Dual-Energy CT: Lower Limits of Iodine Detection and Quantification

Abstract

Background

Assessments of the quantitative limitations among the six commercially available dual-energy (DE) CT acquisition schemes used by major CT manufacturers could aid researchers looking to use iodine quantification as an imaging biomarker.

Purpose

To determine the limits of detection and quantification of DE CT in phantoms by comparing rapid peak kilovoltage switching, dual-source, split-filter, and dual-layer detector systems in six different scanners.

Materials and Methods

Seven 50-mL iohexol solutions were used, with concentrations of 0.03–2.0 mg iodine per milliliter. The solutions and water sample were scanned five times each in two phantoms (small, 20-cm diameter; large, 30 × 40-cm diameter) with six DE CT systems and a total of 10 peak kilovoltage settings or combinations. Iodine maps were created, and the mean iodine signal in each sample was recorded. The limit of blank (LOB) was defined as the upper limit of the 95% confidence interval of the water sample. The limit of detection (LOD) was defined as the concentration with a 95% chance of having a signal above the LOB. The limit of quantification (LOQ) was defined as the lowest concentration where the coefficient of variation was less than 20%.

Results

The LOD range was 0.021–0.26 mg/mL in the small phantom and 0.026–0.55 mg/mL in the large phantom. The LOQ range was 0.07–0.50 mg/mL in the small phantom and 0.20–1.0 mg/mL in the large phantom. The dual-source and rapid peak kilovoltage switching systems had the lowest LODs, and the dual-layer detector systems had the highest LODs.

Conclusion

The iodine limit of detection using dual-energy CT systems varied with scanner and phantom size, but all systems depicted iodine in the small and large phantoms at or below 0.3 and 0.5 mg/mL, respectively, and enabled quantification at concentrations of 0.5 and 1.0 mg/mL, respectively.

© RSNA, 2019

Online supplemental material is available for this article.

See also the editorial by Hindman in this issue.

Summary

The iodine limit of detection using dual-energy CT systems varies with scanner and phantom size, but all systems enable detection of iodine at or below 0.3 and 0.5 mg/mL in the small and large phantoms, respectively.

Key Points

• ■ In a large body phantom, the iodine limits of detection and quantification were found to be at or below 0.5 mg/mL and 1.0 mg/mL, respectively.

• ■ In a small phantom, the iodine limits of detection and quantification were found to be at or below 0.3 mg/mL and 0.5 mg/mL, respectively.

Introduction

Clinical use of dual-energy (DE) CT increased after improvements in material separation (1), quantification (2,3), and workflow (4,5). DE CT provides additional data about materials within the body based on measured changes in x-ray attenuation with photon energy (6,7). Two energy spectra are generated and input into material decomposition algorithms. The low- and high-energy data model tissue as a function of either two or three basis materials consisting of at least one low-atomic-number material and one high-atomic-number material (eg, soft tissue and iodine). There are six commercially available DE CT acquisition schemes (2,7), including dual-source (8), rapid peak kilovoltage switching (9), dual-layered detector (10), split-filter (11), and sequential scanning (12) systems.

Recent studies used DE CT quantitation of iodine concentration to determine diagnostic thresholds between various disease states: benign and malignant lymph nodes for various disease subsets (13–16), papillary and clear cell renal cell carcinomas (17,18), and lung inflammation and metastases (19,20). Before one applies DE CT–derived diagnostic thresholds to patient data, it is crucial to understand the limits of iodine quantitation. It is possible to (a) determine the limit of detection (21) (LOD), which is the iodine concentration above which we can quantitatively assess whether iodine is present, and (b) estimate the limit of quantification (21) (LOQ), which is the iodine concentration above which a measured numeric value is repeatable and accurate. The purpose of this study was to determine the limit of detection and limit of quantification of dual-energy CT in phantoms, comparing rapid peak kilovoltage switching, dual-source, split-filter, and dual-layer detector systems of six different scanners.

Materials and Methods

This phantom study was exempt from institutional review board approval. No industrial support was provided. Authors maintained control of all data and analysis.

Phantom Configuration

We used an elliptical solid water phantom (large phantom, 30 × 40 × 15 cm) (Multi-Energy CT Phantom; Sun Nuclear, Middleton, Wis) with a removable circular insert (ie, small phantom; 20-cm diameter). The large phantom contained 15 removable inserts, while the small phantom contained nine. The inserts around the edge of the small phantom were replaced with 50-mL centrifuge tubes containing iohexol solutions prepared at 2.00, 1.00, 0.50, 0.25, 0.13, 0.06, 0.03, and 0.00 mg of iodine per milliliter (Fig 1). We filled the remaining positions in the large phantom with soft-tissue–equivalent inserts. The iodine concentration within each vial was independently verified in triplicate by using inductively coupled plasma mass spectroscopy (Agilent 7900; Agilent, Santa Clara, Calif).

Figure 1a:

Phantom configuration for the (a) large and (b) small phantoms. Labels in a refer to the nominal iodine concentration present in each sample (in milligrams of iodine per milliliter).

Figure 1b:

Phantom configuration for the (a) large and (b) small phantoms. Labels in a refer to the nominal iodine concentration present in each sample (in milligrams of iodine per milliliter).

CT Scanners and Scanning Protocols

We investigated six dual-energy scanners: GE 750 HD (hereafter, scanner 1) (sles_hd.198; GE Healthcare, Waukesha, Wis), GE Revolution CT (hereafter, scanner 2) (revo_1.5_m3c.120; GE Healthcare), Philips IQon (hereafter, scanner 3) (4.7; Philips Healthcare, Best, the Netherlands), Siemens Definition Flash (hereafter, scanner 4) (VA48A; Siemens Healthineers, Forchheim, Germany), Siemens Force (hereafter, scanner 5) (VA50A; Siemens Healthineers), and Siemens Edge (hereafter, scanner 6) (VA48A; Siemens Healthineers). With scanner 3, two peak kilovoltage settings, 120 and 140 kVp, were tested. Scanners 4 and 5 had multiple peak kilovoltage combinations available. With scanner 4, we used 80/140 Sn kVp and 100/140 Sn kVp (Sn denotes a tin filter on one x-ray tube), and with scanner 5 we used 80/150 Sn kVp, 90/150 Sn kVp, and 100/150 Sn kVp. Scanner 6 inserts a spectral shaping filter comprised of gold and tin along the z-axis at the x-ray tube using a 120-kVp setting. The volumetric CT dose index (CTDI_vol) was 25 mGy, representing the American College of Radiology CT Accreditation Program adult diagnostic reference level (22). Five scans were performed for each phantom (Table 1).

Table 1:

Dual-Energy CT Scanning Parameters Used for Each System and Configuration

Note.—Element symbols after the peak kilovoltage indicate use of that material for additional filtration. Display field of view was constant at 25.0 cm. CTDI_vol = volumetric CT dose index.

Image Processing

We generated DE CT iodine maps using each manufacturer’s software to create images with a 5-mm thickness and interval. Images were generated directly with the GE systems, while thin client software was used with the Siemens and Philips systems (syngo.via, VB20, Siemens Healthineers; Portal, version 4.1, Philips Healthcare). We reconstructed iodine maps directly with a 5-mm section thickness and interval for the Philips and GE systems. For the Siemens systems, we reconstructed low- and high-kilovoltage-peak images with 1.5-mm section thickness and 1.0-mm interval, processed them with virtual non-contrast software, and then reformatted them to a 5-mm image thickness, per the vendor protocol for quantitative imaging.

For each of the five acquisitions per scanner, we placed a cylindrical volume of interest that was approximately 7 mm in diameter to span five consecutive images (25 mm) within each vial in ImageJ software (National Institutes of Health, Bethesda, Md). We recorded the mean of the voxels within the volume of interest and calculated the standard deviation of the mean volume of interest signal over the five acquisitions.

Calibration Curve

For systems with a linear relationship between signal on iodine maps and concentration in samples, a linear least squares fit of all samples was performed (Matlab, r2018b; MathWorks, Natick, Mass) to define a calibration curve of nominal concentration (x-axis) and measured signal (y-axis). For systems with a plateau in measured iodine for low-concentration samples, we defined the calibration curve using only samples that demonstrated a linear signal. The slope of the calibration curve was defined as C_slope, and the y-intercept was defined as C_int.

Limit of Detection

We calculated the LOD for each DE CT system as described by the Clinical and Laboratory Standards Institute (21). First, the limit of blank (LOB)—the highest signal expected from repeated measurements of a blank sample—was calculated for each system in the water sample by using Equation (1):

where LOB_signal is the maximum expected signal level of water (measured in Hounsfield units or milligrams of iodine per milliliter), represents mean water signal, and σ_blank represents standard deviation of the measurements. The multiplicative factor of 1.645 assumes a Gaussian distribution and a false-positive detection rate of 5%.

To calculate the LOD, the standard deviation of a low-concentration sample was required. We chose 0.5 mg of iodine per milliliter because the concentration was expected to be at or above the LOB for all systems. The signal corresponding to the LOD was calculated with Equation (2):

where σ_{low conc} represents the standard deviation of the signal in the 0.5 mg of iodine per milliliter sample for five imaging acquisitions.

By using the aforementioned calibration curve, the LOD can be converted from a signal (in Hounsfield units or milligrams of iodine per milliliter) to the verified concentration value using Equation (3):

Limit of Quantification

The LOQ is defined as a functional metric (23) and indicates the analyte concentration at which measurements meet chosen accuracy or precision criteria. We defined this criterion as a coefficient of variation less than or equal to 20%.

Results

Calibration Curve

The linear equation for the calibration curves is shown in Table 2.

Table 2:

Limit of Detection and Limit of Quantification for All Scanners and Tube Voltage Combinations in Small and Large Phantoms

Note.—Elemental symbols indicate the use of that material as additional filtration.

*Iodine signal is measured in Hounsfield units. Default on other systems is signal in milligrams of iodine per milliliter.

^†Equal to the limit of detection.

Limit of Detection

The range of LODs was 0.021–0.257 mg of iodine per milliliter in the small phantom and 0.026–0.547 mg of iodine per milliliter in the large phantom (Table 2). Except for scanner 5, the LOD was lower in the small phantom (Fig 2a) than in the large phantom (Fig 2b). This is primarily attributable to differences in scatter conditions and beam hardening caused by the size and shape of the phantoms. With scanner 3, the calibration curve used to calculate LOD was fit only to those iodine measurements that demonstrated a linear signal (Fig 3), which resulted in higher calculated LODs relative to other systems (Figs E1–E6 [online]). With scanner 4, 80/140 Sn kVp had a slightly better performance in the small phantom, but results in the large phantom were nearly identical for the 80/140 Sn kVp and 100/140 Sn kVp settings. The optimal LOD in both the small phantom and the large phantom with scanner 5 was 90/150 Sn kVp.

Figure 2a:

Comparison of the measurements performed in the (a) small and (b) large phantoms on scanner 4 at 80 and 140 kVp with tin filtration. Note the lower standard deviation of the blank sample and the tighter distributions of the repetitions of the higher concentrations in the small phantom that lead to a calibration curve with a lower fitting error and, ultimately, a lower limit of detection. The limit of blank (LOB) was defined as the upper limit of the 95% confidence interval of the water sample. The limit of detection (LOD) was defined as the concentration with a 95% chance of having a signal above the LOB. The limit of quantification was defined as the lowest concentration where the coefficient of variation was below 20%.

Figure 2b:

Comparison of the measurements performed in the (a) small and (b) large phantoms on scanner 4 at 80 and 140 kVp with tin filtration. Note the lower standard deviation of the blank sample and the tighter distributions of the repetitions of the higher concentrations in the small phantom that lead to a calibration curve with a lower fitting error and, ultimately, a lower limit of detection. The limit of blank (LOB) was defined as the upper limit of the 95% confidence interval of the water sample. The limit of detection (LOD) was defined as the concentration with a 95% chance of having a signal above the LOB. The limit of quantification was defined as the lowest concentration where the coefficient of variation was below 20%.

Figure 3a:

Comparison of the scanner 3 calibration curves at 140 kVp for the (a) small and (b) large phantoms. Note the plateau in the measured iodine concentration below 0.25 mg/mL in the small and 0.50 mg/mL in the large phantom. The limit of blank (LOB) was defined as the upper limit of the 95% confidence interval of the water sample. The limit of detection (LOD) was defined as the concentration with a 95% chance of having a signal above the LOB. The limit of quantification was defined as the lowest concentration where the coefficient of variation was below 20%.

Figure 3b:

Comparison of the scanner 3 calibration curves at 140 kVp for the (a) small and (b) large phantoms. Note the plateau in the measured iodine concentration below 0.25 mg/mL in the small and 0.50 mg/mL in the large phantom. The limit of blank (LOB) was defined as the upper limit of the 95% confidence interval of the water sample. The limit of detection (LOD) was defined as the concentration with a 95% chance of having a signal above the LOB. The limit of quantification was defined as the lowest concentration where the coefficient of variation was below 20%.

Limit of Quantification

The LOQ for all systems is shown in Table 2. The LOQ in the large phantom was greater than or equal to the LOQ in the small phantom for all systems except scanner 2. Scanner 5 at 80/140 Sn kVp had the lowest LOQ in both the small phantom and the large phantom due to the low standard deviation between successive acquisitions. The LOQ depended primarily on phantom size and did not vary with tube voltage settings on a given system.

Discussion

Assessments of the quantitative limitations among the six commercially available dual-energy (DE) CT acquisition schemes used by major CT manufacturers are largely unavailable, but they could aid researchers looking to use iodine quantification as an imaging biomarker. Thus, we estimated the limit of detection (LOD) for iodinated contrast agents using six DE CT systems from three vendors. All scanners achieved a limit of quantification (LOQ) of 0.50 mg/mL or lower in the small phantom and 1.0 mg/mL or lower in the large phantom. The lower the estimated LOD, the more sensitive a given DE CT system is to the presence of iodine. In low-scatter conditions, we found that the fast peak kilovoltage switching and dual-source CT systems enabled detection of smaller quantities of iodine than did the split-filter and dual-layer detector CT systems.

Literature shows that better spectral separation improves quantitative DE CT results, provided the low-energy spectra are capable of sufficiently penetrating the object being scanned (1). When tested in a large phantom, scanners 2, 4, and 5 had the lowest LODs of the systems. While the dual-source systems both provided consistent LODs and LOQs across kilovoltage pairs, scanner 5 showed improved results in the body phantom. However, our data did not demonstrate a clear trend with spectral separation, and dose partitioning between the low- and high-energy spectra may also affect the results. The LOD of the split-filter system was particularly sensitive to the inclusion of any air outside the phantom. The dual-layer system had an LOD similar to that of the split-filter system due to processing of the iodine image, which demonstrates a clear plateau in the measured iodine concentration below 0.25 mg/mL in the small phantom and 0.50 mg/mL in the large phantom.

Our goal was to determine the limits of each system under ideal circumstances; the small iodine concentrations measured here may not be as apparent in clinical scans. The experimental design used a CTDI_vol of 25 mGy to minimize image noise without adding the additional variable of iterative reconstruction algorithms. This is an acceptable radiation dose for the average adult (22); however, some might consider this relatively high. However, a quantification task is considerably different from a detection one and could justify the use of higher radiation doses for select studies to gain additional diagnostic information about established quantitative imaging biomarkers. Clinically, additional considerations, such as voluntary or involuntary patient motion, pseudoenhancement, and additional beam hardening, would reduce quantitative accuracy. Pseudoenhancement in particular is a problem for small lesions in a highly attenuating background and may be a confounding factor in the interpretation of iodine measurements (24); however, Mileto et al (25) demonstrated that virtual monochromatic images generated from the same DE CT examination may be used to rule out pseudoenhancement. Additionally, contrast agents may have different levels of detectability when the background material has a higher CT number than does water. Therefore, our calculated LOD should be used as a baseline iodine concentration under which reported values may not represent true enhancement. For example, Bahig et al (19) defined pulmonary functional subvolumes based on iodine concentrations as low as 0.2 mg of iodine per milliliter, which is below the LOQ of nearly all DE CT systems in large patients.

Our study had several limitations. This method depends on the standard deviation of the mean signal in the blank vials. Ideally, we would perform up to 20 repetitions of the measurements to ensure accuracy, but we were limited by clinical availability of systems. Second, the large phantom is elliptical rather than round, so beam hardening along the long axis of the phantom could induce bias in the calibration curve. This would be alleviated by tube current modulation, but it is unavailable in DE mode on the fast kilovolt switching systems. Finally, we would ideally sample the concentration domain more finely to more accurately assess the LOQ.

To conclude, the limit of detection (LOD) for dual-energy (DE) CT systems indicates that the presence of iodinated contrast material can be verified for low concentrations of iodine in an ideal setting. For all systems investigated, the limit of quantification (LOQ) was greater than the LOD. Quantitative measurements with current DE CT platforms are likely challenging below 0.5 mg of iodine per milliliter in clinical cases. Larger studies of LOD and LOQ in phantoms across sites are warranted to assess reproducibility. Future studies should investigate the impact of patient and lesion size, total dose, dose partitioning, and material decomposition settings on the LOD and LOQ.