standard SPECT/ CT protocol used at the Nuclear Medicine Department. SPECT acquisition was performed at the following settings: low-energy high-resolution (LEHR) parallel collimators, 128 × 128 matrix, 4.14 mm pixel size, non-circular orbit, step-and-shoot mode with 32 projection angles acquired over 360°and 20 s per projection, two energy windows (140 keV AE 7.5% for 99m Tc photopeak, 120 keV AE 7.5% for Compton down-scatter). The SPECT images were reconstructed using an ordered-subset expectation maximization iterative reconstruction algorithm (OSEM) with four subsets, 10 iterations with attenuation correction (CT data were used to create attenuation-correction maps), scatter correction (dual-energy window method), and resolution recovery correction. Unenhanced CT scans were obtained using helical rotation (x-ray tube high voltage 120 kV, tube current 50 mA, primary beam collimation 20 mm, rotation 1 s, pitch 1.0, and axial field-of-view 50 cm) and reconstructed using a 512 x 512 matrix, slice thickness of 2.5 mm, filtered back projection method (FBP), convolution kernel recommended by the manufacturer, high resolution, and beam-hardening corrections. The angular variation of the x-ray tube current was not available. The final SPECT/CT image was considered as a baseline in further analysis (the reference image). Second, a series of CT scans was carried out with different x-ray tube current values (10,20,30,40,50,60,70,90,110,130,150, and 200 mA) at three different high voltage values (80, 100, and 120 kV). Each CT was utilized for the AC of the SPECT data acquired with baseline parameters. A total of 36 SPECT/ CT images were obtained in a single measurement series for further analysis. This experiment was repeated three times (3 measurement series). The phantom was refilled between measurement series to maintain a comparable activity concentration. For the quantitative analysis, the performance parameters were determined using the cylindrical volumes of interest (VOIs). VOIs were initially created on the CT images and then copied onto the corresponding SPECT images. For each hot source, three VOIs were delineated: in the center (1/2), in 1/4, and in 3/4 of the cylinder height. The VOI diameter corresponded to each cylinder's diameter.

2.A.3 | Image evaluation
The background region was defined by five VOIs (approximately 5 cm 3 each) in the middle of the phantom between hot sources. For each reconstructed SPECT image, the contrast in the i-th hot source (C i ) was calculated using equation 1.
where Ns i is the total number of counts per mL in the i-th hot source VOI and Nb is the mean total number of counts per mL in five background VOIs. To measure the SPECT image noise, the coefficient of variation (CV) was calculated using equation 2.
where SDb SPECT is the mean standard deviation of counts per mL in five background VOIs.
For each CT image, the signal-to-noise ratio in the i-th hot source (SNR i ) was calculated in relation to the background according to equation (3).
where HUi corresponds to the mean reconstructed HU value in the i-th hot-source and SD CT is the mean standard deviation of the pixel value in five background VOIs. The CT image noise was defined by SD CT .

CT dose assessment
The volumetric CT dose index (CTDI vol ) was determined to estimate the CT radiation exposure. Although the CTDI vol values were automatically documented in a dose report, it was measured for each combination of CT parameters using a standardized CTDI vol body phantom (32 cm) and a calibrated dose meter (X2 base unit with an X2 CT sensor, which was a pencil chamber including an electrometer, RaySafe X2, Unfors RaySafe AB, Billdal, Sweden).

2.A.5 | Optimization methodology
The optimal set of parameters for bone SPECT/CT was selected in the following way: 1. Visual assessment of the SPECT/CT images: exclusion of CT parameter sets for which the reconstructed SPECT/CT images were visually scored lower than the reference (defined as a phantom image acquired with the standard SPECT/CT protocol used at the Nuclear Medicine Department described in detail in Section 2.A.2).
2. Quantitative assessment of the SPECT/CT images: exclusion of CT parameter sets for which the mean C and SNR values in at least two hot sources were significantly lower or the mean CV and SD CT values were higher than the reference.
3. Selection of the optimal set of exposure parameters corresponding to the lowest CTDI vol value. and SD CT in the CT images were determined. The contrast (C) in the SPECT images was calculated as follows: For the aorta VOI, the coefficient of variation (CV A ) in the SPECT images was determined according to equation: The SNR in the CT images was calculated analogous to equation (3) as follows: The CT image noise was defined by SD CTA . based only on patient geometry and did not consider the different attenuation of various tissue types as reported in AAPM Report 220. 31 The new report recommended the use of the water-equivalent diameter (D w ), which considers tissue attenuation in addition to patient geometric size. However, the errors resulting from using a patient size-corrected dose estimate only were in the range of a few percent in the abdominal region. 31 Thus, we used the SSDE based on any of the geometric data in accordance with AAPM Report 204.

2.C | Statistics
In study stage I, Welch's t-test was used to compare the mean C, CV, SNR, and SD CT values determined for the reference image and images obtained using the exposition parameter sets. In study stage II, the paired Wilcoxon/Mann-Whitney test or Student's t-test was used to compare two independent groups depending on the data distribution.
The Shapiro-Wilk method was used to test the data set distribution. A p value of less than 0.05 was considered statistically significant.

3.A.1 | Qualitative analysis
The results of the visual assessment of the phantom SPECT/CT images obtained for various combinations of CT exposure parameters are presented in Table 1.

3.A.2 | Quantitative analysis
Detailed results (mean values with standard deviation of C and SNR for hot sources and CV and SD CT in the reconstructed images for each analyzed parameter sets) are presented in Supplementary A (Table A1 and Table A2).

3.A.3 | Optimization of bone SPECT/CT protocol
The decision strategy presenting the next optimization steps according to the adopted criteria are presented in Table 3. The elimination criteria were as follows: V: eliminated based on the visual assessment (a score of 0 or 1 for at least one of the SPECT or CT images).
C: eliminated based on the quantitative assessment of contrast C (a significantly lower mean C value for at least two hot sources compared to the reference image).
CV: eliminated based on the quantitative assessment of the coefficient of variation CV (a significantly higher mean CV value than the reference image).
S: eliminated based on the quantitative assessment of the signalto-noise ratio (SNR) (a significantly lower mean SNR value for at least two hot sources than the reference image).
σ: eliminated based on the quantitative assessment of the noise SD CT (a significantly higher mean SD CT value than the reference image).
D: eliminated based on the quantitative assessment of the radiation exposure (a higher CTDI vol than the reference image).  Table 4). Fig. 4 shows the SPECT and CT MIP images obtained using the old and optimized protocol from two representative patients.

3.B.2 | Qualitative analysis
The CT and SPECT image quality between groups I and II was not

3.B.3 | Quantitative analysis
The quantitative SPECT evaluation showed no difference in contrast C between the groups, with a median C of 6.6 (min 2. The mean SSDE for group I (8 AE 1 mGy) was significantly higher than for group II (6 AE 1 mGy). The percentage difference between the analyzed mean values was 25% (Table 5).

| DISCUSSION
Optimization of diagnostic procedures using ionizing radiation should be dictated by ALARA radiation protection principle (as low as reasonably achievable). In the context of this study, ALARA means the lowest exposure that leads to acceptable image quality for the vast majority of patients. The process of selecting optimal acquisition and exposure parameters (both in terms of image quality and patient exposure) should be carried out in a manner tailored to each clinical problem.
This study proposes a two-step optimization methodology for bone SPECT/CT (stage I: a phantom study and stage II: a clinical study).
At the first stage, the concept of a fillable phantom reflecting realistic bone scintigraphy conditions (accumulation of radiopharmaceuticals in dense structures) was developed. To the best of our knowledge, only two publications presenting comparable solutions can be found in the literature. The first paper concerned a three-dimensional brain phantom with bone structures maintaining a realistic head contour. 32 In the second paper, the authors described a fillable torso phantom containing a material with a density corresponding to bone tissue. 33 Using a specially constructed phantom allowed the detailed assessment of the impact of the exposure parameters on SPECT/CT images of dense structures, both in terms of quality (visual inspection) and quantity (defined measures of image quality).
The best set of exposure parameters in terms of image quality and exposure was determined on this basis. In the vast majority of cases, no statistically significant difference was found in the average con-  T A B L E 3 The decision strategy for finding an optimal image corresponding to the optimal CT parameters.

X-ray tube current [mA]
High   This study had limitations that merit mention.
First, only one type of SPECT/CT device was used in this study. of bone tissue in which the signal-to-noise ratio is naturally high.
The benefits of iterative reconstruction may not be very significant, but it can be expected that their use is an opportunity to further reduce patient exposure. [35][36][37] The optimal bone SPECT image reconstruction has been also studied. Prior studies recommended using the OSEM iterative method and highlighted its superiority over the FBP technique. [38][39][40] However, different OSEM reconstruction parameters were used in each study cited (number of subsets and iterations with or without additional filtration). The optimal reconstruction parameters should be selected for each SPECT procedure taking into account individual preferences of nuclear medicine specialists analyzing diagnostic images at a given nuclear medicine department. The methodology proposed in this paper can be also appropriate for similar types of research studies. However, the reconstruction parameters can be changed and the reconstruction procedure itself can be repeatedly carried out at any time after SPECT/CT data registration. The selected examination protocol (acquisition and exposure parameters) cannot be changed during data acquisition. Raw acquired data cannot be recollected without exposing patients to additional radiation doses. Hence, the proper selection of the acquisition and exposure parameters is important for the optimal performance of any diagnostic procedure using ionizing radiation.

| CONCLUSIONS
In this study, we presented a methodology that finds a compromise between diagnostic information and patient exposure during bone

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.