Optimization of the Differentiation and Quantification of High-Z Nanoparticles Incorporated in Medical Devices for CT-Guided Interventions

Abstract

Purpose:

Material differentiation has been made possible using dual-energy CT (DECT), in which the unique, energy-dependent attenuating characteristics of materials can provide new diagnostic information. One promising application is the clinical integration of biodegradable polymers as temporary implantable medical devices impregnated with high-atomic number (high-Z) materials. The purpose of this study was to explore the incorporation of high-atomic-number (high-Z) contrast materials in a bioresorbable inferior vena cava filter for advanced computed tomography (CT)-based monitoring of its location and differentiating from surrounding materials.

Materials and Methods:

Imaging optimization and calibration studies were performed using a body phantom. The dual-energy CT (DECT) ratios for iron, zirconium, barium, gadolinium, ytterbium, tantalum, tungsten, gold, and bismuth were generated for peak kilovoltage combinations of 80/150Sn, 90/150Sn, and 100/150Sn kVp in dual-source CT via linear regression of the CT numbers at low and high energies. A secondary calibration of the material map to the nominal material concentration was generated to correct for use of materials other than iodine. CT number was calibrated to the material concentration based on single-energy CT (SECT) with additional filtration (150Sn kVp). These quantification methods were applied to monitoring of biodegradable inferior vena cava filters (IVCFs) made of braided poly(p-dioxanone) sutures infused with ultrasmall bismuth nanoparticles (BiNPs) implanted in an adult domestic pig.

Results:

Qualitative material differentiation was optimal for high-Z (>73) contrast agents in DECT. However, quantification became nonlinear and inaccurate as the K-edge of the material increased. Using the high-energy (150Sn kVp) data component as a SECT scan, the linearity of quantification curves was maintained with lower limits of detection than with DECT. Among the materials tested, bismuth had optimal differentiation from iodine in DECT while maintaining increased contrast in high-energy SECT for quantification (11.5% error). Coating the IVCF with BiNPs resulted in markedly greater radiopacity (maximum CT number, 2028 HU) than that of an uncoated IVCF (maximum CT number, 127 HU). Using DECT imaging and processing, the BiNP-IVCF could be clearly differentiated from iodine contrast injected into the inferior vena cava of the pig.

Conclusions:

These findings may improve widespread integration of medical devices incorporated with high-Z materials into the clinic, where technical success, possible complications, and device integrity can be assessed intraoperatively and postoperatively via DECT imaging.

INTRODUCTION

Considerable progress in the development of biodegradable polymeric materials has been made due to the emergence of novel biomedical technologies such as tissue engineering, regenerative medicine, controlled drug release, and bionanotechnology, all of which require biodegradable platforms.^1,2 Gradual degradation of these biodegradable materials is advantageous because it allows for natural tissue regeneration while preventing the initiation of a chronic immune reaction to the presence of permanent foreign material.² For instance, inferior vena cava filters (IVCF) are implanted in the inferior vena cava to prevent thromboembolism in the lungs, which are common in the early hypercoagulable states of acute traumatic injuries. However, after the risk for thromboembolism has passed, patients must undergo a second surgical procedure to retrieve conventional metal IVCFs since leaving the filters may cause complications such as vessel perforation, thrombophlebitis, and thrombosis.³ Therefore, it is beneficial for IVCFs to be made from resorbable materials to prevent potentially severe complications. The rate of conventional IVCF retrieval is less than 60%; thus, a bioresorbable IVCF would mitigate these issues.⁴ Nonetheless, evolution in the development and clinical adaptation of these absorbable biomaterials has been slow, which can be attributed to several unique challenges, including inability to noninvasively monitor the position, integrity, and degradation of a device over an extended period.⁵ The ability to noninvasively monitor and quantify the amount of contrast agent embedded in medical devices would provide insights into the integrity and function of these materials over time.

Few preclinical studies focus on visualizing biodegradable medical devices in vivo. Imaging modalities such as ultrasound elastography, shear wave imaging, photoacoustic imaging, and near-infrared fluorescence are reported to measure the integrity of biomaterials in vivo.^6–9 However, in addition to having limited penetration depth, such imaging modalities are not routinely used nor readily available in the clinic. Researchers have investigated magnetic resonance imaging in monitoring the degradation of polyvinylidene fluoride¹⁰ and collagen¹¹ labeled with ultrasmall superparamagnetic iron oxide, but degradation was not observed in these studies. Furthermore, magnetic resonance imaging is a high-cost imaging modality and is not optimal for frequent device monitoring.

Computed tomography (CT) is a routinely used noninvasive imaging modality that is available in both interventional radiology suites for initial device deployment and outpatient settings for long-term monitoring of medical devices. It offers high in-plane spatial resolution and visualization of anatomical structures are enhanced using intravenous contrast agents.¹² Current clinically used CT contrast agents consist mainly of tri-iodinated benzene rings.¹² Our group has incorporated iodinated molecules in medical devices made of polycaprolactone¹³ and poly(p-dioxanone),¹⁴ resulting in significant radiopacity with a first-order signal degradation rate in CT when implanted in mice. However, iodine has suboptimal X-ray absorption at high X-ray tube potentials (120–140 kVp). Thus, replacing iodine with elements that have higher X-ray absorption coefficients is of significant interest in the field.^15,16

Most researches are focused on the use of gold as a nanoparticle contrast agent.¹⁵ Researchers showed the degradation of collagen conjugated with gold nanoparticles (AuNP) to be strongly correlated with a decrease in CT number during long-term monitoring.¹⁷ Our group has also demonstrated the use of clinical CT imaging in long-term monitoring of AuNP-infused IVCF in pigs.¹⁸ However, due to the high cost of gold, alternative high-atomic-number (high-Z) contrast agents should also be investigated. Additionally, use of high-Z nanoparticles incorporated within a medical device would be advantageous because the nanoparticles can potentially be differentiated from iodinated agents. This is especially valuable for medical devices that require iodinated contrast during the device implantation and/or for subsequent assessment of their functions. In the case of IVCFs, iodine contrast is necessary during deployment and to assess and monitor intravenous blood clots caught by the filter. Furthermore, patients may not be eligible to receive a non-contrast CT scan solely for longitudinal device monitoring, and thus differentiation of the device from iodine is critical.

Material differentiation has been made possible using dual-energy CT (DECT), in which the unique, energy-dependent attenuating characteristics of materials can provide new diagnostic information.^12,19,20,21 Advanced DECT provides further improvements in reducing image artifacts,²² differentiating tissues,^23,24 simulating noncontrast images,²⁵ and improving detection and quantification of contrast material concentrations.^26,27 Additional filtration may also be added to an X-ray tube to optimize imaging of certain high-Z contrast agents,^12,15,16 reduce energy overlap of DECT spectra,²⁸ and lower the radiation dose to the patient.²⁹ Introduction of new contrast agents designed specifically for these imaging technologies could provide the ability to distinguish between intravascular contrast agents and radiopaque medical devices.¹² X-ray attenuation depends on the mass attenuation coefficient (μ/ρ; cm²/g), density (ρ; g/cm³), and X-ray path length through each material. The mass attenuation coefficient is a function of energy, and the K-edge represents the minimum energy required to liberate a K-shell electron from a material, providing a large increase in X-ray attenuation (Figure 1A).³⁰ For materials with K-edges within the typical CT energy range (Figure 1B), DECT allows users to differentiate between contrast agents based on the dual-energy ratio (DER), the ratio between the CT number at the low and high energies.^12,27 Thus, it is advantageous to develop medical devices that incorporate alternative high-Z contrast materials.

Figure 1.

Simulated incident X-ray spectra, mass attenuation coefficients, and K-edges of different materials. (A) Simulated incident X-ray spectra for multiple clinically available peak kilovoltage settings. The presence of additional tin filtration in the X-ray tube is indicated by 150Sn kVp. All spectra were generated using Spektr 3.0 software.³¹ (B) Mass attenuation coefficients (μ/ρ) of six materials with K-edges within the diagnostic energy range of CT.

In this study, we evaluated the use of both DECT material decomposition and high-peak kilovoltage (kVp) single-energy CT (SECT) quantification with additional tin filtration using multiple potential contrast materials (iron, zirconium, barium, gadolinium, ytterbium, tantalum, tungsten, gold, and bismuth). These analysis methods were used to evaluate a biodegradable inferior vena cava filter (IVCF) incorporated with bismuth nanoparticles (BiNPs) implanted in a porcine model in the presence of iodinated contrast.

MATERIALS AND METHODS

The study has two major parts: (1) to determine the optimal contrast agent among different high-Z materials in terms of quantification accuracy and ability to differentiate from iodine using phantom studies in SECT and DECT; and (2) to estimate the amount of material present in a pilot pig study of BiNP-infused IVCF using SECT and DECT. In the first part, material samples were placed in an elliptical solid water phantom (Gammex, Multi-Energy CT Phantom, Sun Nuclear Corporation, Middleton, WI) and scanned with a dual-source CT scanner (SOMATOM Force; Siemens Healthineers, Forchheim, Germany) to determine the optimal DECT and SECT settings, as described in detail below. For the second part, calibrations were repeated on our veterinary CT scanner, a split-filter DECT scanner (SOMATOM Definition Edge; Siemens Healthineers), and were validated with additional material samples and an in vitro IVCF. Because of the small IVCF size, a smaller phantom (Multi Energy CT Quality Assurance Phantom, Kyoto Kagaku, Kyoto, Japan) was used. Figure 2 shows a schematic for the phantoms and scanners used in the study.

Figure 2.

Schematic diagram of imaging methodology. Figure created with BioRender (biorender.com).

Part I: Selection of best performing contrast material in DECT and SECT

Calibration Materials

The materials investigated in this study are listed in Table 1. Solutions used were purchased from Sigma-Aldrich (St. Louis, MO) and Fisher Scientific (Hampton, NH). All chemicals were used without further purification unless otherwise noted. Concentrations of 0, 2, 5, 10, 15, and 20 mg/mL were prepared in 50-mL conical tubes for use in phantom scanning.

TABLE 1.

The materials investigated in the study

Element	Z number	K-edge (keV)	Salt	Purity	Solvent	CAS No.
Iron (Fe)	26	7.1	Fe(III) chloride	≥97.0%	Nitric acid (70%)	7705-08-0
Zirconium (Zr)	40	18.0	Zr(IV) isoproproxide	≥99.9%	Nitric acid (70%)	14717-56-7
Barium (Ba)	56	37.4	Ba(II) oxide	≥97.0%	Nitric acid (70%)	15552-14-4
Gadolinium (Gd)	64	50.2	Gd(III) oxide	≥99.9%	Nitric acid (70%)	12064-62-9
Ytterbium (Yb)	70	61.3	Yb(III) chloride hexahydrate	≥99.9%	Nitric acid (70%)	19423-81-1
Tantalum (Ta)	73	67.4	Ta(V) oxide	≥99.9%	Nitric acid (70%)	1314-61-0
Tungsten	74	69.5	Tungstic acid	≥99.9%	Hydrofluoric acid (48%)	7783-03-1
Gold (Au)	79	80.7	Au(III) chloride trihydrate	≥99.9%	Hydrochloric acid (37%)	16961-25-4
Bismuth (Bi)	81	90.5	Bi(III) acetate	≥99.9%	Nitric acid (70%)	29094-03-9

CAS, Chemical Abstracts Service.

Calibration Phantom

An elliptical solid water phantom measuring 40 × 30 × 15 cm (Multi-Energy CT Phantom; Sun Nuclear, Inc., Middleton, WI) was used to scan the reference standards. Eight inserts within the small body insert were replaced with prepared 50-mL conical tubes (diameter: 3 cm) containing serial dilutions of contrast materials following the methodology described by Jacobsen et al.³² The center insert in the body phantom was replaced with a 50-mL vial of distilled water, whereas the remaining positions in the phantom were filled with soft tissue-equivalent inserts to minimize beam-hardening artifacts.

Phantom Scan Protocol

Optimization of imaging for visualization-based tasks and determination of limits of detection (LODs) were performed using a third-generation dual-source CT scanner (SOMATOM Force; Siemens Healthineers, Forchheim, Germany) with the calibration phantom described above. All DECT kVp combinations appropriate for a large adult (80/150Sn, 90/150Sn, and 100/150Sn kVp) were tested in this study. The pitch and rotation times were held constant at 0.6 and 0.5 s, respectively. Tube current was adjusted for each kVp to maintain a volume CT dose index of 25 mGy for each scan and minimize noise in the calibration. The tube currents were 834/417, 598/374, and 515/258 mA for the 80/150Sn, 90/150Sn, and 100/150Sn kVp scans, respectively. Low- and high-kVp images were reconstructed at a 1.5-mm slice thickness with a 1.0-mm interval and a Q40 kernel and reformatted to a 5.0-mm slice thickness for measurement.

Material-Specific Calibration and Determination of Detection Limits

For the calibration phantom, the low- and high-energy images of the calibration phantom were reformatted using manufacturer’s thin-client software (syngo.via, VB30A_HF08; Siemens Healthineers) with an image thickness and interval of 5 mm as per the vendor’s recommendations for quantitative analysis. Regions of interest 7 mm in diameter were placed within each vial on five consecutive images using ImageJ software (National Institutes of Health, Bethesda, MD). The total volume of interest covered more than 80% of the vial, and the mean CT number in the volume of interest was averaged over each of the five image acquisitions as described by Jacobsen et al.²⁶ The mean CT numbers of the volumes of interest were recorded for both the low- and high-energy images.

The dual-energy ratio (DER) for each material was obtained by performing a linear regression analysis of the mean CT number (in Hounsfield units) of the low-energy scan (80, 90, or 100 kVp) against the CT number on the corresponding high-energy scan (150Sn kVp) across all concentration values. The slope of the curve was defined as the DER and was used as the material-specific calibration value for the “iodine ratio” in the virtual unenhanced application, an implementation of a material decomposition algorithm, in syngo.via to create material-specific maps. The software was limited to input values of at least 1.00, so DER values less than this threshold were set equal to 1.00 by default. Whereas this technique generates a material map proportional to concentration, the concentration values generated by the software are not accurate because the algorithm assumes the contrast agent has the physical properties of iodine. Therefore, the same regions of interest used to generate the initial calibrations were used in the material map, and the resulting software-generated material concentrations were recorded and plotted against the nominal concentrations to obtain a correction factor.

To compare the DECT quantification method with SECT, the CT numbers for 80, 90, 100, 150Sn, and 120Sn kVp were individually plotted against the nominal material concentrations, and linear regression analysis was performed. LODs were calculated for both DECT and SECT calibrations using a method described by Jacobsen et al.^32,17,18 The LOD represents the lowest concentration at which a sample containing a material can statistically be differentiated from a sample water across repeated measurements with 95% confidence.

Part II: Application of optimized analysis method on BiNP-IVCF in vivo

Synthesis of BiNPs and Infusion of the BiNP-IVCF

BiNPs were synthesized by mixing 5 mmol bismuth (III) acetate, 30 mL of oleylamine (), 20 mL of oleic acid, and 50 mL of 1-octadecene and heating the mixture at 100°C for 1 h then at 260°C for 30 min under argon gas protection. The mixture was cooled to room temperature before collecting the nanoparticles by adding excess amounts of ethanol and spinning it down at 1000 × g for 15 min. The collected nanoparticles were washed with ethanol twice and then dispersed in anhydrous dichloromethane. The size and morphology of the BiNPs were determined using transmission electron microscopy (JEOL USA, Inc., Peabody, MA).

Absorbable IVCFs based on poly(p-dioxanone) sutures (Adient Medical, Pearland, TX) were manufactured and infused with BiNPs using a method similar to that described previously.^4,18,33–35 Scanning electron microscopy with energy-dispersive X-ray analysis (Nova NanoSEM microscope; FEI, Hillsboro, OR) was performed to determine the morphology and presence of bismuth on the surface of the BiNP-IVCF.

Calibration Phantom

In vitro imaging of the BiNP-IVCF was performed with a Multi-Energy CT Quality Assurance Phantom (Kyoto Kagaku, Kyoto, Japan) measuring 36.3 × 26.2 × 18.0 cm. The phantom contains eight cylindrical plugs (2 cm diameter each) made of solid water material with space to hold 5-mL dram vials (diameter: 1.2 cm). The phantom also contained two bismuth solutions (5 and 10 mg/mL), 5 mg/mL iodine, a BiNP-IVCF, and an uncoated IVCF. The concentration of bismuth in the 5- and 10-mg/mL samples was validated by performing inductively coupled plasma mass spectrometry (ICP-MS; Agilent 7500 ICP-MS, Santa Clara, CA) using standard procedures. A calibration scan such as that described above was performed for bismuth in split-filter CT, with a wide range of bismuth concentrations (0, 5, 10, 20, 30, 50, 100, and 150 mg/mL) placed in the same phantom used for in vitro scanning.

Phantom Scan Protocol

In vitro and in vivo characterization of the BiNP-IVCF was performed using a split-filter DECT scanner (SOMATOM Definition Edge; Siemens Healthineers) along with split-filter system bismuth calibrations. The scan parameters for the in vitro phantom were as follows: 120AuSn kVp (single-source 120-kVp spectrum with a split gold and tin filter), 351 mA tube current, 0.3 pitch, 1.0-s rotation time, and CT dose index 24.8 mGy. Scan parameters were chosen to match the volume CT dose index of the dual-source CT calibration scans.

Acquisition techniques using both CT systems described above provide full-fidelity diagnostic images for low- and high-energy data sets. Individual energy data sets (i.e., low or high energy with added filtration) were used for the SECT evaluations performed in this study. DECT images use both low- and high-energy data sets to generate material-specific images through basis material decomposition as described below.

Material-specific Calibration

All images were reconstructed at a slice thickness of 1.5 mm with an interval of 1.0 mm and a D30f reconstruction kernel. The low- and high-energy images, treated as individual SECT scans, were exported to 3D Slicer (version 4.10.2; Brigham and Women’s Hospital, Boston, MA). Regions of interest 7 mm in diameter were placed along five consecutive images to create a volume of interest in the 5- and 10-mg/mL bismuth samples. The measured mean concentration in the bismuth-specific DECT images was corrected using the measured correction factor from the calibration phantom. Similarly, the SECT (120Sn kVp) CT number was converted to a bismuth concentration using the calibration phantom data for the split-filter system. Subsequently, the measured and corrected concentrations for DECT and SECT were compared with the ICP-MS–measured concentrations in the 5- and 10-mg/mL bismuth samples to validate these calibrations.

Similar volumes of interest were placed in the vials containing the uncoated IVCF and BiNP-IVCF, where the volume included the entire filter with the exception of the metal tip. The maximum value within the volume of interest was measured to minimize the impact on partial volume effect from the surrounding medium within the vial.

Deployment and Imaging of the BiNP-IVCF In Vivo

This animal study was approved by the Institutional Animal Care and Use Committee. The animal was maintained in a facility approved by AAALAC International and in accordance with current U.S. Department of Agriculture regulations and standards.

The BiNP-IVCF was deployed under fluoroscopic guidance in an adult domestic pig (weight: 43 kg) as previously described.^4,18,33 CT imaging of the pig was performed 1 week after deployment of the BiNP-IVCF with precontrast and postcontrast imaging. Postcontrast imaging was performed with iothalamate meglumine (Conray) injection (USP 60%, 252 mg/mL) at a dose of 2 mL/kg in the arterial phase. The scan parameters were as follows: 120AuSn kVp, average tube current of 501 mA, rotation time of 0.33 s, pitch of 0.45, and volume CT dose index of 7.89 mGy. The overall dose was lower for in vivo imaging due to the use of tube current modulation. The pitch and rotation time were adjusted from the settings used during phantom imaging to lower the scan time for imaging during a single breath hold. Material decomposition was performed to generate both bismuth-specific and iodine-specific images using the previously calculated DECT ratio for bismuth and the manufacturer’s default calibration for iodine, respectively. Analysis of bismuth and iodine radiopacity was performed as described above while ensuring that the volume of interest encompassed the entire IVCF.

RESULTS

Part I: Selection of best performing contrast material in DECT and SECT

The relative attenuation for each material in the calibration phantom is shown in Figure 3A, with the slope of each line equal to the measured DER. Materials with relatively low Z values (≤73; e.g., iron, zirconium, barium, gadolinium, ytterbium, tantalum) tend to have decreasing DERs as the energy increases (values ranging from 1.33 to 3.24 for 80/150Sn kVp, 1.00 to 2.74 for 90/150Sn kVp, and 0.70 to 2.37 for 100/150Sn kVp). The highest DER was consistently that of barium (Z = 56) for all energies, with DERs of 3.24, 2.74, and 2.37 at 80/150Sn, 90/150Sn, and 100/150Sn kVp, respectively. In contrast, the DERs for elements with higher Z values (>73; e.g., tungsten, gold, bismuth) exhibited less dependence on the kVp combination and remained relatively constant (values ranging from 0.97 to 1.20 for 80/150Sn kVp, 0.92 to 1.37 for 90/150Sn kVp, and 0.93 to 1.40 kVp for 100/150Sn kVp). In comparison, the DER for iodine is set at 3.46, 3.01, and 2.64 for 80/150Sn, 90/150Sn, and 100/150Sn kVp, respectively.

Figure 3.

Dual energy ratios and calibration curves of the different materials in DECT. (A) Relationship between the low and high X-ray spectra for all materials, with the DER (slope of the linear fit) shown in parentheses. (B) Correction factors based on DECT software-calculated material concentrations using the DER that were generated via linear regression analysis of measured and nominal material concentrations.

DECT-measured concentrations from material decomposition using each material’s DER were plotted against the nominal concentrations of the material to generate the correction factor as shown in Figure 3B. Materials with Z values lower than 73 demonstrated highly linear calibration curves (R², 0.991–0.999), whereas those with Z values of 73 or higher were more likely to be biased and have negative slopes (R², 0.279–0.952), with the exception of tungsten (R², 0.980–0.997).

To circumvent the challenges associated with quantification of high-K-edge materials in DECT, we assessed the utility of high-energy SECT (150Sn kVp) for this purpose. The relative image contrasts of the different materials with increasing kVp are shown in Figure 4A. All materials demonstrated highly linear calibration curves when using SECT (R², 0.993–1.000). When using the low-energy DECT data set as a SECT scan (80, 90, and 100 kVp), gadolinium, ytterbium, and tantalum (Z = 64–73) consistently had the highest CT numbers (HU) per mass concentration. Barium (Z = 56) had high CT number at 80 kVp but had the sharpest and most rapid decline at higher energies. Materials with higher Z values (gold and bismuth; Z = 79–83) had greater CT number at 150Sn kVp, whereas the other materials had stark decreases in image contrast at this energy. This phenomenon is further demonstrated in Figure 4B, in which the CT numbers at the same concentration (10 mg/mL) for all materials at different energies are compared. As the attenuation of other materials declined in the high-energy component (150Sn kVp), the CT numbers of high-Z materials (bismuth and gold) remained steady or increased slightly in attenuation throughout all energies. Furthermore, the LOD for each material when quantified using SECT and DECT is shown in Supplemental Table 1. Materials with lower Z values (≤70) had similar LODs in SECT and DECT, ranging from −0.05 to 1.12 mg/mL and from −0.06 to 1.68 mg/mL, respectively. We noted higher LODs in high-energy SECT than in low-energy SECT for these materials. In contrast, high-Z materials (Z ≥ 73) had higher LODs in DECT (−11.50 to 19.67 mg/mL) than in SECT (−0.16 to 1.24 mg/mL).

Figure 4.

Calibration curves of the different materials in high-energy SECT. (A) Calibrations based on mean CT intensity in SECT for each energy. (B) Mean CT intensity of the different materials (10 mg/mL) for different X-ray spectra.

Part II: Application of optimized analysis method on BiNP-IVCF in vivo

Synthesis and Characterization of BiNPs and Their Infusion in Poly(p-dioxanone) IVCF Sutures

Given the results of the calibration phantom studies, bismuth had the highest contrast at high energies, and so we chose it as the contrast agent for infusion into the IVCF. Synthesized BiNPs were homogenous and uniform in size (mean ± standard deviation, 3.44 nm ± 0.59 nm) as shown in the transmission electron microscopic images in Supplemental Figure 1A. Scanning electron microscopy of the IVCFs showed coating on the surface with BiNPs, which was confirmed by energy-dispersive X-ray analysis (Supplemental Figure 1B).

Calibration of BiNP in phantom

The calibrated DER of bismuth for the split-filter system (120AuSn kVp) was 0.99 while the default iodine DER is 1.47. Material images of 5-mg/mL samples of bismuth (Z = 83) and iodine (Z = 53) obtained using DECT at 120AuSn kVp are shown in Figure 5. The bismuth-specific images show differentiation of bismuth with concurrent suppression of iodine signal. Conversely, inputting the DER for iodine reduced the visibility of bismuth while facilitating enhanced visualization of iodine. Signal enhancement in both DECT bismuth-specific images and SECT demonstrated highly linear trends with bismuth concentration (R² > 0.99) as shown in Figure 6A. The highest bismuth concentration (150 mg/mL) was excluded from analysis due to saturation of the attenuation attributed to the maximum bit depth of the images limiting the maximum CT number as shown in Figure 6B. We compared DECT- and SECT-based quantification with ICP-MS, which is the gold standard for determining the elemental concentration. Quantification using SECT resulted in an error of 7.7% for 5 mg/mL and 11.5% for 10 mg/mL bismuth solution. Comparing with DECT, percent error was calculated to be 253.8% for 5 mg/mL and 139.6% for 10 mg/mL bismuth solution. Thus, SECT quantification was more accurate than DECT.

Figure 5.

Bismuth and iodine differentiation using DECT (120Au/Sn kVp). Shown are software-generated images of the material decomposition of bismuth and iodine in virtual unenhanced mode. Colormap legend: from 0 HU to 200 HU, where the signal is the amount of attenuation in HU attributed to the material.

Figure 6.

Bismuth calibration curves in split-filter CT with asynchronous DECT imaging and quantification of bismuth solutions. (A) Virtual unenhanced scan of different bismuth concentrations in a body phantom (left), with calibrations based on DECT-calculated Bi concentration (middle) and maximum CT intensity for SECT (right). Values for the highest concentration (150 mg Bi/mL) were excluded (marked by *) to obtain linear curves. (B) SECT (left) and DECT (middle) imaging of bismuth solutions in a body phantom shows increasing signal intensity as concentration is increased. Quantification with SECT is shown to be more accurate as bismuth concentrations based on SECT and DECT intensities were compared with Bi concentrations measured with inductively coupled plasma mass spectrometry (ICP-MS) (B; right).

Imaging of BiNP-IVCF in the phantom was performed to determine the increase in radiopacity imparted by BiNP infusion. Figure 7 shows imaging of uncoated IVCF and BiNP-IVCF using SECT and DECT. We observed high contrast with the BiNP-IVCF in high-energy SECT (120Sn kVp), with a maximum CT number of 2028 HU, compared with a maximum CT number of 127 HU for the uncoated IVCF. Additionally, we visually observed material differentiation of bismuth from the background media using bismuth-specific images in DECT with the BiNP-IVCF but not with the uncoated IVCF.

Figure 7.

Signal enhancement and material decomposition of IVCF and BiNP-IVCF. Photographs (top), SECT images (middle), and DECT images (bottom) of the uncoated IVCF (left) and BiNP-IVCF (right). Increased radiopacity can be visualized in the SECT images, and material differentiation can be visualized in the DECT images. PPDO, poly(p-dioxanone).

Imaging of the BiNP-IVCF In Vivo

As shown in Figure 8, we observed increased radiopacity in both high-energy SECT (120Sn kVp) and the DECT blended image (120AuSn kVp) in both noncontrast and contrast scans. Using the iodine-specific DER, iodine can be detected inside the inferior vena cava on a contrast scan at a maximum of 271 HU versus 93 HU in the noncontrast scan. The bismuth-specific images in Figure 8 clearly highlight the BiNP-IVCF in both noncontrast and contrast scans, with concurrent suppression of iodine inside the inferior vena cava evident in the contrast scan.

Figure 8.

Material differentiation using DECT (120Au/Sn kVP) in a pig implanted with the BiNP-IVCF. The bismuth-specific images show the intensity of the BiNP-IVCF with concurrent suppression of iodine signal inside the inferior vena cava.

DISCUSSION

In this study, we explored the use of CT with high-Z elements for material decomposition using DECT and quantification using SECT. In addition, a proof-of-concept imaging of implanted BiNP-IVCF in a swine model showed differentiation of bismuth from iodine intravenous contrast and strong signal enhancement for image-guided implantation and potential long-term monitoring of the medical device.

Material decomposition refers to the digital reconstruction of DECT data to represent fractions of two or more materials as according to their unique attenuation characteristics at different X-ray spectra.^12,19,20 Materials with relatively low Z values (≤73) tend to have decreasing DERs as the kVp of the low-energy spectrum increases (Figure 3A). This is expected because the X-ray attenuation of these materials diminishes at higher kVp settings farther from their K-edges.^12,36 In contrast, elements with higher Z values (>73) have greater K-edge values and do not lose as much X-ray attenuation at higher energies, resulting in DERs that remain stable as the low-energy kVp increases. Large differences in material DER relative to iodine allows for improved material decomposition between high-Z elements and iodinated contrast.

K-edges that fall within the diagnostic energy range cause difficulties in quantification using DECT because the algorithms assume a monotonically decreasing model of photoelectric attenuation.^37,38 In Figure 3B, quantification of high-Z materials (Z≥73) resulted in negative slopes with the exception of tungsten (Z=74) likely because the anode material in an X-ray tube is made of tungsten, minimizing the high K-edge effects. Materials with high K-edges have higher attenuation at higher kVp than at lower ones, resulting in DERs below 1.0. Our investigation of DECT’s utility was somewhat limited by the software disallowing inputs below this threshold. To maintain the improved material attenuation of high-Z materials while enabling accurate quantification in a single scan, we used the high-energy component of DECT, termed SECT with additional tin filtration, to circumvent the adverse effect of a high K-edge on material decomposition.

An element ideal for use as a DECT contrast agent would be one that provides high attenuation per mass concentration and allows for simultaneous imaging of iodine- and barium-based contrast agents, with which image contrast decreases substantially at high kVp settings.¹² When using low kVp (80, 90, and 100 kVp), we found that gadolinium, ytterbium, and tantalum (Z = 64–73) consistently had the highest CT numbers per mass concentration (Figure 4). This is consistent with results of previous studies and is explained by the locations of the materials’ K-edges, which fall well below the peak tube voltage, resulting in high overall attenuation.³⁹ Barium (Z = 56), which has a K-edge closest to that of iodine (33.2 keV), demonstrated a high signal at 80 kVp but had the most rapid decline in contrast at higher energies because the K-edge is near the minimum energy in the CT spectrum. Materials with the highest Z values (gold and bismuth; Z = 79–83) had the highest attenuation in the high-energy component (150Sn kVp) compared to all other materials, which had stark decreases in contrast at this energy. This is because at higher energies, particularly when the beam is hardened by the tin filter, higher numbers of photons are generated at the K-edges. The additional tin filtration present on the tube preferentially filters out the majority of low-energy photons, which are optimal for imaging materials with low K-edges. Thus, the maximum CT intensity at high energies used for quantification is likely to be attributed to high-Z materials even in the presence of iodine or barium when introduced simultaneously.

Of the high-Z materials we tested, we used bismuth over gold in subsequent studies because of its cost-effectiveness and ease of handling. In SECT, CT attenuation using bismuth remained relatively steady across all energies and had the highest attenuation among all materials in the high-energy scans. Bismuth can be differentiated from iodine on high-energy SECT with tin filtration because of the large difference in attenuation between these two materials at these energies. Using DECT material decomposition, bismuth-specific images show clear differentiation of bismuth from iodine through suppression of the iodine signal, whereas iodine-specific images show the inverse (Figure 5). Bismuth had a DER less than 1.0 for multiple kilovoltage combinations, which could not be entered into the vendor software. In the future, we will investigate alternative methods for bismuth quantification for these situations, such as a DER-based method proposed by Lambert et. al.²⁷ The slight difference in measured DER relative to the entered value of 1.0 may result in additional errors in bismuth quantification, but the difference was not large enough in magnitude to substantially affect the task of visual discrimination for this application.

As a proof-of-concept study, we synthesized BiNPs and infused them into a poly(p-dioxanone) IVCF which was implanted in a pig model. Although we performed our initial calibration studies for bismuth using dual-source CT, we used split-filter CT for in vivo scanning because it was the only available CT scanner that could be used for our large animal imaging study. Therefore, we performed a machine-specific calibration of the split-filter scanner and used higher bismuth concentrations than in the previous calibration (Figure 6A). At a bismuth concentration of 150 mg/mL, the signal for both SECT and DECT no longer increased due to signal saturation caused by exceeding the bit depth of the stored images. Validating the proposed quantification process using SECT and DECT with asynchronous scans of two different concentrations of bismuth further confirmed that SECT with additional tin filtration is more accurate than DECT (Figure 6B). Quantitative values obtained via DECT-based bismuth-specific images overestimated the bismuth concentration by up to 250%, whereas values obtained via SECT-based quantification stayed within the standard error of 11.5%.

Infusion of BiNPs in an absorbable poly(p-dioxanone) IVCF markedly increased filter radiopacity, particularly with additional tin filtration of the X-ray spectrum. Additionally, material differentiation was evident on bismuth-specific images obtained using DECT. Bismuth-specific images of the phantom (Figure 7) and pig (Figure 8) showed enhanced visualization of the BiNP-IVCF relative to the uncoated IVCF using both noncontrast and contrast-enhanced scans. Notably, the attenuation of iodine in both bismuth-specific images and scans with tin filtration was diminished, allowing for clear differentiation of bismuth from iodine. Furthermore, using high-energy SECT quantification of bismuth in implanted BiNP-IVCF in pigs showed similar maximum CT number in both noncontrast and contrast-enhanced scans due to the suppression of iodine attenuation in this energy, compared to bismuth which has a higher K-edge. This demonstrated that the maximum CT intensity is attributable to BiNPs and is unaffected by concurrent iodine exposure when tin filtration is used with the X-ray tube. Thus, material decomposition and quantification of high-Z elements in implantable medical devices can be achieved using this optimized postprocessing method. Additionally, use of X-ray filtration results in more accurate quantification when the contrast agent includes a material with a high K-edge.

This optimized method of using high energy SECT and DECT to noninvasively quantify high-Z materials as well as differentiate them from concurrent iodine contrast can be valuable in monitoring a multitude of medical devices, apart from biodegradable IVCF. This is especially true for biodegradable drug delivery devices where quantitative accuracy of multiple contrast materials administered simultaneously may be employed.

This study has several limitations. Firstly, we did not include iodine in our phantom experiments. This is because iodine is well calibrated by default on the system in which the default iodine material decomposition settings on a variety of clinical DECT systems were generally accurate to within 10% of the nominal iodine concentration in phantoms for a wide variety of kVp combinations and CT manufacturers.²⁶ Secondly, we used only one pig and one timepoint in the application of the CT-based quantification in BiNP-IVCF as a proof-of-concept. We are currently doing a long-term multi-animal study to evaluate the utility of CT-based quantification in monitoring the device integrity. Thirdly, we used a larger body phantom (containing solutions in 50-mL conical tubes) for our initial study of the different materials and then shifted to a slightly smaller phantom (using 5-mL vials) to image BiNP-IVCF. We opted to use the smaller phantom because of the small IVCF size since the dram vials for the phantom minimized filter motion within the background media. Lastly, although our previous study shows that dual-source and fast kilovolt-switching DECT scanners generally provided the most accurate results for iodine quantification,²⁶ we were limited to use of a split-filter DECT system for our pig imaging due to its location in our large animal facility. We repeated the calibration in the same manner in both the dual-source and split-filter DECTs using their corresponding phantom and the calibration curves obtained were applied to their respective analyses. Development of imaging agents and devices using photon-counting detector CT (PCD-CT) would be more advantageous as PCD-CT allows users to place bins on either side of the K-edge to accurately quantify the contrast agent concentration, known as K-edge imaging. However, no human PCD-CT systems are commercially available at this time. We are exploring this option for our future studies.

CONCLUSIONS

We showed that high-Z materials are superior to materials with lower Z values when used in imageable, nanoparticle-infused implantable medical devices. Problems in DECT-based quantification of high-Z materials using material decomposition algorithms are solved by quantifying the attenuation on high-energy SECT with additional tin filtration, for which high-Z materials with high K-edges produce the greatest signals in devices, resulting in linear calibration curves and accurate quantification validated by elemental analysis. We applied an optimized DECT material decomposition for bismuth imaging and the SECT quantification method to imaging of a biodegradable BiNP-IVCF in phantom studies and an adult domestic pig. Infusion of BiNPs in IVCF resulted in markedly increased radiopacity and improved material decomposition in the presence of an iodinated contrast agent on both DECT and SECT scans. This work describes an optimized postprocessing method that can potentially be used for noninvasively monitoring implantable medical devices longitudinally and in real time.