Influence of the X-ray beam quality on the dose increment in CT with iodinated contrast medium

2.1 Principle of the approach

In a previous work, we introduced an analytical method to relate the radiation dose increase in selected organs with the increment in CT numbers (expressed as Hounsfield Units) observed between the native CT scan, and the contrast-enhanced CT scan. [24]

Following the formulation introduced in Ref. [24], the increment in Hounsfield Units (HU) observed in a tissue T up taking an amount of iodine quantifiable with a iodine mass fraction φ_I, is given by: Δ HU = HU TI - HU T = ρ T γ φ I 1 - φ I(1) where γ = μ I ρ I μ H 2 O · 10 3(2)

In Equations 1 and 2, ρ _T and ρ _I are the tissue and iodine densities, respectively, while μ _I and μ _{H 2 O} are the linear attenuation coefficients of iodine and water, respectively.

On the other hand, the average absorbed dose into each organ considered in the anthropomorphic phantoms of the neck or abdomen in presence of a known amount on contrast iodine medium can be evaluated by means of Monte Carlo simulations. It was shown in Ref. [24] that the relative dose increase is in analytical relationship with φ_I, through the function: Δ D D = D TI - D T D T = a φ I b(3) where a and b are fit parameters which, of course, will depend on the geometry of the phantom (i.e. shape, volume and position of the target organ) and on the X-ray spectrum (kilo-Voltage Peak Potential and filtration).

So that the relative dose increase in an organ when loaded with iodine can be evaluated by an in-vivo measurement of the HU increment, ΔHU, exploiting the above formulations and the proper values for the parameters a, b and γ: Δ D D = a ( Δ HU ρ T γ + Δ HU ) b(4)

Here, ΔHU is the difference between CT numbers of a given organ in presence of iodine and those without iodine.

In the present study, we enlarge the range of applicability of the method previously introduced, by characterizing the effects of X-ray beam quality (maximum energy, determined by tube kilo-Voltage Peak Potential, and shape of the X-ray spectrum, influenced by tube filtration) on the experimental (γ) and simulated (a, b) parameters of the model.

A first set of simulations in Geant4 [26–28] was aimed to compute the expected dose increases in the organs of interest, for different X-ray filtrations and maximum energies (kVp), thus enabling us to study the variation of the fit parameters a and b in these conditions.

The computational anthropomorphic phantoms used in this work are those introduced in Ref. [24]: neck is represented as an elliptical tube of soft tissue including a portion of spinal column of bone tissue and the thyroid gland, consisting of two ellipsoids; abdomen phantom is another elliptical tube of soft tissue, containing different ellipsoids representing the various organs: kidneys, spleen, liver and pancreas. Both phantoms are described in full detail in Ref. [24], also with supporting figures.

The X-ray source is represented by a photon source whose spectrum is sampled on the basis of data obtained according to Ref. [29], characterized by a filtration of 2.5 mm Al, rotating at 60 cm from the center of the phantom. Figure 1 shows the X-ray spectra at 90, 100, 120 and 140 kV, calculated according to Ref. [29], with the same value of mAs.

In order to analyze the influence of beam quality on the organ dosimetry in the anthropomorphic phantom, two different X-ray tubes were simulated by introducing proper additional filtrations: 0.025 mm Nb as representative of the Toshiba Aquilion 64 slices, and 1.2 mm Ti as representative of the Philips Brilliance 16 slices.

Four kilo-voltage peak potential levels were adopted for both X-ray tube models, even if the Philips CT system allows a selection of 90-120-140 kV, while the Toshiba system permits 100-120-135 kV.

The Geant4 code used in this study was derived from the one used for the previous works [30, 31] with the same anthropomorphic phantom as in Ref. [24]; the major changes include the increased accuracy of the code (updated to version 10.00.p03) and the choice of different photon source spectra.

The Geant4 low energy Livermore package was adopted for electromagnetic interactions, with a 10-μm cut in range and 10⁶ events per run for all the simulations; in this way, the statistical uncertainty (2σ) associated with the present result was below 1% [32].

2.3 Experimental determination of gamma

Following the methodology introduced in Ref. [24], we conducted a series of measurements of HU increments as a function of the iodine concentration, on several CT scanners and exploiting all the available kVp values of each tomograph. In Table 1, the main technical data of the CT scanners employed are summarized.

For this experimental measurement, six cylindrical vials, 5-cm diameter and 5.1-cm height, giving a volume of 100 ml, containing known dilutions of iodinated contrast medium were prepared, with 0, 1, 2, 5, 7, and 10 mg/ml of concentration χ _I in distilled water.

The six vials were aligned on the bed, along the tomographic axis, without any other phantom, and imaged in a unique CT scan at the various kVp values available. A general purpose helical protocol was employed, characterized by a unitary pitch, a filter size “head” and a beam collimation of 12 mm, corresponding to 4 slices 3 mm each. Beam charge (mAs) was adjusted in order to obtain a standard deviation in HU values lower than 4 HU for each concentration value and for all kVp, consistent with CT quality check specifications in many jurisdictions [33, 34]. A satisfactory value resulted to be 150 mAs.

For each CT acquisition, five ROIs were selected on five contiguous slices to acquire the HU experimental average values using the open source software ImageJ [35]. ROI size was set to 2/3 of the vial diameter, to reduce the statistical errors. The HU values were obtained using FC70 and Standard reconstruction filters, respectively for Toshiba and Philips scanners in order to choose the same operative conditions related to image Quality Assurance procedures.

A preliminary test was conducted in order to verify the stability of HU increment measurements with time between the preparation of iodine contrast dilutions and their CT scans. Stability tests up to two weeks gave positive results.

2.4 Experimental characterization of the attenuation properties of iodinated media

In order to obtain a measurement of the attenuation properties of iodinated contrast media, an experimental set-up was prepared which allowed us to measure doses inside a cylinder containing known amounts of iodinated contrast in water.

In detail, a cylindrical phantom (20 cm of inner diameter) made of PMMA 5-mm thick was provided with an axial cylindrical housing (PMMA 2 mm thick, 10 mm of inner diameter) in which a pencil ionization chamber Wellhofer WDCT-10 was positioned inside the phantom. Picture and scheme of the phantom is provided in Fig. 2.

In this second experimental procedure, dose measurements were carried out by filling alternatively the phantom with 2.3 l of known dilutions of iodinated contrast medium (χ _I  =  0, 1, 2, 5, 7 and 10 mg/ml), using the Toshiba Aquilion CT scanner with an exposure of 450 mAs, by acquiring in axial mode with a collimation of 4×3 mm in each experimental condition. In each measurement, the ionization chamber was carefully positioned at the tomograph isocenter. Phantoms were accurately rinsed with distilled water after each emptying. Furthermore, iodinated solutions were kept as homogeneous as possible, using an ultrasonic mixer for several minutes. The degree of homogeneity was evaluated through a test scan, aimed at identifying potential differences in HU values at various ROI positions in the phantom.

Moreover, each dose value was obtained as an average of five measurements. If D₀ is the dose in distilled water (0 mg/ml) and D _x are the doses in known dilutions of iodinated contrast medium, we obtained the overall error on percentage dose increment: Δ D D 0 = D 0 - D x D 0(5) as a result of the propagation of the statistical errors and taking into account the tolerance in the calibration factor of the ionization chamber. The overall error on dose measurement resulted below the 4%.

The outputs from these experiments are decrements of signal registered from the ionization chamber as a result of an increase of concentration of iodine in the surrounding solution. Iodinated solution acts as a radiation shield with respect to the radiation detector, whose active volume is separated from the volume in which iodine is present, and the increase of energy deposited (and thus in dose) in iodinated solution leads to a decrease of dose registered by the instrument.

The experimental dose decrements observed were compared with those expected from two independent Monte Carlo simulations of the same experimental set up. Both simulations assumed the source filtration of 2.5 mm Al plus 0.025 mm Nb, characteristic of the Toshiba Aquilion CT scanner.

In particular, the geometrical layout of the phantom and pencil chamber was implemented in a Geant4 Monte Carlo simulation, exploiting the same code, physics packages and transport parameters as reported in Sec. 2.2 and, for independent comparison and validation, also in a MCNPX (Monte Carlo N Particle Extended) Monte Carlo simulation [36, 37].

The MCNPX simulation was carried out in ‘p e’ mode, with the default energy cuts of 1 keV both for photons and electrons. Dose values were estimated by using the f6 tally function providing the energy deposition averaged over a cell. The default data libraries, mcplib02 and el03, were adopted for photons and electrons, respectively. MCNPX simulation data show relative errors less than 1%.

MCNPX tally estimations are accompanied by a relative error, R, which is an indication of the acceptance grade of the obtained results. Relative errors less than 0.1 (10%) indicate meaningful, than acceptable, results [37]. R is calculated as R = S x ¯ x ¯(6) where S x ¯ is the square root of the variance of the sample mean. For the Monte Carlo method, it can be demonstrated that R ≈ 1 / N where N is the number of simulated particle histories; this means that a reduction of the relative errors can be achieved by properly enhancing the number of the particle histories. Depending on the probability associated to a physical process, the number of histories can be adjusted thus to obtain relative errors less than 10%.

In these set of simulations, due to the small size and low density of the sensitive volume and especially as a result of the increasing shielding effect exhibited by the contrast medium, the probability of interaction of the photon beam with the scoring volume of air inside the ionization chamber was considerably reduced.

Therefore, we had to increase the number of simulated events, up to 5E+8 events per run, in order to obtain a good counting statistics as well as a satisfactory stability and accuracy of the results.

3.1 Simulations of anthropomorphic phantoms

The relative per cent dose increases, for the four Nb-filtered X-ray sources and for the four kVp values considered, are reported in Fig. 3. For each data series, lines represent the corresponding analytical fits with Equation (3). The fit parameters a and b are reported in Table 2.

From the Monte Carlo simulations of the anthropomorphic phantom we observe only slight variations of the relative dose increments with the kVp value and tube filtration. Referring to Fig. 3, the main differences are found between organ types, with liver exhibiting the lowest relative dose increments and pancreas the highest values, in agreement to the results presented in the previous work [24]. The characteristic dose increment of each organ is dependent upon the specific shape, depth and tissue composition of the phantom model, and it is influenced by the possible shielding effect of all the surrounding tissues, such as the other soft tissue components of the phantoms and also the bone inserts representing the spinal cord.

The fact that X-ray maximum energy has a low impact on the dose increase at a given iodine concentration, can be explained by considering that, when increasing the X-ray energy, the interaction cross section in the iodinated target organ volume decreases, while the X-ray fluency reaching the same volume increases due to the enhancement in the mean free path of photons in the surrounding soft tissue.

As a consequence, organ-specific average values for the fit parameters a and b could be calculated and adopted irrespective of X-ray energy and tube filtration.

3.2 Experimental determination of gamma

The HU increments as a function of the iodine mass fraction were plotted and fitted with Eq. (1), in order to evaluate the experimental values of γ. Figure 4 shows, as an example, the results obtained using a Siemens Definition AS 64 slices CT scanner and the respective analytical fits.

The concordance between data series at different beam collimations is apparent in the data reported in Fig. 4, where the maximum relative per cent differences between the two collimations were below 3%. The obtained γ factors are reported in Table 3, for all CT scanners and every available kVp value. For all fits, the maximum relative per cent error was 2%, as results from the standard deviations on the parameter γ evaluated by means of a scaled Levemberg-Marquardt fit algorithm implemented in QtiPlot software [38].

These measurements demonstrate that, at a given value of iodine concentration, the increment in HU decreases with the X-ray energy (kVp). The effect is observed in all the CT scanners, independently from the collimation adopted.

It can be explained, since the presence of iodine increases mostly the photoelectric cross section of the solution and, for higher X-ray energies, the photoelectric cross section decreases more rapidly with respect to the Compton cross section.

The characterization of a CT scanner requires then the experimental determination of γ at each kVp value through simple measurements of HU increments, exploiting samples of known dilution of contrast medium.

3.3 Experimental characterization of the attenuation properties of iodinated media

The decrease of dose measured by the ionization chamber on the axis of the cylindrical phantom filled with the various dilutions of iodine contrast medium is reported in Table 4, as the absolute value of the relative per cent deviation of dose with respect to the baseline measurement conducted in water.

These measurements can be regarded as a further validation of our Monte Carlo simulations. A comparison with the dose decrements expected from Monte Carlo simulations of the same experimental set up are reported in Table 4 and, as exemplary data relative to 120 kV, in Fig. 5.

Table 5 reports the comparison between the fit parameters a and b of Equation (3), as obtained from the experimental and simulative data series.

In these measurements, a systematic decrease of average absorbed dose is registered by the ionization chamber located along the axis of the cylindrical phantom. Since the iodine solution acts as a shield with respect to the radiation exposing the active volume of the dosimeter, the chamber allows us to obtain an indirect measurement of the dose increase in the iodinated solution through a measurement of the decrease of radiation flux transmitted by the solution.

The maximum relative dose decrement, measured at 10 mg/ml, decrease when the X-ray energy increases, as expected from the enhanced penetration capability of harder X-rays.

These data and relative trends are qualitatively and quantitatively confirmed by the Monte Carlo simulations (Fig. 5), and this result can be regarded as a further validation of our simulative data about anthropomorphic phantoms.