Dual-Energy CT: What the Neuroradiologist Should Know
- 1.7k Downloads
Because of the different attenuations of tissues at different energy levels, dual-energy CT offers tissue differentiation and characterization, reduction of artifacts, and remodeling of contrast-to-noise ratio (CNR) and signal-to-noise ratio (SNR), hereby creating new opportunities and insights in CT imaging. The applications for dual-energy imaging in neuroradiology are various and still expanding. Automated bone removal is used in CT angiography and CT venography of the intracranial vessels. Monoenergetic reconstructions can be used in patients with or without metal implants in the brain and spine to reduce artifacts, improve CNR and SNR, or to improve iodine conspicuity. Differentiation of iodine and hemorrhage is used in high-density lesions, after intra-arterial recanalization in stroke patients or after administration of contrast media. Detection of underlying (vascular and non-vascular) pathology and spot sign can be used in patients presenting with (acute) intracranial hemorrhage.
KeywordsDual-energy CT Virtual non-contrast Iodine overlay Circle of Willis Bone removal Intra-arterial recanalization
Blood brain barrier breakdown
Dual-energy CT venography
Digital subtraction angiography
Internal carotid artery
Kilo electron voltage
Peak kilo voltage
Middle cerebral artery
Magnetic resonance imaging
Maximum intensity projection
With the introduction of dual-energy CT (DECT) in neuroradiology, imaging of various important indications shifted back from MRI to CT and new indications for CT scanning were emerged.
For neuroradiologists in today’s practice, MRI is the cornerstone of imaging. It allows direct imaging of the entire brain and its feeding vessels, and grants for functional imaging as perfusion, diffusion, and spectroscopy. Its ability in tissue characterization in the brain is one of the reasons MRI has replaced CT in clinical routine imaging for most indications. However, CT remains the preferred imaging method in the acute setting, e.g., in neurotrauma, in patients presenting with signs of acute hemorrhage and in patients with contra-indications to MR imaging. However, with the introduction of DECT, tissue characterization became feasible and CT indications are again adapting.
Neuroimaging before CT was limited to indirect imaging of the neuraxis. Diagnostic possibilities mostly consisted of conventional X-rays, myelography, pneumoencephalography and arteriography.
Even after the introduction of the first CT-scanner in 1971, due to the long acquisition time, only imaging of the brain was possible. In 1974, technical progress allowed whole-body imaging and in 1975, for the first time, syringomyelia was detected on CT [1, 2].
In the next decades, CT underwent substantial technical improvements, with its most important advancements: helical acquisition, which became feasible due to the implementation of the slip-ring technology and furthermore the introduction of multidetector-row CT (MDCT). In general, MDCT has improved spatial and temporal resolution as well as z-coverage. Isotropic voxels allow multiplanar reformations (MPR) and 3D imaging. Perfusion of the brain and dynamic (“4D”) imaging of the brain are nowadays commercially available and increasingly used in daily routine. Although the concept of dual-energy (DE) scanning was already described in the late 1970s, due to lack of computational power, broad clinical implementation was not possible until the early 2000s [3, 4, 5, 6].
CT scanners with DECT, based on the use of two simultaneous working X-ray tubes (Siemens Healthcare), fast peak kilovoltage switching (GE Healthcare, Toshiba), as well as dual-layer detector systems (Phillips Healthcare) were introduced.
The advantage of DECT is the ability for material characterization and differentiation, based on high- and low-peak voltage acquisitions. Materials with equal Hounsfield densities at 120 kVp imaging can be differentiated by analyzing energy dependent changes of the attenuation of materials (e.g. 80 vs 140 kVp).
The neuroradiological indications for DECT imaging lag behind other indications as abdominal imaging, although the use of DECT in neuroradiological and head neck imaging is increasingly valued and literature is expanding [7, 8, 9, 10, 11, 12, 13•, 14, 15•, 16••, 17, 18].
On conventional CT hemorrhage and iodine are difficult to discriminate due to similar Hounsfield densities. DECT grants characterization and differentiation of materials, like iodine, calcium and hemorrhage [19, 20, 21].
Applications allows bone removal in CT angiography (CTA) and CT venography (CTV) [22, 23, 24, 25, 26, 27, 28, 29, 30]. Iodine can be virtually extracted from the images, resulting in iodine maps (IOM). These can be subtracted from the blended images (synonym: mixed/combined images), resulting in virtual non-enhanced images (synonym: virtual non-contrast, VNC). These images discriminate hemorrhage from iodine after intra-arterial recanalization (IAR) in stroke patients [13•, 15•]. The iodine images are useful in patients presenting with intracranial hemorrhage (ICH) to improve the detection of underlying (vascular and non-vascular) pathology and the spot sign, a sign of active bleeding [16••, 31]. Calculation of monoenergetic images can be helpful in CTA because of increased vascular opacification at lower energies [14, 32]. In patients with metal implants, the use of higher kV at monoenergetic imaging increases the delineation of the metal implants and reduces beam-hardening artifacts [33, 34, 35], e.g., in CT myelography in patients with metal spondylodesis .
Conventional MDCT uses a polychromatic X-ray spectrum, provided by one X-ray tube. DECT uses two different energy spectra, produced by two different kVp settings either by two X-ray sources or by other techniques, as mentioned below. Typical energies used are 80–100 kVp for lower energy spectrum and about 140–150 kVp for the high-energy spectrum. The attenuation of tissues is energy dependent; at the higher energy spectrum, the attenuation of tissues is lower, whereas attenuation is increased at the lower energy spectrum.
DECT can be accomplished with several different techniques, vendor dependent.
Using two almost perpendicular X-ray tubes provides simultaneously two energy spectra with high and low kVp (e.g. 140 and 80 kVp) (Siemens Healthcare). In a one-tube system, DE scanning can be accomplished by fast kV switching, in which the energy of the X-ray tube changes between 80 and 140 kVp in a near-real-time mode (<0.5 ms) (Toshiba, GE Healthcare). In a dual-layer detector system, the detector consists of two scintillation layers, one layer is capable of measuring the intensity of the low-energy photons and the other of the high-energy photons, to acquire DE data from a single polychromatic X-ray source. Another, non- ‘real-time’-option is single source double scan or sequential scanning [36, 37].
The attenuation of materials is based on the Compton scattering and the photoelectric effect. The photoelectric effect, responsible for the largest part of the attenuation, depends on the energy of the X-ray beam and the atomic number (Z) of the tissue and materials. The Compton effect prevails in materials with lower atomic number. In materials with higher atomic number, the attenuation of the X-ray beam is larger, mainly because of the photoelectric effect. When the photon energy exceeds the binding energy of the K-shell electron of the material, attenuation suddenly increases. This effect is called the K-edge.
In material quantification, the concentration of the elements can be determined and quantified.
Iodine has a high atomic number, whereas brain parenchyma, brain hemorrhage, and CSF have low atomic numbers . When two materials with low atomic numbers with different attenuations at a certain energy are put in a diagram with a material with high atomic number (calcium or iodine), the algorithm can decompose a single voxel in a mixture of these three elements and the contribution of the high-density material (calcium or iodine) can be calculated [17, 21], this being called three-material decomposition. The density of the voxel and the amount of the high atomic number substance can be reconstructed. Thus, it is possible to calculate a blended image, a calcium or iodine map. By subtracting calcium or iodine from the blended image, a virtual non-calcium or virtual non-enhanced image (VNC) is generated.
In virtual monoenergetic or monochromatic imaging, images from a DECT dataset are reconstructed at different virtual monoenergies. These monoenergetic images reflect the HU values of the scanned region at one specific energy, opposed to the HU values derived at a polychromatic spectrum. Calculation of these images is based on material decomposition and the specific mass attenuation of these materials, respectively .
Recently, an advanced method for calculation of monoenergetic images was published, in which the advantages of high CNR at low keV are combined with the high SNR at higher keV using a frequency split principle . The low keV and the high keV images are decomposed into two sets of subimages. The first set contains only lower spatial frequencies and thus object information, the second set contains the remaining high spatial frequencies, image noise. By combining the lower frequency range from the low kV with the higher frequency range of the higher kV, the benefits of both result in better CNR at higher SNR .
Monochromatic images have the potential to facilitate quantitative measurements and reduce beam hardening artifacts, in particular artifacts from metallic structures .
Material Characterization and Differentiation
Material characterization and differentiation is one of the most important advances of DECT compared to standard MDCT in neuroradiology and is used in various indications.
Bone Removal CTA
The numbers of CTA over the past decade increased, and CTA of the intracranial and extracranial vessels largely replaced digital subtraction angiography (DSA) for diagnostics.
There are various ways of looking at CTA images: axial source images, MPR, maximum intensity projections (MIP), and volume rendering (VR). Especially in MIP and VR, the bone can hamper interpretation of the vessels. Previously, various methods have been used to remove the bone of skull and skull base to facilitate interpretation of the vessels [22, 42, 43].
Commonly applied methods of bone removal are
Manual segmentation of bone and vessels. This method is time consuming and operator dependent.
Semiautomatic segmentation by (vendor specific) threshold-based software based on density differences of structures. Disadvantages are possible subtraction artifacts when HU of vessels, calcified plaques and bone are very similar.
Subtraction segmentation, subtracting the non-enhanced CT (NECT) from the contrast enhanced CT (CECT). Even though this method is not threshold dependent, radiation dose is increased and vulnerable for motion artifacts between the two scans [22, 42, 43].
Several authors published the application of DECT-based bone removal in the intracranial vessels. Sometimes comparing DECTA with gold-standard DSA, sometimes to conventional CTA, or with bone-removed CTA [22, 24, 25, 27, 44].
Watanabe et al. compared bone-removed DECTA and CTA with the gold-standard (DSA) in a small patient group and concluded that bone removal in DECTA was good in most cases. Visualization of aneurysms next to the skull base and calcified aneurysms was superior to conventional CTA. DECTA was superior to conventional CTA for the detection and quantification of internal carotid artery (ICA) stenosis, mainly attributed to the presence of calcifications of the intracavernous and paraclinoid region of the ICA . However, bone-removed DECTA lead to an overestimation of the degree of stenosis compared to DSA, due to the blooming effect of the calcifications.
Zhang et al. compared DECTA with digital subtraction CTA and 3D DSA for the detection of intracranial aneurysms in 46 and 80 patients, respectively [30, 44]. They stated that image quality of DECTA was not different from digital subtraction CTA, but at a lower radiation dose. Diagnostic accuracy of DECTA, compared to DSA, was not different for the detection of intracranial aneurysms, visualization of aneurysm morphology, and prediction of treatment (coiling vs surgery) . DECTA showed a sensitivity of 97 % and a specificity of 100 % for aneurysm detection on a per-patient basis, compared to DSA.
Morhard et al. compared conventional automated bone removal with DE-based bone removal. They found that vessel delineation was better with DECT, resulting in faster reading times and a dose reduction of 29–43 % .
Buerke et al. compared DECTA with bone removal to time-of-flight (TOF)-MRA and standard CTA; they concluded that standard CTA and TOF-MRA were comparable, but DECTA overestimated stenosis quantification in the C2 segments, in close relationship to the bony structures. In all other segments, all three methods were similar .
Korn et al. investigated the grading of carotid artery stenosis in the presence of extensive calcifications by comparing DECTA and CE-MRA with the gold standard (DSA). DECTA-MPR had slightly better agreement in measuring the degree of stenosis in ACI than CE-MRA .
Bone Removal CTA and CTV
In conclusion, DECTA is a powerful tool for evaluation of the intracranial vessels and presence of vascular malformations like aneurysms. DECTA is equal or superior to CTA for the detection of aneurysms and is superior to CTA for the estimation of vessel integrity in the presence of calcifications.
However, DECTA tends to overestimate the degree of stenosis in comparison to DSA when calcifications, particularly close to the skull base, are present [22, 27]. Blooming artifacts caused by calcifications, especially at the skull base can induce artifacts due to excessive removal of bone at these locations or cause ill-defined borders. This underlines the necessity (as with all post-processing methods) to check the source images carefully.
Characterization of Iodine and Calculation of Virtual Non-enhanced Images and Iodine Maps
After Intravenous or Intra-arterial Contrast
Morhard et al. found that hyperdensities at CT immediately after IAR was a frequent finding in 48 of their 60 patients [13•]. In 10.4 % (5/48), this was present on VNC and was rated as hemorrhage; in 89.6 % (43/48), this was only present on iodine images and was related to blood brain barrier breakdown (BBBB). They encountered no false positive or negative ICH compared to follow-up CT and/or MRI. They stated that DECT allows for a reliable differentiation between BBBB and ICH.
Gupta et al. studied 15 patients with high density lesions at CT after administration of intravenous or intra-arterial iodinated CM, 12 of these were scanned after IAR. They found a sensitivity of 100 % with a specificity of 91 % for the identification of hemorrhage, with an accuracy of 93 % .
Phan et al. used DECT in a group of 40 patients after contrast administration. Indications were evaluation of carotid stenosis, trauma, tumors, and after IAR. Of 148 hyperdensities on CT, DECT correctly classified 142 lesions as hemorrhage or iodine. Sensitivity was 100 %, specificity 84–100 %, and accuracy 87–100 %, depending on the location of the density (parenchymal, subarachnoid, or intraventricular). Radiation dose in their study was comparable with a conventional CT (CTDI 66 mGy). The iodine map and VNC derived from DECT were noisier than mixed images but were rated sufficient for diagnosis. In their study, calcifications (as in the pineal gland) were a confounder and could not be discriminated from iodine. The latter is confirmed by Dinkel et al. who concluded that the presence of a fourth material cannot be depicted by the three-material decomposition algorithm and will be misclassified [54•].
Brockman et al. used DECT after peri-interventional subarachnoid hemorrhage SAH. Allowing for discrimination and subtraction of blood and iodine mixed within the subarachnoid space in patients with peri-interventional re-SAH. They stated that overestimation of SAH is reduced after peri-interventional rebleeding and is therefore valuable .
Intracranial Hemorrhage: Identification Vascular Malformation, Tumor, and Spot Sign
Watanabe et al. performed a retrospective study in 36 patients presenting with acute intracranial hematomas [16••]. They performed a NECT and a DECTA and compared the latter with standard CT images with delayed enhancement. The patient population consisted of 16 idiopathic ICH, 16 traumatic ICH, 2 patients with SAH combined with ICH, and two patients with intra-tumor hemorrhage. They evaluated the combined images of DECTA, simulating 120 kVp and iodine images with late enhancement standard CT images. In 57, delayed enhancement in the hemorrhage was seen. Using the iodine images compared to combined imaging, sensitivity, negative, and PPVs of contrast leakage or enhancement increased with constant specificity (95.7/82.6 %; 95.7/95 %; 94.1/80 %; and 94.1/94.1 %, respectively). The authors found three lesions in the iodine images that could not be detected on combined imaging alone, and in four patients, enhancement of lesions was better depicted on the iodine images, compared to combined imaging. They concluded that DECT emphasized iodine enhancement and facilitated detection of contrast enhancement or leakage.
Kim et al. evaluated 56 patients, presenting with intracerebral hematoma with DECT to differentiate primary ICH from tumor bleeding . They evaluated the accuracy of mixed images, mixed images combined with true non-enhanced (TNC) images and fusion images in detection of underlying tumors. Of 18 tumorous lesions, they identified 17 with fusion imaging, 12 on mixed images combined with TNC, and 11 on mixed images alone. They encountered 1 false positive with fusion images, which turned out to be a cavernoma. They concluded that the diagnostic performance of fusion imaging was higher than that of mixed imaging, whether or not combined with a TNC. Sensitivity and specificity of fusion images were 94.4 and 97.4 %, respectively. There was a significant difference with mixed imaging (sens. 61.1 % and spec. 92.3 %; p = 0.006) and with mixed imaging/TNC (sens. 66.7 %; spec. 89.7 %; p = 0.011). The image quality of TNC images was superior to VNC images, but the latter had lower noise in the measurements. CNR was higher in VNC images compared to TNC. Although MRI is known to be superior in detection of underlying tumor in ICH, this is known to be difficult in the acute stage. DECT can play an important role in the acute and subacute stage, thereby enabling early diagnosis .
Virtual Monoenergetic Imaging/Reconstructions
When scanned in DE mode, monoenergetic or monochromatic images can be reconstructed. As in polychromatic CT with increasing (tube) voltage, the SNR increases while tissue contrast decreases. Lowering the tube voltage will increase the CNR while the SNR decreases. DECT takes advantage of both [63, 64, 65].
Increased Attenuation of Iodine
In cerebral CTA, virtual monoenergetic imaging can increase vascular enhancement by increasing CNR at lower virtual monoenergetic images [66, 67] (Fig. 6). The increase of vascular enhancement provides opportunities to lower the amount of iodine [32, 68]. Schneider et al. determined the optimal settings in cerebral and cervical DECTA with monoenergetic imaging . They found an optimum vessel attenuation and CNR for cerebral CTA at 60 keV when compared to 120 kVp polychromatic imaging.
Optimized Brain Imaging
Shinohara et al. used monoenergetic imaging in CTA to reduce artifacts surrounding platinum coils. They used high keV and additional metal artifacts reduction software. At increasing keV, the SNR increased and metallic artifacts were reduced by reducing beam-hardening artifacts .
Next to brain imaging, DECT can be useful in spinal imaging, especially when metallic implants are present. In patients with metallic implants, like in spinal fusion implants, imaging of the spine and spinal canal can be challenging with CT as well as MRI . It is applicable to identify complications of the hardware, like malpositioning, loosening, and disruption. In patients with contra-indications for MRI or inability to perform MRI due to susceptibility artifacts, CT myelography is an alternative for spine imaging. The presence of metal leads to beam-hardening and streak artifacts due to photon starvation. Possible solutions to improve image quality are using high kVp, reduced voxel size, oversampling, and high definition scanning. Increasing the kVp reduces beam-hardening artifacts, even though at the price of increased radiation dose. Iterative reconstruction techniques improve image quality by reducing artifacts, as well as dose reduction .
Srinivasan described the use of DECT virtual monochromatic series for assessment of spinal transpedicular hardware-bone interface in 92 patients . They concluded that monochromatic images at 110 keV or higher were beneficial in the reduction of metallic streak artifacts and enabled better visualization of the hardware-bone interface than 70 keV (comparable with 120 kVp standard CT) series in patients with spinal transpedicular screw fixation. Guggenberger et al. evaluated the optimal monoenergetic DECT settings for artifact reduction in posterior spine implants from different vendors in a phantom . They demonstrated that monoenergetic DECT images provided better image quality and less metallic artifacts than standard CT. They showed that image quality was superior at higher monoenergies of DECT with artifact reduction at higher energies, with an optimum at 120–140 keV, dependent on vendor and spine level. They concluded that keV values in monoenergetic reconstructions should be individually tailored dependent on brand of implants and spine-level characteristics .
Wang et al. evaluated monochromatic images in pedicle screws in patients with scoliosis and concluded that at higher monochromatic photon energy image quality was increased compared to polychromatic imaging, and more accurate measurements of the width of the screws were obtained. They found an optimum range for reducing metal artifacts at 110–140 keV .
DECT myelography can be a solution in patients with spinal instrumentation . Grams et al. demonstrated that DECT myelography after lumbar osteosynthesis with DECT provided minimal artifact disturbance by reducing beam-hardening artifacts using the tube voltage of 140 kVp, compared to 80 and 120 kVp blended images. Virtual myelography was superior to conventional myelography in their group of 30 patients.
DECT myelography can be an opportunity to reduce dose compared to standard CT. The use of monoenergetic reconstructions can minimize beam-hardening artifacts, and application of material decomposition offers the possibility of virtual myelography .
Future Developments and Challenges
DECT in neuroradiology is catching up with other radiological subspecialties. The number of neuroradiological publications is steadily increasing and its strength is recognized. This encourages further refinement of DECT algorithms and applications in brain imaging and asks for solution of the limitations in DECT, as the confounding fourth element in three-material decomposition.
Recently, advanced monoenergetic reconstructions were introduced, combining the advantages of low and high keV reconstructions using only a part of the frequency spectrum of both. This has the potential to put monoenergetic reconstructions to a higher quality level.
The use of iterative reconstructions in DECT can improve SNR, especially in non-full dose VNC and IOM.
DE scanning in neuroradiology does clearly have an added value. Bone removal, material characterization, and monoenergetic reconstructions allow for improved detection of aneurysms, vascular malformations, and iodine enhancement all within one scan. Monoenergetic reconstructions help improving image quality of the brain and spine and/or improved iodine attenuation.
When material characterization is desired, DECT offers more information at the same image quality while maintaining dose neutrality.
Recently published papers of particular interest have been highlighted as: • Of importance •• Of major importance
- 13.• Morhard D, Ertl L, Gerdsmeier-Petz W, Ertl-Wagner B, Schulte-Altedorneburg G. Dual-energy CT immediately after endovascular stroke intervention: prognostic implications. Cardiovasc Intervent Radiol. 2014;37:1171–178. Evaluated the incidence of ICH and BBBB immediately after endovascular recanalization therapy. They found no correlation between BBBB and infarct size. No false positive and false negative ICH were found when DECT was used for differentiation of ICH and BBBB.Google Scholar
- 15.• Tijssen MP, Hofman PA, Stadler AA et al. The role of dual energy CT in differentiating between brain hemorrhage and contrast medium after mechanical revascularisation in acute ischaemic stroke. 2014;Eur Radiol. 24:834–40. Evaluated DECT in patients after IAR. They concluded that diagnostic accuracy and confidence increased in early differentiation between hemorrhage and contrast for DECT, even at lower dose than conventional CT. Google Scholar
- 16.•• Watanabe Y, Tsukabe A, Kunitomi Y et al. Dual-energy CT for detection of contrast enhancement or leakage within high-density haematomas in patients with intracranial hemorrhage. Neuroradiology. 2014;56:291–95. Compared DECT with delayed enhanced conventional CT for the presence of contrast enhancement in intracranial hematomas. They found that 57.5 % of hematomas demonstrated contrast enhancement or extravasation. They found a sensitivity, specificity, PPV and NPV of 82.6, 94.1, 95.80 % for the combination of non-contrast CT and combined CT. This increased to 95.7, 94.1, 95.7 and 94.1 % when non-contrast CT and combined CT were combined with iodine images.Google Scholar
- 32.Cho ES, Chung TS, Oh DK, et al. Cerebral computed tomography angiography using a low tube voltage (80 kVp) and a moderate concentration of iodine contrast material: a quantitative and qualitative comparison with conventional computed tomography angiography. Invest Radiol. 2012;47:142–7.CrossRefPubMedGoogle Scholar
- 41.Grant KL, Flohr TG, Krauss B, Sedlmair M, Thomas C, Schmidt B. Assessment of an advanced image-based technique to calculate virtual monoenergetic computed tomographic images from a dual-energy examination to improve contrast-to-noise ratio in examinations using iodinated contrast media. Invest Radiol. 2014;49:586–92.CrossRefPubMedGoogle Scholar
- 43.Romijn M, Gratama van Andel HA, van Walderveen MA, et al. Diagnostic accuracy of CT angiography with matched mask bone elimination for detection of intracranial aneurysms: comparison with digital subtraction angiography and 3D rotational angiography. AJNR Am J Neuroradiol. 2008;29:134–9.CrossRefPubMedGoogle Scholar
- 48.Fiorelli M, Bastianello S, von Kummer R, et al. Hemorrhagic transformation within 36 hours of a cerebral infarct: relationships with early clinical deterioration and 3-month outcome in the European Cooperative Acute Stroke Study I (ECASS I) cohort. Stroke. 1999;30:2280–4.CrossRefPubMedGoogle Scholar
- 54.• Dinkel J, Khalilzadeh O, Phan CM et al. Technical limitations of dual-energy CT in neuroradiology: 30-month institutional experience and review of literature. J Neurointerv Surg. 2014. doi: 10.1136/neurintsurg-2014-011241. Evaluated the failure modes and limitations of DECT in neuroimaging applications. Ninety percent of DECT analysis was successful, but beam-hardening artifacts, the presence of a fourth material or motion can impair DECT.
- 69.•• Pomerantz SR, Kamalian S, Zhang D et al. Virtual monochromatic reconstruction of dual-energy unenhanced head CT at 65-75 keV maximizes image quality compared with conventional polychromatic CT. Radiology. 2013;266:318–25. The authors evaluated monochromatic reconstructions from unenhanced brain DECT to define optimum levels of SNR and CNR in the supratentorial brain and posterior fossa and compared them to conventional CT.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.