Imaging bone turnover assessment through volumetric density-adjusted standardized uptake value using quantitative bone SPECT/CT in osteoporosis

Abstract

Background

Serum bone turnover markers offer limited insight into metabolic activity at the individual vertebra level in osteoporosis. This study introduces a novel image-derived bone turnover marker for individual vertebrae to address this limitation, utilizing volumetric density-adjusted quantitative bone single-photon emission computed tomography/computed tomography (SPECT/CT) with [^99mTc]Tc-DPD. This retrospective study included 177 lumbar vertebrae from 55 postmenopausal South Korean women. The mean standardized uptake value (SUV_mean, g/cm³) and volumetric bone mineral density (vBMD, mg/cm³) were determined within a 2-cm³ volume of interest in the trabecular portion of each vertebra using quantitative SPECT and CT. The density-adjusted mean standardized uptake value (dSUV_mean) was calculated by dividing the SUV_mean by the vBMD and multiplying by 1,000.

Results

SUV_mean correlated positively with vBMD (r = 0.60, p < 0.001). Conversely, dSUV_mean correlated negatively with vBMD (ρ = −0.66, p < 0.001), highlighting the inverse relationship between bone mass and turnover after density adjustment of SUV_mean. Patients with major osteoporotic fractures had lower vBMD (62.5 ± 29.4 vs. 92.3 ± 27.4 mg/cm³, p = 0.001) but higher dSUV_mean (100.8 ± 60.7 vs. 62.6 ± 17.5, p = 0.001) compared to those without fractures, reinforcing the association between fracture prevalence, low bone mass, and high bone turnover.

Conclusion

Volumetric density-adjusted quantitative bone SPECT/CT offers a novel image-derived bone turnover marker for assessing bone turnover in osteoporosis. This method provides a precise assessment of fragility at the individual vertebra level, which may enhance personalized osteoporosis management.

Background

The incidence of osteoporosis and the associated socioeconomic burden of osteoporotic fracture have increased worldwide due to the aging population [1]. The primary goal of osteoporosis treatment is to minimize fracture risk and prevent subsequent fragility fractures, especially among high-risk individuals [2]. Thus, there is a high demand for accurate assessment of bone strength-related fracture risk.

Although bone mineral density (BMD), typically measured by dual-energy X-ray absorptiometry (DXA), is a crucial determinant of fracture risk, higher bone mass does not entirely eliminate fracture risk [3,4,5]. This has led to growing recognition of “bone quality” as an integral aspect of bone strength, distinct from bone mass. Bone quality encompasses various factors, including bone turnover, microarchitecture, mineralization, and microdamage [6,7,8]. Among them, several parameters provide quantitative indicators of bone turnover. Although bone histomorphometry is considered the gold standard for measuring bone turnover [9], it is invasive and not widely used. As a result, serum bone turnover markers (sBTMs), which include bone formation and bone resorption markers, are favored in clinical practice due to their simplicity and cost-effectiveness [8]. However, sBTMs offer a cumulative measure of whole-body bone turnover, lacking the specificity needed for individual bone assessment.

Bone scintigraphy using bone-seeking radiopharmaceuticals has been extensively used to track active bone turnover and metabolism [10]. These radiopharmaceuticals bind to hydroxyapatite in bone tissue, reflecting osteoblastic activity and mineralization rates [11]. However, traditional bone scintigraphy produces two-dimensional images primarily subject to visual analysis, limiting their quantitative utility. Recent advancements in imaging technology, particularly the introduction of hybrid single-photon emission computed tomography/computed tomography (SPECT/CT) system, have enabled three-dimensional scintigraphic imaging [12]. This allows for precise quantification of radiotracer activity and provides accurate anatomical information [13, 14]. CT-based attenuation correction in SPECT offers quantitative turnover rate, such as standardized uptake value (SUV), similar to those in static PET. Additionally, the CT component provides bone density information in Hounsfield units (HU), comparable to quantitative CT (QCT) and BMD measurements [15, 16].

Clinically, bone SPECT/CT is valuable for differentiating the cause of lower back pain and for quantitative analysis [17]. Previous studies have attempted to quantify bone turnover in the lumbar spine (L-spine) using a volume of interest (VOI) approach [18,19,20]. These studies facilitated direct comparison with BMD and showed a positive relationship between SUV and L-spine bone density parameters. However, this contradicts the established inverse correlation between bone turnover and bone mass [21, 22], suggesting that SUV alone may not accurately reflect bone turnover, similar to sBTMs.

The avidity of bone-seeking radiotracer at active bone turnover sites is influenced by the available bone surface or volume for tracer uptake [23, 24]. Since bone formation rate per unit of bone surface or volume equates to bone turnover by histomorphometry definitions [9], further dimensional or volumetric adjustment of SUV is necessary for accurate normalization of turnover rates in individual vertebrae. Given that BMD influences tracer accumulation, we hypothesize that normalizing SUV by BMD, rather than using SUV alone, could more accurately reflect the bone turnover rate at these sites. In this context, our study explores the feasibility of using volumetric bone mineral density (vBMD)-adjusted SUV from quantitative bone SPECT/CT as a novel marker for bone turnover. We aimed to elucidate its significance by correlating it with the osteoporosis-related parameters.

Methods

Patient selection

We retrospectively recruited postmenopausal women who underwent both bone SPECT/CT of the L-spine and DXA within a three-month interval from May 2017 to August 2021 after excluding males and pre-menopausal women. Following the International Society for Clinical Densitometry recommendations [25], post-menopausal women with at least two vertebrae suitable for quantification without compression fractures, degenerative changes, or spinal implants were included. A comprehensive review of electronic medical records and radiologic images was conducted to assess clinical characteristics as outlined in the Fracture Risk Assessment Tool (FRAX), including the presence of prevalent major osteoporotic fractures (MOF) in the hip, spine, radius, or humerus. The 10-year probability of a MOF or hip fracture was calculated using the South Korean reference available on the FRAX website (www.sheffield.ac.uk/FRAX). Reasons for undergoing bone SPECT/CT were also identified. This study was approved by the Institutional Review Board (IRB No. 2020 − 1061) of our institution, and the need for informed consent was waived.

Bone SPECT/CT acquisition protocol and quantitative image analysis

All patients underwent whole-body planar scintigraphy at least three hours after an intravenous injection of approximately 740 MBq (20 mCi) of [^99mTc]Tc-labeled 3,3-diphosphono-1,2-propanodicarboxylic acid ([^99mTc]Tc-DPD). Data including patients’ body weight and height, initial and residual radioactivity before and after injection, and the timing of radioactivity measurements were recorded for precise dose calculation after decay correction. SPECT/CT images of the L-pine and/or pelvis were acquired using a hybrid SPECT/CT camera (Symbia Intevo 16, Siemens). For quantification, the ^99mTc activity was cross-calibrated between the dose calibrator and SPECT scanner. SPECT images were acquired over 360° using a step-and-shoot mode with 45 frames per detector head view, each lasting 22 s, and non-circular orbit. A non-enhanced CT scan was subsequently acquired with 110 kV and 40-reference mAs using adaptive dose modulation (CARE Dose 4D) and a slice thickness of 2 mm. Images were reconstructed to generate attenuation-corrected SPECT data using the xSPECT-Bone™ reconstruction algorithm (Siemens) with 1 subset and 24 iterations, enabling post-processed image quantification with a 256 × 256 matrix size.

For quantitative analysis, the body weight-based mean standardized uptake value (SUV_mean, g/cm³) and mean HU (HU_mean) of each vertebra from the first (L1) to the fourth (L4) lumbar vertebrae were measured by dedicated software (Syngo Via, Siemens). A 2 cm³ spherical VOI was drawn in the trabecular center of each vertebra, avoiding cortical bone and the basivertebral vein (Fig. 1a).

Fig. 1

Representative images of the quantitation in measurable and non-measurable vertebra. (a) Example of measuring the HU_mean and SUV_mean parameters using a 2 cm³ spherical volume of interest (VOI) from L-spine SPECT/CT in one shot by a dedicated software. The image show axial and sagittal views with the green VOI placed in the measurable trabecular center of each L1–L3 body, while the non-measurable L4 body, with vertebral wedging and active degenerative changes, is also presented. (b) Example of measuring bone density parameters from QCT using a 2 cm³ cylindrical VOI in the same area as (a)

The SUV_mean equation was as follows:

$$\begin{aligned}\it \:SUVmean\:(g/{cm}^{3})= & \frac{\begin{array}{c}\text{Mean radioactivity concentration} \\ \text{in the volume of}\\\text{interest of tissue} \, (MBq/{cm}^{3}\:)\end{array}}{Injected \, dose \, (MBq)/ Patient's \, body \, weight \,(g)}\end{aligned}$$

QCT measurement

Using the CT component of SPECT/CT, we analyzed the vBMD (mg/cm³) for the same vertebral bodies retrospectively using QCT Pro (Mindways Software Inc.), with asynchronous phantom-based calibration. A cylindrical VOI equivalent to 2 cm³ was semi-automatically placed in the trabecular center of each vertebra and manually adjusted to avoid cortical bone and the basivertebral vein (Fig. 1b).

Deriving image-based bone turnover parameters

Given that trabecular bones in the vertebrae exhibit higher turnover rates than cortical bones and are crucial in determining the response to anti-osteoporotic medication [26], adopting nearly simultaneous VOIs from SPECT and QCT for density adjustment is ideal for assessing bone turnover. We derived an additional variable, bone density-adjusted SUV_mean (dSUV_mean), by multiplying 1,000 by SUV_mean and then dividing the vBMD in each vertebra:

$$\it \:dSUVmean=\frac{\text{S}\text{U}\text{V}\text{m}\text{e}\text{a}\text{n} \, (g/{cm}^{3})\:\times\:\:\:\text{1,000}\:}{vBMD\:\left(mg/{cm}^{3}\right)}$$

Similarly, HU-adjusted SUV_mean (hSUV_mean) was calculated by multiplying 1,000 by SUV_mean and then dividing by HU_mean:

$$\it \:hSUVmean\:(g/{cm}^{3})=\frac{\text{S}\text{U}\text{V}\text{m}\text{e}\text{a}\text{n} \, (g/{cm}^{3})\:\times\:\:\text{1,000}\:}{HUmean}$$

DXA acquisition and image analysis

Areal BMD (aBMD, g/cm²) from L1 to L4 was measured using DXA equipment (GE Lunar Prodigy Advance). The region of interest (ROI) was automatically generated by the equipment’s software and manually adjusted by a skilled technologist when necessary. Trabecular bone score (TBS) was retrospectively analyzed from L-spine DXA scans using iNsight software version 3.0 (Med-Imaps) with the ROI corresponding to those designated for BMD measurement.

Statistical analysis

Descriptive statistics, expressed as mean ± standard deviation, were used to summarize the baseline characteristics of the study population. Pearson’s (r) or Spearman’s correlation coefficients (ρ) were employed for patient- and vertebra-based analyses to examine relationships between variables. In the patient-based analysis, measured values from L1 to L4 were averaged to create a single representative value per patient. Partial correlation coefficients were additionally used to adjust for potential confounding factors. Group comparisons were conducted using the Mann–Whitney U test. A p-value of < 0.05 was considered statistically significant. All analyses were performed using SPSS version 23.0 (IBM Corp.).

Results

Baseline demographics

Figure 2 presents the patients’ recruitment process. After excluding males, pre-menopausal women, and those with non-quantifiable vertebrae, 55 postmenopausal women with at least two measurable vertebrae were included in the study. Among them, 24 patients (43.6%) had prevalent MOF. The baseline characteristics of the patients are summarized in Table 1. The majority (45/55, 81.8%) underwent SPECT/CT due to low back or hip pain. Nine patients (16.3%) had a history of glucocorticoid intake, and 25 patients (45.5%) exhibited clinical features suggestive of secondary osteoporosis, such as diabetes mellitus, Parkinson’s disease, or were on anticoagulation therapy. A total of 177 intact vertebrae were included in the quantitative analysis.

Fig. 2

Flow chart of patient selection

Table 1 Baseline characteristics of the patients

Comparison of SPECT/CT, DXA, and QCT parameters

The average value (mean ± SD) of SUV_mean, HU_mean, and vBMD from SPECT/CT for all patients was 5.2 ± 1.3 g/cm³, 99.1 ± 39.9, and 79.3 ± 31.7 mg/cm³, respectively. The aBMD, T-score, and TBS values from DXA were 0.90 ± 0.14 g/cm², − 2.1 ± 1.1, and 1.4 ± 0.1, respectively. Both patient- and vertebra-based analyses demonstrated a very strong correlation between HU_mean from SPECT/CT and vBMD from QCT (r = 0.99, p < 0.001; r = 0.99, p < 0.001, respectively; Supplementary Fig. 1a and 2a) and a moderate-to-weak correlation between HU_mean from SPECT/CT and aBMD from DXA (r = 0.47, p < 0.001; r = 0.36, p < 0.001). Additionally, dSUV_mean and hSUV_mean showed a very strong correlation (r = 0.98, p < 0.001; r = 0.97, p < 0.001; Supplementary Fig. 1b and Fig. 2b).

SUV_mean showed a strong positive correlation with HU_mean (r = 0.60, p < 0.001; r = 0.60, p < 0.001) and vBMD (r = 0.60, p < 0.001; r = 0.60, p < 0.001) and a weak positive correlation with aBMD (r = 0.25, p = 0.062; r = 0.20, p = 0.009), indicating that higher bone density was generally associated with SUV_mean. However, TBS showed no significant relationship with SUV_mean (r = 0.18, p = 0.18; r = 0.12, p = 0.20), as depicted in Fig. 3.

Fig. 3

The graphs show the correlations of SUV_mean with vBMD, aBMD, and TBS using Pearson’s correlation coefficient. Panels (a-c) represent the patient-based analysis, and panel (d-f) represent the vertebra-based analysis

Conversely, dSUV_mean showed a strong negative correlation with HU_mean (ρ = −0.69, p < 0.001; ρ = −0.66, p < 0.001), vBMD (ρ = −0.69, p < 0.001; ρ = −0.66, p < 0.001) and a moderate or weak negative correlation with aBMD (ρ = −0.43, p < 0.001; ρ = −0.36, p < 0.001; Fig. 4). Interestingly, inverse correlations in exponential functions between dSUV_mean and density parameters (aBMD, vBMD) were observed regardless of fracture history, and these trends remained significant even after adjusting for age and BMI, as indicated in Table 2. The dSUV_mean showed a weak-to-very weak negative correlation with TBS (ρ = −0.37, p = 0.006; ρ = −0.18, p = 0.047).

Fig. 4

The graphs show the correlations of dSUV_mean with vBMD, aBMD, and TBS using Spearman’s correlation coefficient. Panels (a-c) represent the patient-based analysis, and panel (d-f) represent the vertebra-based analysis

Table 2 Partial correlation coefficients among quantitative parameters of bone strength corrected for age and BMI

Comparison of image-derived parameters and clinical factors

In terms of age, there was a moderate negative correlation with both SUV_mean (r = − 0.44, p = 0.001) and vBMD (r = − 0.59, p < 0.001), whereas dSUV_mean showed a weak positive correlation (r = 0.34, p = 0.01). BMI did not show any significant correlation with these parameters. Compared with patients without MOF (Table 3), those with prevalent MOF were older (69.6 ± 9.9 years vs. 63.5 ± 5.8 years, p = 0.012) and had a higher 10-year probability of MOF (13.1 ± 6.4% vs. 7.4 ± 3.5%, p < 0.001) and hip fractures (5.2 ± 4.2% vs. 2.2 ± 2.1%, p < 0.001) according to FRAX. They also tended to lower HU_mean (78.9 ± 38.0 vs. 114.7 ± 34.3, p = 0.001), lower vBMD (62.5 ± 29.4 mg/cm³ vs. 92.3 ± 27.4 mg/cm³, p = 0.001), lower aBMD (0.85 ± 0.16 g/cm² vs. 0.93 ± 0.11 g/cm², p = 0.010), and lower TBS (1.34 ± 0.06 vs. 1.38 ± 0.09, p = 0.019), but higher dSUV_mean (100.8 ± 60.7 vs. 62.6 ± 17.5, p = 0.001) and hSUV_mean (84.6 ± 59.8 g/cm³ vs. 50.6 ± 14.7 g/cm³, p = 0.004) as seen in Fig. 5, while SUV_mean did not show significant difference (5.0 ± 1.4 g/cm³ vs. 5.5 ± 1.2 g/cm³, p = 0.131). After adjusting for age and BMI, vBMD (r = − 0.35, p = 0.011) and dSUV_mean (ρ = 0.33, p = 0.017) remained significantly associated with the presence of MOFs, while SUV_mean (r = − 0.03, p = 0.83) did not demonstrate such a relationship.

Table 3 Comparison of parameters between the groups upon MOF

Fig. 5

The graphs show the differential values of SUVmean (a), vBMD (b) and dSUVmean (c) in patient-based analysis according to the major osteoporotic fractures (MOF) status. Box plots display the median, interquartile range, and outliers

Discussion

In this study, we introduced a novel image-based parameter of bone turnover, the equivalent vBMD-adjusted SUV_mean (dSUV_mean), using quantitative bone SPECT/CT. Our findings revealed an inverse correlation between dSUV_mean and bone density parameters, supporting the established concept that a high bone turnover rate correlates with lower bone mass. In contrast to previous studies and our findings, which showed a positive correlation between SUV_mean and bone density parameters [18,19,20], SUV_mean appears to be more influenced by bone mass. On the other hand, dSUV_mean serves as an indicator of turnover rate per unit density, offering a novel perspective distinct from SUV_mean. These findings underscore the importance of adjusting SUV for bone density to accurately assess bone turnover.

Our findings align with prior research that utilized [¹⁸F]NaF PET for evaluating bone turnover [23, 27, 28]. Frost et al. normalized bone turnover by using DXA-derived bone mineral apparent density (BMAD)-corrected net fluoride uptake, demonstrating higher turnover in osteopenic and osteoporotic groups compared to the normal group [23]. These findings are consistent with our observations, emphasizing the importance of density adjustment for accurate bone turnover assessment. However, BMAD remains a surrogate measure [29], whereas our study utilized direct vBMD measurements from the same VOI, providing more precise data. Additionally, the dynamic [¹⁸F]NaF PET approach has practical limitations, including extended acquisition time and the necessity for input function.

Our study expands the foundational SPECT-only works of Front et al. [30] and Israel et al. [31]. They demonstrated the potential of SPECT for quantifying bone metabolism but lacked the precision offered by CT integration. Key advancements in our methodology include improved quantitation accuracy through CT-based attenuation correction and density adjustment, higher resolution for defining a region of interest in the trabecular bone with detailed anatomical information from CT, and the use of advanced algorithms, such as xSPECT-Bone™, for more reliable value, SUV.

Our approach of adjusting SUV for vBMD using quantitative bone SPECT/CT is cost-effective and efficient. Unlike usual static SUV measurement [32], which assumes a soft tissue density of approximately 1 g/cm³, our method corrects for the inherent differences in bone density. This adjustment is crucial for accurately reflecting bone turnover. However, static bone SPECT/CT measurements face challenges such as partial volume effects, non-specific tracer binding, and signal noise, which must be considered.

The inverse relationships between the dSUV_mean and density parameters (aBMD or vBMD) mirror those between sBTM and bone strength parameters [21, 22]. The significant inverse correlation between dSUV_mean and vBMD aligns with the established non-linear relationship between fracture risk and bone density [33], suggesting dSUV_mean could serve as a potential new biomarker, or imaging bone turnover marker (iBTM). Further research is needed to explore the characteristics of iBTM and its potential clinical applications, such as diagnosing metabolic bone diseases, predicting fracture risk in osteoporosis, and monitoring therapy efficacy [34].

Patients with prevalent MOF exhibited traits similar to those noted in prior studies, such as lower BMD, TBS, and advanced age [35], alongside lower SUV_mean and higher dSUV_mean, supporting the link between high bone turnover, low bone mass, and spinal fractures [36].

Previous studies often used inconsistent VOIs that included the cortical bone and/or basivertebral vein [18,19,20]. In contrast, we exclusively measured SUV from the trabecular center, using a VOI nearly identical to conventional QCT [15], likely resulting in more accurate density adjustment.

Quantitative L-spine SPECT/CT has significant opportunistic value for osteoporosis screening, offering one-shot information on bone density and quality-related markers [37, 38]. However, precise radioactivity measurement, standardized image acquisition, and accurate quantitative data from both SPECT and CT components are essential prerequisites for this application [39].

This study has limitations, including a small sample size of postmenopausal Asian women, limiting the generalizability of the results. Future research should establish normal SUV values with a coefficient of variance according to radiotracer, age, sex, and race. Additionally, due to the retrospective nature of this study, we could not compare dSUV_mean with sBTM. Finally, the semi-automatic VOI drawing limited the representation to a restricted trabecular area in the spine. A more refined VOI encompassing the entire trabecular area would be preferable [40].

Conclusion

Volumetric density-adjusted quantitative bone SPECT/CT offers a novel image-derived bone turnover marker for assessing bone turnover in osteoporosis, providing a precise assessment of fragility at the individual vertebra level, which may enhance personalized osteoporosis management.