Liquid tissue surrogates for X‐ray and CT phantom studies

Abstract

Purpose

To develop a simple method for producing liquid‐tissue‐surrogate (LTS) materials that accurately represent human soft tissues in terms of density and X‐ray attenuation coefficient.

Methods and materials

We evaluated hypothetical mixtures of water, glycerol, butanol, methanol, sodium chloride, and potassium nitrate; these mixtures were intended to emulate human adipose, blood, brain, kidney, liver, muscle, pancreas, and skin. We compared the hypothetical densities, effective atomic numbers (Z_eff), and calculated discrete‐energy CT attenuation [Hounsfield Units (HU)] of the proposed materials with those of human tissue elemental composition as specified in International Commission on Radiation Units (ICRU) Report 46. We then physically produced the proposed LTS materials for adipose, liver, and pancreas tissue, and we measured the polyenergetic CT attenuation (also expressed as HU) of these materials within a 32 cm phantom using a 64‐slice clinical CT scanner at 80 kVp, 100 kVp, 120 kVp, and 140 kVp.

Results

The predicted densities, Z_eff, and calculated discrete‐energy CT attenuation of our proposed formulations generally agreed with those of ICRU within < 1% or < 10 HU. For example, the densities of our hypothetical materials agreed precisely with ICRU's reported values and were 0.95 g/mL for adipose tissue, 1.04 g/mL for pancreatic tissue, and 1.06 g/mL for liver tissue; the discrete‐energy CT attenuation at 60 keV of our hypothetical materials (and ICRU‐specified compositions) were −107 HU (−113 HU) for adipose #3, −89 HU (−90 HU) for adipose #2, 56 HU (55 HU) for liver tissue, and 31 HU (31 HU) for pancreatic tissue. The densities of our physically produced materials (compared to ICRU‐specified compositions) were 0.947 g/mL (0.0%) for adipose #2, 1.061 g/mL (+2.0%) for pancreatic tissue, and 1.074 g/mL (+1.3%) for liver tissue. The empirical polyenergetic CT attenuation measurements of our LTS materials (and the discrete‐energy HU of the ICRU compositions at the mean energy of each spectrum) at 80 kVp were −104 HU (−113 HU) for adipose #3, −87 HU (−90 HU) for adipose #2, 59 HU (55 HU) for liver tissue, and 33 HU (31 HU) for pancreatic tissue; at 120 kVp, these were −83 HU (−83 HU) for adipose #3, −68 HU (−63 HU) for adipose #2, 55 HU (52 HU) for liver tissue, and 35 HU (33 HU) for pancreatic tissue.

Conclusion

Our method for formulating tissue surrogates allowed straightforward production of solutions with CT attenuation that closely matched the target tissues' expected CT attenuation values and trends with kVp. The LTSs' inexpensive and widely available constituent chemicals, combined with their liquid state, should enable rapid production and versatile use among different phantom and experiment types. Further study is warranted, such as the inclusion of contrast agents. These liquid tissue surrogates may potentially accelerate development and testing of advanced CT imaging techniques and technologies.

1. Introduction

Currently, several vendors offer phantoms for computed tomography (CT) and X‐ray imaging studies. Some of these phantoms use tissue‐equivalent materials (TEMs), often plastics, that are intended to emulate the energy‐dependent X‐ray attenuation properties of human tissues. Phantom production with plastic TEMs is fairly mature; rather sophisticated phantoms have been developed¹ and several groups have reported on TEMs and methods to characterize these.^{2, 3, 4, 5} Plastic TEMs are convenient in that they can be formed by casting or machining, they are generally stable in shape and composition, and their attenuation can be reasonably well matched to that of human tissues. However, if an experiment calls for custom shapes and/or various compositions such as added contrast agent, costs can be high and lead times can be long. Furthermore, solid‐phase materials cannot be used for dynamic experiments, e.g., perfusion studies. Liquid soft‐tissue surrogates (LTSs), if available, could offer flexibility in shape (i.e. contained within a plastic envelope) and composition (e.g., blends of different tissue materials or a tissue material and a contrast agent). Recently, there has been interest in 3D‐printed phantoms^{6, 7, 8, 9}; as this technology develops, increasingly complex phantoms could be constructed by combining 3D‐printed plastic TEMs with LTSs. For example, one might construct an organ such as liver by 3D printing liver‐equivalent plastic, incorporating a vascular network and perforated parenchyma as voids. A blood‐equivalent LTS could be pumped through the phantom, first without and later with contrast agent added. In this way, one could, in principle, perform dynamic liver phantom imaging studies. To enable these and other possibilities, we sought to develop a simple method to produce LTSs for major organ tissues.

2. Materials and methods

2.A. Selection of constituent materials

To correctly emulate the tissues of interest and provide accurate X‐ray attenuation over the diagnostic imaging energy range, the energy‐dependent linear attenuation coefficient (LAC, μ, cm⁻¹) must be correct; this quantity is equal to the product of the mass attenuation coefficient (MAC, μ/ρ, cm²/g) and the density (ρ, g/cm³). The MAC is determined by the atomic numbers (Zs) and the mass fractions of the individual elements comprising the material; effective Z (Z_eff) can be used to estimate the MAC of a material. We used the equation^{10, 11}

$$Z_{eff}=\sqrt[2.94]{\sum f_i{Z}_i^{2.94}}$$ (1)

where f _i is the electron fraction of an element in the material and Z _i is the atomic number of that element.

The composition of the tissues of interest, taken from ICRU Report 46,¹² is given in Table 1. Elemental composition of major organ tissue. All data are from ICRU Report 46¹² (we used normal adult tissue compositions) except the effective Z (Z_eff), which was calculated from the ICRU data using Eq. (1).

Table 1.

Elemental composition of major organ tissue. All data are from ICRU Report 46¹² except the effective Z (Z_eff), which was calculated from the ICRU data using Eq. (1)

	Density (g/mL)	Elemental mass fractions (%)											Zeff
		H	C	N	O	Na	P	S	Cl	K	Ca	Fe
ICRU adipose #3	0.93	11.6	68.1	0.2	19.8	0.1		0.1	0.1				6.14
ICRU adipose #2	0.95	11.4	59.8	0.7	27.8	0.1		0.1	0.1				6.33
ICRU blood	1.06	10.2	11.0	3.3	74.5	0.1	0.1	0.2	0.3	0.2		0.1	7.53
ICRU brain	1.04	10.7	14.5	2.2	71.2	0.2	0.4	0.2	0.3	0.3			7.45
ICRU kidney	1.05	10.3	13.2	3.0	72.4	0.2	0.2	0.2	0.2	0.2	0.1		7.44
ICRU liver	1.06	10.2	13.9	3.0	71.6	0.2	0.3	0.3	0.2	0.3			7.46
ICRU muscle	1.05	10.2	14.3	3.4	71.0	0.1	0.2	0.3	0.1	0.4			7.45
ICRU pancreas	1.04	10.6	16.9	2.2	69.4	0.2	0.2	0.1	0.2	0.2			7.32
ICRU skin	1.09	10.0	20.4	4.2	64.5	0.2	0.1	0.2	0.3	0.1			7.26
ICRU spleen	1.06	10.3	11.3	3.2	74.1	0.1	0.3	0.2	0.2	0.3			7.48

From these data, we see that although all soft tissues are predominantly composed of H, C, and O, the fraction of H is nearly constant. Therefore, in terms of elemental composition, it is primarily the fraction of carbon (Z = 6) and oxygen (Z = 8) that distinguish soft tissues from each other, with small contributions coming from small variations in the presence of higher‐Z elements (sodium, Z = 11; phosphorus, Z = 15; sulfur, Z = 16; chlorine, Z = 17; potassium, Z = 19; calcium, Z = 20; and iron, Z = 26).

Soft tissues fall into three classes: (a) Adipose tissues, which are carbon‐rich and have a low mineral content, leading to Z_eff less than that of water; also, these are hypodense to water. All other tissues are rich in oxygen, resulting in Z_eff in the range of that of water; also, these are hyperdense to water. As will be shown in the Results section, these nonadipose tissues can be grouped into those that have (b) lower and (c) higher mineral content. The ICRU tissues are plotted in Fig. 1 and summarized in Table 2. Table 2 includes carbon, oxygen, and minerals because their mass fractions vary substantially between materials and this variation dominates the range of Z_eff. For simplicity and clarity, Table 2 excludes hydrogen and nitrogen as the small variations in the mass fractions of these elements have minimal effect on Z_eff. Clearly, the Z_eff of adipose versus other tissues strongly depends on the relative carbon versus oxygen content (i.e., the carbon‐to‐oxygen ratio); therefore, to develop TEMs for these materials this should be a primary, initial consideration.

Figure 1.

Plot of effective Z and density of selected human soft tissues as defined in ICRU Report 46.¹² [Color figure can be viewed at wileyonlinelibrary.com]

Table 2.

Summary of the relevant characteristics of adipose and other tissues

Tissue class	Elemental mass fractions (%)			Zeff	Density (g/mL)
	Carbon	Oxygen	Minerals
Adipose	60–70	20–30	0.3	6.1–6.3	0.93–0.95
Water		89		7.4	1.00
Muscle, blood, organs	10–20	65–75	> 1	7.3–7.5	1.04–1.09

To reproduce the LAC of these tissues using liquids, we hypothesized that if we had a library of water‐soluble compounds with various densities near water and with various carbon‐to‐oxygen ratios, we may be able to mix these compounds to produce reasonable LTSs. We identified glycerin (C₃H₈O₃), butanol (C₄H₁₀O), methanol (CH₄O), and water as candidate liquids. We further hypothesized that small quantities of sodium chloride (NaCl) and potassium nitrate (KNO₃) could be added to approximate the small fraction of slightly higher‐Z elements in the reported ICRU tissue compositions. The relevant characteristics of these compounds are summarized in Table 3.

Table 3.

Summary of the relevant characteristics of LTS constituent compounds

Compound	Elemental mass fractions (%)			Zeff	Density (g/mL)
	Carbon	Oxygen	Minerals
Butanol	64.8	21.6		6.03	0.81
Methanol	37.4	50.0		6.68	0.79
Glycerin	39.1	52.1		6.85	1.26
Water		88.8		7.42	1.00
KNO3		47.5	38.7	14.2	2.11
NaCl			100	15.2	2.17

2.B. Development and evaluation of mixtures as potential LTS materials

We used a two‐step process to develop LTS formulas. First, to quickly and simply develop an initial formula for further, more rigorous evaluation, we developed a spreadsheet to calculate the density and Z_eff of mixtures of the compounds in Table 3 and we used this tool to iteratively evaluate candidate mixtures. Once we found a mixture that agreed well with the corresponding ICRU tissue, we used CatSim software¹³ to calculate the LAC at discrete energies from 0.5 keV to 140 keV in increments of 0.5 keV. We then applied the definition of the Hounsfield Unit (HU) to normalize the calculated LAC for the material to the LAC for water:

$$HU\left(E\right)=1000\times \frac{\mathit{\mu }{\left(E\right)}_{material}-\mathit{\mu }{\left(E\right)}_{water}}{\mathit{\mu }{\left(E\right)}_{water}}$$ (2)

and we compared the LAC of the ICRU‐specified material to that of the proposed LTS material.

We chose four LTS materials to physically formulate and test for agreement in CT attenuation (HU). These materials were chosen to represent the three classes of soft tissue that we had identified: low Z_eff and hypodense to water (adipose #2 and adipose #3), water‐like Z_eff and hyperdense to water with low mineral content (represented by pancreas), and water‐like Z_eff and hyperdense to water with high mineral content (represented by liver). We then physically produced four of these mixtures, measured the density of each, and measured the CT attenuation of each sample using a method previously described in detail¹⁴ (Fig. 2). Briefly summarizing this method, we scanned each sample at four X‐ray tube voltages (80 kVp, 100 kVp, 120 kVp, and 140 kVp) within a 19‐mm‐diameter vial placed in the center of a 32‐cm diameter CT dose index (CTDI) phantom that had the center hole enlarged to accept the vial, and we measured the average CT attenuation (HU) of the samples in a 10‐mm‐diameter region of interest (ROI) over 16 images per material. To reduce the effect of scatter, we used 1‐cm source collimation. For further details about the CT scan and analysis method, we refer the reader to our earlier report.¹⁴ As described in Data S1 of that article, this method robustly produces results with standard error of approximately 2–3 HU. Using CatSim,¹³ we estimated the mean photon energy of the X‐ray spectra produced by this scanner, after attenuation by the bow‐tie filter and the CTDI phantom, at the four tube voltages used. CatSim uses XSPECT¹⁵ for spectrum estimation in terms of photon flux at discrete energies; CatSim conveniently simulates the spectrum after attenuation by the bow‐tie filter and the phantom; the photon flux is then converted to energy flux at each discrete energy, and the mean energy is calculated from the result.

Figure 2.

Apparatus for measurement of CT attenuation of LTS materials. [Color figure can be viewed at wileyonlinelibrary.com]

2.C. Method validation

As a first step in evaluating our physically produced materials, we measured their densities. Measurement of Z_eff and HU(E) is impractical, and empirical validation with CT scans is challenging because there are no reference‐standard ICRU materials to compare with our LTS materials; however, from the CT scans, we evaluated the agreement between observed versus expected CT attenuation for the four physically produced LTS materials. To further validate our overall method, we applied our material evaluation procedure (estimating density, Z_eff, and LAC) to well‐known materials that span the range of the tissues of interest and scanned those materials. For this validation, we chose normal saline (0.9% NaCl in water), polymethylmethacrylate (PMMA, C₅O₂H₈), and canola oil (a well‐known mixture of carbon‐rich acids). The characteristics of these materials are plotted in Fig. 3 and summarized in Table 4. We propose that if good agreement is shown between our estimation of discrete‐energy CT attenuation and empirical measurements of polyenergetic CT attenuation for these well‐defined and measurable materials, then it is valid to apply the same method to evaluate agreement between the well‐defined but unmeasurable ICRU materials and our well‐defined and measurable proposed LTS materials.

Figure 3.

Plot of effective Z and density of selected human soft tissues and validation materials. [Color figure can be viewed at wileyonlinelibrary.com]

Table 4.

Summary of the relevant characteristics of validation compounds

Compound	Elemental mass fractions (%)			Zeff	Density (g/mL)
	Carbon	Oxygen	Minerals
Canola oil	76.7	11.4		5.84	0.92
PMMA	60.0	32.0		6.47	1.18
Normal saline		88.8	0.9	7.56	1.01

3. Results

3.A. Evaluation of ICRU tissues

The CT attenuation of the ICRU materials and water is shown in Fig. 4.

Figure 4.

Plot of calculated discrete‐energy CT attenuation of selected ICRU tissues and water. [Color figure can be viewed at wileyonlinelibrary.com]

3.B. Formulation and evaluation of hypothetical LTS materials

We developed formulas for mixtures as shown in Table 5. The hypothetical elemental composition, density, and Z_eff of these materials is shown in Table 6 and plotted in Fig. 5; their CT attenuation (HU) is detailed in Data S1 and plotted in Fig. 6. For the density and Z_eff values, errors in our hypothetical LTS materials versus ICRU materials were < 1% except for Adipose #3, for which the Z_eff value for the LTS material was 3.3% higher than that of the ICRU tissue. As detailed in Data S1 and shown in Fig. 6, at low energies there were small discrepancies between some ICRU materials and the LTS materials. Above 5 keV, these discrepancies were smaller than 10 HU for all materials except Adipose #2, Adipose #3, and blood. For blood, the disagreement was greater than 10 HU only from approximately 3.5 to 7.0 keV. Adipose #2 agreed within 10 HU above 18 keV. At energies above 40 keV, agreement was within 2 HU for all materials except adipose #3 and brain; above 40 keV, brain agreed within 4 HU. Adipose #3 agreed within 20 HU above 35 keV, 15 HU above 40 keV, and 10 HU above 50 keV.

Table 5.

Formulations for proposed LTS materials

	Compound mass fractions (%)
	Glycerin	Butanol	Methanol	NaCl	KNO3	Water
LTS adipose #3	35.5	64.5
LTS adipose #2	40.5	59.5
LTS blood	26.3		2.9	0.8	1.1	68.9
LTS brain	27.9		10.7	0.9	0.8	59.8
LTS kidney	28.4		7.1	0.7	0.8	63.0
LTS liver	28.0		9.0	0.6	1.2	61.2
LTS muscle	29.0		8.0	0.4	1.3	60.0
LTS pancreas	27.9		16.7	0.5	0.6	54.3
LTS skin	47.6		8.0	0.6	0.4	43.9
LTS spleen	27.5		3.2	0.8	0.8	67.7

Table 6.

Predicted characteristics of proposed LTS materials

	Density (g/mL)	Elemental mass fractions (%)											Zeff
		H	C	N	O	Na	P	S	Cl	K	Ca	Fe
LTS adipose #3	0.93	11.9	55.7		32.4								6.34
LTS adipose #2	0.95	11.6	54.4		34.0								6.38
LTS blood	1.06	10.4	11.4	0.2	76.9	0.3			0.5	0.4			7.53
LTS brain	1.04	10.5	14.9	0.1	73.3	0.4			0.5	0.3			7.45
LTS kidney	1.05	10.4	13.8	0.1	74.7	0.3			0.4	0.3			7.44
LTS liver	1.06	10.4	14.3	0.2	74.0	0.2			0.4	0.5			7.46
LTS muscle	1.05	10.4	14.3	0.2	74.2	0.2			0.2	0.5			7.45
LTS pancreas	1.04	10.6	17.2	0.1	71.4	0.2			0.3	0.2			7.32
LTS skin	1.09	10.0	21.6	0.1	67.5	0.2			0.4	0.2			7.26
LTS spleen	1.06	10.4	12.0	0.1	76.4	0.3			0.5	0.3			7.48

Figure 5.

Plot of effective Z and density of ICRU tissues and hypothetical LTS materials. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 6.

Plot of calculated discrete‐energy CT attenuation of selected ICRU tissues and hypothetical LTS materials. [Color figure can be viewed at wileyonlinelibrary.com]

3.C. Empirical evaluation of proposed LTS materials

The measured densities of our physically produced materials (compared to ICRU's reported values) were 0.947 g/mL (0.0%) for adipose tissue, 1.061 g/mL (+ 2.0%) for pancreatic tissue, and 1.074 g/mL (+ 1.3%) for liver tissue. The results of our measurements of four LTS materials are summarized in Table 7 and plotted in Fig. 7. The curves from Fig. 6 for ICRU‐specified compositions are duplicated in Fig. 7 for comparison with the measurements. The measured CT attenuation values are plotted at the mean energy of each spectrum used, as reported in Table 7.

Table 7.

Calculated discrete‐energy CT attenuation of selected ICRU tissues and the measured CT attenuation of the physically produced LTS materials at the estimated mean energy of the scanner's spectra at the four tube potentials used (80 kVp, 100 kVp, 120 kVp, and 140 kVp)

CT attenuation (HU)
Discrete energies:		60 keV	70 keV	81 keV	93 keV
Calculated attenuation of ICRU‐specified compositions	Adipose #3	−113	−99	−89	−83
	Adipose #2	−90	−77	−68	−63
	Liver	55	54	52	52
	Pancreas	31	32	33	33
Estimated mean energies:		60 keV	70 keV	81 keV	93 keV
CT measurements of physically produced LTS materials	Adipose #3	−104	−93	−86	−83
	Adipose #2	−87	−79	−72	−68
	Liver	59	55	56	55
	Pancreas	33	34	35	35
Difference	Adipose #3	10	5	3	0
	Adipose #2	2	−3	−4	−5
	Liver	3	1	3	3
	Pancreas	1	2	2	2

Figure 7.

Plot of calculated discrete‐energy CT attenuation of selected ICRU tissues (lines) and the measured CT attenuation of the physically produced LTS materials (diamonds). The discrete‐energy attenuation was calculated and plotted in 0.5‐keV increments, whereas the measured attenuation was plotted at the estimated mean energy of the scanner's spectra at the four tube potentials used (80 kVp, 100 kVp, 120 kVp, and 140 kVp). [Color figure can be viewed at wileyonlinelibrary.com]

3.D. Evaluation of validation materials

The calculated discrete‐energy CT attenuation of each validation material is detailed in Data S1, summarized in Table 8, and plotted with the results of the CT measurements, also summarized in Table 8.

Table 8.

Calculated discrete‐energy CT attenuation and measured CT attenuation of the validation materials at the estimated mean energy of the scanner's spectra at the four tube potentials used (80 kVp, 100 kVp, 120 kVp, and 140 kVp)

CT attenuation (HU)
Discrete energies:		60 keV	70 keV	81 keV	93 keV
Calculated attenuation of theoretical compositions	0.9% saline	12	9	7	6
	PMMA	102	117	126	132
	Canola oil	−129	−112	−101	−93
Estimated mean energies:		60 keV	70 keV	81 keV	93 keV
CT measurements of physical materials	0.9% saline	14	13	9	8
	PMMA	102	111	120	124
	Canola oil	−129	−115	−107	−100
Difference	0.9% saline	2	4	2	1
	PMMA	−1	−6	−5	−8
	Canola oil	1	−3	−6	−6

4. Discussion

From Tables 1, 2, and Fig. 1, we can see that soft tissues can be grouped into two broad classes: (a) adipose; and (b) muscle, blood, and organs. Adipose tissues are hypodense to water and have lower Z_eff than that of water, whereas the other tissues are hyperdense to water and have Z_eff in the range of that of water. In combination, these two characteristics result in negative CT attenuation for adipose (approximately −50 to −100 HU in the diagnostic energy range) and positive CT attenuation for all other tissues (approximately + 50 HU), as shown in Fig. 4. Figure 4 also shows that as energy decreases, the CT attenuation for adipose decreases monotonically, because carbon (with its lower Z than oxygen‐rich water) dominates the attenuation at lower energies. This is also true of oxygen‐rich soft tissues that have low mineral content (e.g., skin and pancreas), again due to their carbon component, which is less dominant than adipose but, even at this lower carbon concentration, reduces the low‐energy attenuation of the material when normalized to water. However, when carbon‐rich materials contain a sufficient concentration of minerals, the attenuation of the higher‐Z minerals dominates the low‐energy CT attenuation, and the CT attenuation of these materials increases with decreasing energy.

We were able to develop formulas for hypothetical LTSs (Table 5) that closely matched the densities and Z_eff of ICRU tissues (Fig. 5). The resulting CT attenuation of the LTS materials (Fig. 6) shows the same trends with energy as the ICRU tissues, reflecting good matches to the ICRU tissues' carbon‐to‐oxygen ratios, with the mineral content of the ICRU tissues emulated accurately by the NaCl and KNO₃. We had hypothesized that the four LTS materials that we physically produced would produce similar CT attenuation magnitudes and trends as calculated for the ICRU tissues and hypothetical LTSs. When the results from the CT scanning are superimposed on the plot of calculated CT attenuation (Fig. 7), it is clear that our LTS formulas reasonably emulate the ICRU tissues in both attenuation magnitude (within 10 HU, Table 7) and trend. Likewise, our method is further validated by the excellent agreement (within 8 HU, Table 8) between predicted and measured CT attenuation for our validation materials (Fig. 8). As expected, the canola oil, like adipose tissue, has negative CT attenuation and, due to its high carbon content, exhibits decreasing CT attenuation with decreasing keV and kVp. The PMMA shows the same trend, also due to its high carbon content, but the PMMA shows positive CT attenuation, due to its density, which is 19% higher than water. However, the saline solution, due to its high oxygen and mineral content, shows the opposite trend with energy.

Figure 8.

Plot of calculated discrete‐energy attenuation of validation materials (lines) and the measured CT attenuation of these materials (circles), plotted at the estimated mean energy of the scanner's spectra at the four tube potentials used (80 kVp, 100 kVp, 120 kVp, and 140 kVp). [Color figure can be viewed at wileyonlinelibrary.com]

For diagnostic CT and X‐ray modalities, almost no X rays are detected in the low‐energy range (below approximately 30 keV) because there is typically substantial prepatient filtration and the patient absorbs nearly all flux at low energies. Therefore, small discrepancies between the attenuation of LTS materials and ICRU‐specified compositions at low energies do not compromise the utility of LTS materials for these modalities unless extraordinarily precise tissue equivalence is required. The small discrepancies that we found between ICRU blood and LTS blood were because we are emulating the 0.1% iron in blood using potassium (Z = 19) rather than iron (Z = 26), which is a reasonable approximation except between the k‐edge energies of potassium at 3.6 keV and iron at 7.1 keV and attenuation in this energy range has essentially no influence on CT attenuation. For the adipose tissues, our method does not achieve perfect agreement at low energies because butanol and glycerin, the most carbon‐rich ingredients used in our method, must be blended to have densities like the ICRU adipose tissues and the resulting mixtures contain more oxygen and less carbon than the ICRU adipose tissues. As shown in Table 7, the LTS materials that we tested agreed with ICRU‐specified materials within 10 HU for CT spectra. We did dot evaluate LTS materials for mammography; if contemplating using LTS materials with low‐energy spectra it will be important to consider the intended purpose of the experiment; LTS may be sufficient for such applications unless there is a need for precise tissue equivalence.

We have not evaluated the thermal dependence of these materials. Most clinical CT systems are installed in temperature‐controlled rooms and these systems are typically calibrated to water at the scanner room temperature. We therefore anticipate that LTS materials will be used at the scanner room temperature and thermal dependence will not be a concern.

Air bubbles are always a concern with liquids in phantoms. Of course, water phantoms are used routinely, and experimentalists have learned to manage this issue. In our experience working with LTS materials, we have not found these to be more prone to air bubbles than water.

It is important to appreciate that there is intrinsically good adipose‐to‐organ image contrast at CT. This is because adipose materials are lower in both density and in Z_eff relative to water, whereas the other tissues are hyperdense to water and have similar Z_eff to water. In comparison, CT image contrast between and within muscle, blood, and organs depends on the relatively small variations in Z_eff and density, which leads to the poor inter‐ and intraorgan image contrast at CT. This contrast can be enhanced using contrast agents; current clinical X‐ray/CT contrast agents use iodine (Z = 53) for intravascular contrast agents and barium (Z = 56) for enteric contrast enhancement. For contrast‐enhanced exams, the image contrast between/within organs is therefore determined by the LAC of contrast agent‐containing blood within vessel lumen and/or within blood‐perfused organ parenchyma, or contrast agent‐containing intestinal lumen. Currently, the X‐ray attenuation of contrast‐enhanced tissues follows the LAC of iodine or barium, which decreases monotonically in the energy range used for diagnostic imaging. Therefore, a lower X‐ray tube voltage (kVp) is preferable for all exams using current clinical agents. However, there is interest in exploring new contrast agents based on higher‐Z elements than iodine or barium,¹⁶ especially in the form of nanoparticles,¹⁷ which have a range of biodistribution characteristics.¹⁸ There is particular interest in using high‐Z contrast agents in conjunction with various forms of spectral CT.¹⁹ A method to quickly produce tissue‐emulating phantoms where selected tissues are perfused with defined concentrations of contrast‐enhancing elements might facilitate development and evaluation of both spectral CT imaging systems and contrast agents.

We anticipate that the formulations that we have reported could be readily applied for unenhanced normal tissues or for normal tissues with added contrast agent. An extension of this work might be to develop variations of these formulations that simulate tissues at specific disease states, such as fatty infiltration, edema, calcification, and hemorrhage within a given tissue. Some such tissues are specified in ICRU Report 46; exploration of such materials could be the focus of future studies.

5. Conclusion

We have successfully developed a simple method for producing liquids that are appropriate surrogates for human soft tissues in diagnostic X‐ray/CT imaging experiments; these materials can be easily produced from readily available, common chemicals. Based on ICRU tissue specifications, we have provided formulations for liquid surrogates for 10 common soft tissues: adipose #2, adipose #3, blood, brain, kidney, liver, muscle, pancreas, skin, and spleen. These formulations produce liquids that accurately represent human soft tissues in terms of density and X‐ray attenuation coefficient; when scanned in a clinical CT scanner, these materials produced the correct absolute CT attenuation (HU) and trends with energy. Therefore, these materials are appropriate to evaluate conventional single‐kVp imaging as well as spectral imaging systems. As liquids, these materials lend themselves to experiments using current or proposed clinical X‐ray/CT contrast agents of any type. We anticipate evaluating and reporting the use of these materials in combination with iodinated contrast agent in the future.

Conflict of Interest

The authors declare that they do not have any conflict of interest.

Supporting information

Data S1. Linear attenuation coefficients from CatSim.