This study evaluated five commercially available calibration phantoms composed of different bone-surrogate materials for use with MECT. The study was designed to assess phantoms based on their attenuation response at clinical MECT energies, using patient data to validate phantom results. Several studies have highlighted the importance of quality control in CT-based BMD quantification, and several phantoms and software packages are commercially available for use with both synchronous and asynchronous calibration in conventional CT [35]. However, to date, little has been published on the appropriateness or limitations of QC phantoms for MECT BMD applications. This study shows that conventional BMD phantoms can introduce bias and inaccuracies in BMD calibration. Using a phantom specifically designed for spectral CT is necessary for accurate calibration, as MECT-specific phantoms are constructed from materials that mimic the energy-dependent attenuation characteristics of human tissues across the energy spectrum used in MECT imaging.

MECT offers several advantages over SECT for BMD imaging, including reduced beam hardening, reduced metal artifacts, and material-specific quantification [30]. MECT enables the execution of material decomposition algorithms and the creation of calcium-specific images [31]. Calcium-specific images may increase the accuracy of BMD estimates by removing the influence of adipose tissue in the bone marrow, which can artificially lower CT numbers and, therefore, BMD [32]. This effect can be minimized with MECT by correctly identifying and quantifying a specific material within a given voxel [14]. The impact of adipose tissue on BMD estimates is particularly relevant to spine-based measurements [33]. MECT can also enable differentiation between cortical and trabecular bone, which is important as trabecular bone is more metabolically active and thus susceptible to early changes in density [34]. Another advantage of using spectral CT over conventional CT for BMD assessment is that three-material decomposition can be applied to remove contributions from iodinated contrast and adipose tissue, thereby enabling BMD quantification in patients undergoing contrast-enhanced CT studies.

Based on the results of this study, the MECT-developed calcium phantom (phantom A) was identified as the most appropriate surrogate for patient-derived DER representation because of its numerical agreement with the attenuation response measured via DER, high measurement precision, and a clinically relevant design spanning the range of calcium concentrations expected in patients. At 100/150Sn, phantom A exhibited a DER of 1.55 (95% CI [1.54–1.57]), with the narrowest confidence interval among all phantoms. The mean difference in the high/low energy ratio between phantom A and patient data was 0.04 (95% CI: -0.14, 0.21) and was not statistically significant (adjusted p-value = 0.9830). Phantom A also demonstrated the highest degree of linearity (R² = 1.000) and the lowest RMSE (2.43), indicating highly linear and reproducible behavior. Importantly, its multiple inserts spanned the range of calcium concentrations observed in the patient cohort, enabling representative calibration. Using the dual-energy response of calcium, a custom material decomposition profile was used to generate a calibration curve from which calcium concentration can be determined from the signal (HU) in calcium-specific images. This capability will be helpful for future applications in which calcium concentration must be directly determined from MECT images.

Phantom B demonstrated a DER of 1.48 (95% CI [1.11, 1.84]) but had the largest RMSE (41.68) among the phantoms, indicating substantial variability in the data. In addition, the negative y-intercept indicates a bias in the data, suggesting that the phantom background material is not a suitable surrogate for soft tissue at the energies used in diagnostic MECT. Phantom C exhibited the most significant deviation from the patient-derived DER (DER = 1.69 (95% CI: 1.56, 1.81)), with a statistically significant mean difference in the high/low energy ratio of -0.26 (adjusted p-value = 0.0253), and was therefore deemed unsuitable. Phantoms D and E yielded the same DER values at 100/150Sn kVp, but differed in their concentration-specific high/low energy ratios. The mean differences relative to patient data were − 0.09 (95% CI: -0.31, 0.12) for phantom D and − 0.09 (95% CI: -0.40, 0.26) for phantom E. Phantom D also exhibited a relatively large RMSE (18.45) in DER, while phantom E was excluded from comparative analysis entirely because its DER was derived from only two data points, precluding reliable estimation of uncertainty. Furthermore, neither phantom D nor E included more than two inserts with HU values within the range observed in the patient cohort, rendering both phantoms unacceptable for clinical application.

Bone-mimicking phantom inserts should be constructed with calcium or other compounds that accurately represent the attenuation characteristics of bony anatomy. Traditional substitutes, such as dipotassium phosphate (K2HPO4), commonly used as a bone substitute in conventional CT phantoms, may not be appropriate for MECT, as they do not accurately reflect the attenuation response observed in patients. This is evidenced by the relatively large RMSE for measurements taken in Phantom B (K2HPO4), indicating a nonlinear relationship between attenuation at high and low energies. The concentrations used in the phantom design should cover the clinically expected range of 0-150 mg Ca/ml [21, 36]. Phantom A included eight calcium rod inserts spanning the clinical range with a maximum calcium concentration of 300 mg/ml. Concentrations beyond this range will result in CT signal saturation and beam hardening artifacts at low kVp, introducing bias into calibration curves and subsequent BMD estimates.

Phantom background material is another important consideration when selecting phantoms for use with MECT. Phantoms designed for spectral CT applications are composed of background materials that exhibit similar attenuation across the energy spectrum used in MECT imaging. Background materials in the phantom inserts (0 mg/mL calcium) should not contain calcium or potassium, as this will bias the measured calcium concentration. Among the phantoms included in this study, phantom A was the only phantom specifically designed for MECT and most closely matched the attenuation response observed in patients. This highlights the importance of matching phantom background materials to tissue properties for both calibration accuracy and the development of reliable quality control protocols.

Results showed that the tube voltage strongly influences material quantification in SECT and MECT. Other protocol parameters, such as radiation dose and image reconstruction approach, are unlikely to affect BMD quantification substantially, provided image quality is sufficient. Consequently, opportunistic BMD measurements should be performed only when phantom calibration data are available at the same tube voltage(s) used in the patient scan. A primary goal of image acquisition in QCT analysis is to minimize the extent to which non-anatomic factors affect calculated results. This is accomplished by establishing a consistent protocol that standardizes as many acquisition parameters as possible. While BMD calibration data would ideally be available at multiple dose levels for validation, maintaining additional calibrations can be burdensome for clinical staff and is of limited relevance when tube current modulation is employed in clinical CT examinations.

Phantom A is comprised of two components – a body phantom with a head phantom insert. Throughout this study, the body phantom was used. At least one study reports that the smaller head phantom insert provides more stable and accurate attenuation estimates than measurements in the body phantom [37]. However, that study used a prototype phantom with higher background attenuation than the commercially available model, so beam hardening affected CT numbers in the inserts more dramatically. Secondly, the prior study used a scanner with 80/140 kVp, and the low kVp beam may not be optimal for larger phantoms and patients. Consideration of clinical application is also important, as the body phantom provides scan conditions that more accurately represent those of abdominal and pelvic scans, where opportunistic BMD evaluation will be integrated into our clinical workflow.

There are several limitations to this study. First, the results are specific to dual-source systems from one manufacturer. There are various approaches to MECT, from system design to acquisition, and these differences could result in some variation [38, 39]. Additionally, spectral separation can affect the accuracy of material quantification, which varies across MECT technologies [38, 39]. The patient data sample size (N = 10) was small. However, the measured calcium concentrations are comparable to those reported in the literature and enabled correlation to phantom data. Many studies report calcium concentrations obtained from vendor or third-party software, but hydroxyapatite and calcium measurements can vary across vendors, necessitating cross-calibration of MECT systems [39, 40]. Independent validation using dedicated phantoms remains underutilized but is critical for ensuring reliable BMD quantification, particularly when proprietary software is unavailable or when comparing results across platforms.