Abstract
Purpose
Weighted average dual-energy computed tomography (DE-CT) reconstructions are considered a proxy of standard CT images of the brain, recommended for routine clinical use and used as a reference standard in DE-CT research. However, their image quality has not been assessed, which was the aim of our study.
Methods
Images from 81 consecutive patients who underwent both non-contrast single-energy (SE)-CT and DE-CT of the brain on the same scanner were retrospectively evaluated. Attenuation values (HU) and SD of grey matter/white matter (GM/WM) pairs, along with SD in the posterior fossa and subcalvarial region were measured. Four readers evaluated image noise, GM/WM contrast, posterior fossa and subcalvarial artefacts, as well as overall image quality.
Results
Weighted average DE-CT GM and WM HU were significantly lower and noise higher compared to SE-CT (GM HU 36.46 v. 41.82; WM HU 28.18 v. 29.94; GM SD 2.93 v. 2.49; and WM SD 3.16 v. 2.44, all p < 0.0001). After correcting the measured SE-CT noise for 37% higher acquisition dose, DE-CT GM noise became significantly lower (2.93 v. 3.11, p = 0.0121). Measured and dose corrected SE-CT GM/WM contrast-to-noise ratio was superior to weighted average DE-CT (3.42 and 2.74 v. 1.95, both p < 0.0001). Weighted average DE-CT had significantly less artifacts on qualitative analysis.
Conclusion
Weighted average DE-CT images of the brain yield less artefacts at 37% dose reduction and lower noise at SE-CT equivalent dose. Dose-adjusted GM/WM contrast-to-noise ratio of weighted average DE-CT with 0.4 weighting factor remains inferior to SE-CT images.
Keywords: Brain, dual-energy computed tomography, artefacts, image quality
Introduction
Dual-energy computed tomography (DE-CT) of the brain is being used for routine evaluation and specific clinical scenarios. DE-CT applications based on material decomposition and virtual monoenergetic imaging have been found to be superior to SE-CT in multiple studies for specific clinical indications such as bone subtraction in CT angiography, differentiation of haemorrhage from iodinated contrast, detection of acute ischemia after intra-arterial recanalization in stroke patients, prediction of haemorrhagic complications after mechanical thrombectomy based on the quantity of iodine contrast leakage, and evaluation of aneurysm or arteriovenous malformation complicated by intracranial hemorrhage.1,2 Weighted average DE-CT reconstructions have been recommended for routine analysis as a surrogate for single-energy computed tomography (SE-CT) images of the brain at 120 kV.3,4 However, to the best of our knowledge, those recommendations are based on the calculated theoretical assumptions only without published confirmatory data. Weighted average images can be automatically reconstructed by blended linear assimilation of low (70 or 80 kV) and high (140 or 150 kV) energy DE-CT data at different weighting factors (WF). Changing the WF influences image-quality metrics, principally image contrast and noise: images at higher WF have the benefit of higher contrast, but with increased noise, whereas with lower WF the noise is significantly reduced at the expense of lower contrast-to-noise ratio (CNR). Images reconstructed with the recommended WF of 0.3 (30% of the image information is derived from low kV data and 70% from high kV data) are assumed to have the quality similar to 120 kV SE-CT images and are routinely used.4
A number of studies have evaluated image quality (IQ) and diagnostic accuracy of virtual monochromatic and virtual non-contrast DE-CT reconstructions and compared them to either weighted average DE-CT or SE-CT images.5–9 To the best of our knowledge, IQ of weighted average DE-CT reconstructions, as the SE-CT surrogate, was assessed in only a single study in paediatric patients. Weinman et al.10 compared quantitative and subjective IQ metrics of SE-CT and weighted average DE-CT images of the brain in a small group of paediatric patients, and found generally improved subjective IQ along with inferior grey matter/white matter (GM/WM) contrast-to-noise ratio of weighted average DE-CT in patients over 6 years of age, at lower radiation exposure. However, they did not compare images of the same patients, the patient population was heterogeneous, and sample size small (27 patients under and 17 patients over 6 years in the SE-CT group; 16 patients under and 15 patients over 6 years in the DE-CT group).
In our opinion, IQ of weighted average DE-CT reconstructions of the brain should be investigated more extensively since they are recommended and routinely used in clinical practice as a SE-CT proxy, and serve as a common reference standard in research of other DE-CT modalities. Therefore, the aim of our study was to compare the IQ of weighted average DE-CT and SE-CT images of the brain in patients who were subjected to both techniques on the same scanner.
Methods
The Institutional Review Board approved this retrospective study and the informed consent was waived.
Patients
Consecutive adult patients who had two non-contrast CT studies of the brain, both a SE-CT and a DE-CT on the same dual source DE-CT scanner (Siemens Somatom Definition Flash, Siemens, Erlangen, Germany) from July 2015 to January 2019 were included in the study by reviewing the Picture Archiving and Communication System (PACS). In order to ensure reliable pairwise comparison, we excluded patients whose brain morphology changed significantly between the initial and follow-up CT studies (due to progression of the existing condition) or whose IQ was substantially different between the two scans, due to severe motion artefacts on only one of the two studies. Furthermore, patients with bilateral brain pathology that precluded image-quality measurements and evaluation due to a significant GM and WM density or volume changes were excluded as well. In cases of unilateral pathology, the non-affected side was assessed.
CT scan protocols
DE-CT tube voltage was 80 kV and 140 kV. SE-CT was acquired with 100 kV or 120 kV, as semi-automatic adjustment of the tube voltage (Care kV®) was activated. Automatic tube current modulation (CareDose® 4D) was activated on both SE-CT and DE-CT studies with referent tube currents set at 396 or 270 mA, and 310 and 145 mA, respectively. SE-CT and DE-CT scanning protocols were the same for all patients, regardless of the clinical indication. Two different energy levels for SE-CT of the brain are the consequence of the protocol changes that were implemented during the time period of our investigation. Detector configuration was 16 x 0.6 mm2 for both studies. All SE-CT and DE-CT reconstructions were generated using Siemens iterative reconstruction software (SAFIRE®) at level 3. Weighted average DE-CT images were reconstructed with a weighting factor of 0.4, as suggested by the vendor. Axial soft tissue (SE-CT J30s and DE-CT J30f reconstruction kernel) 5 mm thick reconstructions were available for evaluation on PACS.
Quantitative analysis
A neuroradiologist (D.D., 8 years of neuroradiology experience) reviewed SE-CT and DE-CT images side by side and measured quantitative image-quality parameters: attenuation values (HU) and SD were measured by placing a fixed round 25 mm2 regions of interest (ROI) at three pairs of identical grey and juxtacortical WM locations – frontal and parietal lobes, thalamus and posterior limb of the internal capsule; posterior fossa and subcalvarial beam hardening artefacts (SCA) were quantified as SD within fixed round 200 and 25 mm2 ROIs at the level of the midbrain with the most artefacts, and near the skull, respectively.8,11 Measurement method is demonstrated in Figure 1. Image-quality index, calculated as the square root of mean SE-CT and DE-CT CT dose index volume (CTDIvol) quotient, was used to correct the SE-CT image noise levels due to a higher radiation dose compared to DE-CT studies.5,8,11 HU and SD measurements at three different GM and WM locations were averaged for comparison and further calculations. GM and WM signal-to-noise ratio (SNR) and GM/WM contrast-to-noise ratio (CNR) were calculated for each individual ROI measurement, and for the averaged values: SNR was calculated as the mean CT number divided by its respective SD, and CNR as the mean GM and WM HU value difference divided by the square root of the sum of their variances.5,8,11–13 Figure of merit (FOM) was calculated as CNR2/CTDIvol, representing dose-normalized CNR and quantifying image-quality improvement per exposure risk to the patient.14
Figure 1.
Circles mark areas of regions of interest for placement of attenuation value and SD measurements at pairs of grey (GM) and white matter (WM) in: (a) the frontal and parietal lobes, thalamus and posterior limb of the internal capsule; (b) the posterior fossa; and (c) the subcalvarial region.
Subjective analysis
Anonymized images were randomly presented to four readers: two neuroradiologists (S.K., 15 years of neuroradiology experience with limited DE-CT experience, and D.D., 8 years of neuroradiology experience with 3 years of DE-CT experience) and two general radiologists (Z.M.K., limited neuroradiology and DE-CT experience, I.Ž., limited neuroradiology and 3 years of DE-CT experience), one in each group with and without experience in reading DE-CT images. Readers blinded to the image type, as well as all patient personal and clinical data, independently evaluated noise level, GM/WM contrast, level of posterior fossa and subcalvarial artefacts, and overall IQ on a four point Likert scale:
1. no or minimal noise or artefacts, excellent GM/WM contrast and overall IQ;
2. some noise and artefacts that do not influence image evaluation, very good GM/WM contrast and overall IQ;
3. noise and artefacts that allow limited evaluation, poor GM/WM contrast and overall IQ; and
4. non-diagnostic images.
The evaluation was performed on the same diagnostic monitor with a fixed window width and level set at 80 and 35 HU. Figure 2 shows an example of SE-CT and DE-CT images of the same patient at the level of basal ganglia, posterior fossa, and subcalvarial regions.
Figure 2.
Example of single (SE-CT) and dual-energy computed tomography (DE-CT) images of the brain in the same patient. Weighted average DE-CT images were assessed by the readers to have: (a) inferior grey to white matter contrast and higher noise at the level of the basal ganglia; less artefacts in the (b) posterior fossa, and (c) subcalvarial regions.
Statistical analysis
Kolmogorov–Smirnov test was used to establish normal data distribution. Normally distributed data were presented as mean ± SD, and nonparametric data as median and interquartile range (IQR). The normally distributed independent continuous variables were compared by t-test, dependent continuous variables by paired t-test, and the paired ordinal data by Wilcoxon signed rank test. Interrater consistency was assessed by a two way intraclass correlation coefficient (ICC) model for multiple readers.15 Guidelines by Cicchetti16 were used for the interpretation of ICC results with the coefficients less than 0.4 indicating poor, from 0.4 to 0.59 fair, from 0.6 to 0.74 good, and above 0.75 excellent interrater agreement. Statistical analysis was done using MedCalc (MedCalc Software, Osten, Belgium) software. Statistical significance level was set at 0.05.
Results
There were 140 patients that had both non-contrast SE-CT and DE-CT of the brain, on average within 14 months (median 414 days, interquartile range 198–661 days). Fifty-nine patients were excluded: 55 had significant bilateral cerebral pathology that prevented measurements and image evaluation (severe bilateral encephalomalacia, confluent white matter hypodensities and/or brain atrophy) and four had severe motion artefacts on one of the two studies. Eighty-one patients were included in the study: 43 (53%) male, 18–87 years of age (median 67, IQR 56–76) at the time of SE-CT exam, and 18–88 years of age (median 69, IQR 58–78 years) at the time of DE-CT exam (p = 0.6). SE-CT/DE-CT scan indications were, respectively: suspected acute stroke (21/27); headache (11/7); transient ischemic attack (7/10); transient global amnesia (3/1); vertigo (5/3); tremor, dementia or cerebral microangiopathy (7/4); altered consciousness and infection (4/2); convulsions (5/6); brain metastases (4/6); trauma (2/9); follow up stroke (2/1); follow up hematoma (2/3); syncope (1/1); hydrocephalus (1/1); tetraplegia (1/0); unknown (5/0). SE-CT/DE-CT imaging findings were respectively: normal (53/49); mild microangiopathy (16/17); acute stroke (4/7); subarachnoid haemorrhage (SAH) (1/1); intra-axial haemorrhage (ICH) (2/2); subdural haematoma (SDH) (1/0); small encephalomalacia (3/3); intra-axial mass (1/1); extra-axial mass (1/1); arachnoid cyst (1/1); and anterior communicating artery (AcomA) aneurysm (0/1). Of the follow-up studies 85% (69/81) were considered unchanged. Five patients developed infarction, two showed resolution of SAH, one resolution of ICH, one resolution of SDH, one underwent elective embolization of AcomA aneurysm, one had intra-axial neoplasm removed and one developed mild microangiopathy. In five patients the pathology (lacunar infarct, SAH, SDH, ICH and intra-axial mass – one each) was present on SE-CT and in seven (four acute infarcts; SAH, microangiopathy, AcomA aneurysm – one each) on DE-CT study.
In total 84% (68/81) of SE-CT studies were acquired with 100 kV and 16% (13/81) with 120 kV. Dose parameters CTDIvol and DLP were significantly lower for DE-CT compared to SE-CT (21.4 ± 2.4 mGy v. 33 ± 5.2 mGy, p < 0.0001; 337 ± 39 mGycm v. 511 ± 83 mGycm, p < 0.0001, respectively). IQ index was 1.25.
Quantitative analysis
Results of the HU and SD measurements in the GM and WM of frontal and parietal lobes, thalamus and posterior limb of the internal capsule are shown in Table 1. Attenuation values in the GM and WM of frontal and parietal lobes on weighted average DE-CT images were significantly lower compared to SE-CT images and no difference was found in the thalamus and posterior limb of the internal capsule. Noise levels were significantly higher on weighted average DE-CT images for all GM and WM measurements.
Table 1.
Results of grey (GM) and white matter (WM) attenuation values and noise measurements in SE-CT and weighted average DE-CT images expressed as mean (SD).
|
HU |
SD |
|||||
|---|---|---|---|---|---|---|
| Variable | SE-CT | DE-CT | p value | SE-CT | DE-CT | p value |
| Frontal GM | 43.09 (3.96) | 35.6 (1.78) | <0.0001a | 2.44 (0.72) | 2.79 (0.63) | <0.0012a |
| Frontal WM | 30.89 (1.99) | 28.38 (1.51) | <0.0001a | 2.37 (0.62) | 2.95 (0.7) | <0.0001a |
| Parietal GM | 45.12 (4.47) | 36.81 (1.95) | <0.0001a | 2.42 (0.67) | 2.76 (0.64) | <0.0007a |
| Parietal WM | 30.83 (3.38) | 28.33 (1.74) | <0.0001a | 2.35 (0.62) | 3.09 (0.66) | <0.0001a |
| Thalamus | 37.25 (2.96) | 36.95 (1.62) | 0.403 | 2.58 (0.61) | 3.25 (0.72) | <0.0001a |
| Posterior limb of IC | 28.1 (1.96) | 27.84 (1.86) | 0.3236 | 2.62 (0.6) | 3.44 (0.74) | <0.0001a |
HU: Hounsfeld units; SE-CT: single-energy computed tomography; DE-CT: dual-energy computed tomography; IC: internal capsule.
Statistically significant.
After correcting the SE-CT noise values for higher acquisition dose compared to DE-CT, the SE-CT noise levels became significantly higher in the GM, while the difference in the WM noise levels was annulled. Measured and dose corrected GM and WM SNR, GW/WM CNR, and FOM were significantly inferior in weighted average DE-CT images. There was no difference in posterior fossa and SCAs. Results are shown in Table 2.
Table 2.
Results of quantitative image quality (IQ) metrics of SE-CT and weighted average DE-CT images expressed as mean (SD).
| IQ index | SE-CT | SE-CT correctedb | DE-CT | p value | p value correctedb |
|---|---|---|---|---|---|
| GM HU | 41.82 (2.96) | 36.46 (1.28) | <0.0001a | ||
| WM HU | 29.94 (1.88) | 28.18 (1.33) | <0.0001a | ||
| GM noise | 2.49 (0.43) | 3.11 (0.53) | 2.93 (0.45) | <0.0001a | 0.0121a |
| WM noise | 2.44 (0.4) | 3.05 (0.5) | 3.16 (0.46) | <0.0001a | 0.1506 |
| GM SNR | 17.25 (2.91) | 13.8 (2.33) | 12.72 (2.05) | <0.0001a | <0.0008a |
| WM SNR | 12.55 (2.06) | 10.04 (1.65) | 9.09 (1.38) | <0.0001a | 0.0001a |
| GM-WM CNR | 3.42 (0.65) | 2.74 (0.52) | 1.95 (0.4) | <0.0001a | <0.0001a |
| FOM | 0.37 (0.13) | 0.24 (0.08) | 0.19 (0.09) | <0.0001a | <0.0005a |
| PFAI | 4.37 (1.3) | 4.53 (0.79) | 0.2511 | ||
| SCA | 3.69 (2.36) | 4.06 (1.93) | 0.1651 |
SE-CT, single-energy computed tomography; DE-CT, dual-energy computed tomography; GM, grey matter; WM, white matter; HU, Hounsfeld units; SNR, signal-to-noise ratio; CNR, contrast-to-noise ratio; FOM, figure of merit; PFAI, posterior fossa artefact index; SCA, subcalvarial artefact.
Statistically significant; b SE-CT IQ parameters corrected for the difference in radiation dose between SE-CT and DE-CT with an image quality index of 1.25 and respective p value.
The subgroup analysis comparing 100 kV SE-CT and 120 kV SE-CT to weighted average DE-CT images was comparable with the results of the whole cohort, with the exception of significantly lower SCA in 100 kV SE-CT compared to weighted average DE-CT images (3.63 ± 2.24 and 4.27 ± 2.08, mean ± SD, p = 0.037, respectively). Results are shown in Table 3. Compared to 100 kV SE-CT, GM/WM CNR was significantly lower in 120 kV SE-CT images (p = 0.0466) and SCA insignificantly higher (p = 0.7055).
Table 3.
Results of quantitative subgroup analysis comparing SE-CT 100 and 120 kV to weighted average DE-CT images.
| IQ index | SE-CT 100 kV | DE-CT | p value | SE-CT 120 kV | DE-CT | p value |
|---|---|---|---|---|---|---|
| GM HU | 42.57 (2.91) | 36.4 (1.36) | <0.0001 | 39.66 (1.97) | 36.7 (1.03) | <0.0001 |
| WM HU | 30.1 (1.91) | 28.23 (1.34) | <0.0001 | 29.43 (1.74) | 28.1 (1.35) | 0.0006 |
| GM noise | 2.53 (0.37) | 2.98 (0.46) | <0.0001 | 2.3 (0.55) | 2.78 (0.4) | 0.0022 |
| WM noise | 2.49 (0.36) | 3.25 (0.44) | <0.0001 | 2.3 (0.48) | 2.92 (0.46) | 0.0005 |
| GM SNR | 17.07 (1.99) | 12.46 (1.89) | <0.0001 | 17.93 (4.63) | 13.53 (2.33) | 0.0003 |
| WM SNR | 12.32 (1.86) | 8.84 (1.25) | <0.0001 | 13.25 (2.51) | 9.83 (1.52) | <0.0001 |
| GM-WM CNR | 3.52 (0.63) | 1.86 (0.35) | <0.0001 | 3.19 (0.64) | 2.18 (0.45) | <0.0001 |
| PFAI | 4.52 (1.34) | 4.59 (0.83) | 0.6663 | 3.9 (1.09) | 4.38 (0.67) | 0.1254 |
| SCA | 3.63 (2.24) | 4.27 (2.08) | 0.037a | 3.86 (2.76) | 3.52 (1.36) | 0.5475 |
IQ, image quality; SE-CT, single-energy computed tomography; DE-CT, dual-energy computed tomography; GM, grey matter; WM, white matter; HU, Hounsfeld units; SNR, signal-to-noise ratio; CNR, contrast-to-noise ratio; PFAI, posterior fossa artefact index; SCA, subcalvarial artefact.
Significant difference of subgroup compared to cohort analysis.
Subjective analysis
The results of subjective analysis and interrater agreement are shown in Table 4. GM/WM contrast and diagnostic IQ of weighted average DE-CT reconstructions were deemed inferior to SE-CT images by experienced DE-CT readers, while the evaluation of the two DE-CT inexperienced readers showed opposite results. Two experienced DE-CT readers estimated noise levels to be higher in weighted average DE-CT images and reader 4 (general radiologist with limited DE-CT experience) showed the same trend, but the results were not statistically significant. All four readers found less posterior fossa and subcalvarial artefacts on weighted average DE-CT images, the results were not statistically significant for subcalvarial artefacts evaluated by reader 2 (general radiologist, experienced in DE-CT). Only the evaluation of subcalvarial artefacts on SE-CT images showed good interrater agreement of 0.69, while the interrater agreement of all other IQ indices was poor to fair, in the range of 0.11 to 0.5.
Table 4.
Results of subjective image quality evaluation and interrater agreement.
|
Reader 1 |
Reader 2 |
Reader 3 |
Reader 4 |
ICC |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SE-CT vsDE-CT | Test statisticZ | p value | Test statisticZ | p value | Test statisticZ | p value | Test statisticZ | p value | SE-CT | 95% CI | DE-CT | 95% CI |
| Noise | –5.97 | <0.0001 | –2.88 | 0.0039 | 45b | 0.1454 | –1.62 | 0.1049 | 0.46 | 0.22–0.63 | 0.27 | 0.02–0.48 |
| GM/WM contrast | –5.77 | <0.0001 | –5.11 | <0.0001 | 2.65a | 0.008 | 30.5b | 0.0507 | 0.33 | 0.09–0.52 | 0.14 | –0.037–0.32 |
| SCA | 6.74a | <0.0001 | 1.33a | 0.1846 | 3.93a | 0.0001 | 4.16a | <0.0001 | 0.69 | 0.55–0.8 | 0.11 | –0.15–0.35 |
| PFAI | 4.95a | <0.0001 | 3.39a | 0.0007 | 5.32a | <0.0001 | 5.6a | <0.0001 | 0.44 | 0.23–0.61 | 0.47 | 0.27–0.63 |
| Overall IQ | –2.19 | 0.0285 | –1.64 | 0.1005 | 4.64a | <0.0001 | 3.7a | 0.0002 | 0.5 | 0.28–0.66 | 0.21 | –0.01–0.42 |
SE-CT, single-energy computed tomography; DE-CT, dual-energy computed tomography; Reader 1, neuroradiologist, experienced in DE-CT; Reader 2, general radiologist, experienced in DE-CT; Reader 3. neuroradiologist, inexperienced in DE-CT; Reader 4, general radiologist, inexperienced in DE-CT; ICC, intraclass correlation coefficient; CI, confidence interval; GM, grey matter; WM, white matter; SCA, subcalvarial artefact; PFAI, posterior fossa artefact index; IQ: image quality.
Wilcoxon signed rank test for paired nonparametric data. Positive Z indicates DE-CT is superior to SE-CT.
Wilcoxon statistic expressed instead of test statistic Z when number of different values is ≤20.
Discussion
Analysis of quantitative and subjective IQ evaluation of weighted average DE-CT and SE-CT images showed fairly consistent results: less posterior fossa and subcalvarial artefacts along with inferior GM/WM contrast and noise on weighted average DE-CT images, at significantly lower dose levels compared to SE-CT.
The posterior fossa and subcalvarial artefacts were deemed less prominent on weighted average DE-CT images by all four readers, while the quantitative analysis did not show significant difference, although there were slightly higher artefact levels on weighted average DE-CT reconstructions. The SD of HU values at certain locations is commonly used to quantify the level of artefacts; however, in our opinion, the subjective estimation of artefact interference with the diagnostic value of the images should take preference because it is clinically more relevant. Moreover, the influence of a severe artefact in a single slice only has smaller clinical impact on image evaluation than a less prominent artefact distributed over many slices. However, the measured SD would be higher within the area of a single slice severe artefact, leading to conflicting results of the subjective and quantitative analysis. The study by Weinman et al.10 found significantly less posterior fossa artefacts in weighted average DE-CT images in children over 6 years of age. In comparison to our results, the mean weighted average DE-CT SD values were almost identical, 4.56 v. 4.57, respectively, and so the discordance was due to a considerable difference in the acquisition parameters of SE-CT and DE-CT studies, with a CTDIvol reduction of 37% (from 33 to 21 mGy respectively) in our study, compared to 11% (from 30.9 to 27.5 mGy, respectively) in the Weinman study.10 These findings indicate that posterior fossa artefacts are improved in weighted average DE-CT images compared to SE-CT, at a significantly lower dose.
Significant dose reduction led to higher measured and perceived noise levels in the GM and WM in weighted average DE-CT images, compared to SE-CT. A dose reduction of 37% in our study was much higher than DE-CT dose reductions in the range of 5 to 30% reported by others.5,10,17,18 Therefore, our protocol could be regarded as a low dose DE-CT exam with CTDIvol of 21 mGy, comparable to the parameters in the study by Hwang et al.7 Their subjective analysis showed that equal image noise, GM/WM contrast and posterior fossa artefacts in regular (CTDIvol 40.7 mGy) and low dose (CTDIvol 28 mGy) DE-CT are achievable in certain virtual monoenergetic (VM) DE-CT reconstructions compared to SE-CT (CTDIvol 45 mGy), and concluded that DE-CT compares favourably with SE-CT even with up to 37% dose reduction.7 However, they did not compare the IQ of weighted average DE-CT images, and pointed out that at least 2 VM reconstructions should be available in routine clinical practice to match the SE-CT IQ for evaluation of supra and infratentorial brain; specifically between 60 and 70 keV for overall image noise and at 80 to 100 keV for the posterior fossa.7 After correcting for the acquisition dose differences, noise levels in weighted average DE-CT images in our study became significantly lower or equal to SE-CT. These results indicate that the noise in weighted average DE-CT reconstructions is superior to SE-CT at equivalent dose to the patient. However, prospects of DE-CT image noise improvements at reasonably lower radiation doses need further investigation, and weighted average or VM reconstructions that provide optimal balance between IQ and dose need to be established.
Significantly lower attenuation values of the GM and WM in weighted average DE-CT reconstructions compared to SE-CT images in our study may be the result of a low weighting factor of 0.4, which retrieves 60% of the data from 140 kV thereby resulting in lower tissue contrast. Contrary to noise levels, the GM/WM SNR and CNR in weighted average DE-CT images, as well as FOM as an indicator of dose-normalized CNR, did not significantly change with dose corrections. Furthermore, comparison of SE-CT images at 100 and 120 kV showed that significantly lower noise and inferior contrast resulted in overall inferior GM/WM CNR with 120 kV. All these results indicate that contrast influences GM/WM CNR more strongly than noise. Our results are in keeping with the study by Weinman et al.,10 in which lower GM/WM CNR was found in weighted average DE-CT images with weighting factor of 0.3 compared to SE-CT in children over 6 years of age. Tijssen et al.18 reported that significantly higher image noise in weighted average DE-CT images, due to a 30% lower CTDIvol, compared to SE-CT, did not deteriorate diagnostic IQ for detection of hyperdense brain lesions. These findings suggest that noise is not a determining IQ metric for DE-CT of the brain and that significant dose reduction is achievable. However, to retain diagnostically acceptable IQ, GM/WM contrast needs to be optimized. Weighting factor of 0.3 is commonly recommended for weighted average DE-CT reconstructions as it is thought to be the equivalent of 120 kV SE-CT images.4 However, several studies have investigated the influence of weighting factor on DE-CT IQ and the findings suggest that the factor of 0.6 should be preffered.4,19 Paul et al.19 evaluated the IQ of postcontrast head and neck DE-CT, which included measurements in the temporal lobes, and found that a weighting factor of 0.6 provides optimal IQ of the postcontrast weighted average DE-CT images of the brain. Furthermore, scanning the brain at 120 kV is merely the best compromise between image noise, tissue contrast and artefacts achievable with single-energy imaging. Thus the approach to image reconstruction should not be translated linearly to dual-energy imaging, whose inherent post-acquisition possibilities allow for independent IQ optimization, which could surpass what is achievable with single-energy. We argue that DE-CT IQ could be further optimized by increasing the weighting factor, while retaining lower dose levels compared to SE-CT.
Subjective image evaluation of the two DE-CT experienced readers was consistent with the results of IQ metrics, the interrater agreement was predominantly poor for DE-CT image evaluation and fair to good for the assessment of SE-CT images. These findings suggest that the adaptation to and experience of a specific imaging modality play an important role.
The main strengths of our study are a large sample size and a selection of patients that underwent both non-contrast SE-CT and DE-CT exam of the brain on the same scanner; however, retrospective and non-randomized study design could have led to some selection bias. There were some additional limitations of our study. Firstly, the SE-CT protocol was not completely consistent within the patient cohort, however the majority (84%) of patients were scanned with 100 kV and the subgroup analysis did not show any deviation from the whole cohort results, except for significantly lower subcalvarial artefacts with 100 kV SE-CT compared to weighted average DE-CT images. Secondly, within the median 14-month time interval between the scans, remarkable changes in brain appearance could have developed and could have influenced the measurements and subjective assessment. However, in only a minority (15%) of the studies unilateral interval changes were noted and were deemed acceptable for evaluation by all readers. Furthermore, 65% of SE-CT and 60% of DE-CT studies were considered normal. Thirdly, the mathematical dose correction is only theoretical and might not reflect the reality, however this method has been repeatedly used in previous studies.5,8,11 Lastly, we compared only images reconstructed from our default DE-CT protocol, at 37% lower dose compared to SE-CT, and did not test other post-acquisition possibilities to optimize weighted average images, nor did we evaluate the diagnostic value of the images for assessment of brain pathology. Nevertheless, we argue that significant reduced artefacts of DE-CT images facilitate posterior fossa and subcalvarial pathology evaluation, particularly extra-axial haemorrhage, and that GM/WM contrast needs to be further optimized to reach the same diagnostic quality as SE-CT for routine analysis. Although general quantitative and subjective IQ level does provide a basic estimate and direction, verification of the method in assessment of pathological changes should be the standard of reference. However, since weighted average DE-CT images are used as a reference standard for objective and clinical IQ assessment of other DE-CT reconstructions, we find the results of our study relevant.
In conclusion, qualitative analysis showed significantly less posterior fossa and subcalvarial artefacts in weighted average DE-CT images at 37% lower dose. Lower noise levels are achievable in weighted average DE-CT images at equivalent SE-CT dose. Dose adjusted GM/WM CNR in weighted average DE-CT images remained inferior to SE-CT at weighting factor of 0.4. Impact of weighting factor on GM/WM CNR and clinical value of weighted average DE-CT images should be further investigated.
Supplemental Material
Supplemental material, sj-pdf-1-neu-10.1177_1971400920920785 for Comparing image quality of single- and dual-energy computed tomography of the brain by Doris Dodig, Slavica Kovačić, Zrinka Matana Kaštelan, Iva Žuža, Filip Benić, Jurković Slaven, Damir Miletić and Zoran Rumboldt in The Neuroradiology Journal
Supplemental material, sj-pdf-2-neu-10.1177_1971400920920785 for Comparing image quality of single- and dual-energy computed tomography of the brain by Doris Dodig, Slavica Kovačić, Zrinka Matana Kaštelan, Iva Žuža, Filip Benić, Jurković Slaven, Damir Miletić and Zoran Rumboldt in The Neuroradiology Journal
Author contribution
All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Dodig Doris, Benić Filip, Kovačić Slavica, Žuža Iva, and Matana Kaštelan Zrinka. The first draft of the manuscript was written by Dodig Doris and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Conflict of interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Ethical approval
All procedures performed in the studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee (Institutional ethical committee, Clinical Hospital Centre Rijeka, Reference number: 003-05/19-1/29) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The informed consent in this retrospective study was waived by the Institutional ethical committee.
Funding
The authors received no financial support for the research, authorship and/or publication of this article.
ORCID iD
Doris Dodig https://orcid.org/0000-0001-9462-4516
References
- 1.Naruto N, Itoh T, Noguchi K. Dual energy computed tomography for the head. Jpn J Radiol 2017; 36: 69–80. [DOI] [PubMed] [Google Scholar]
- 2.Bonatti M, Lombardo F, Zamboni GA, et al. Iodine extravasation quantification on dual-energy CT of the brain performed after mechanical thrombectomy for acute ischemic stroke can predict hemorrhagic complications. AJNR Am J Neuroradiol 2018; 39: 441–447. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Forghani R, De Man B, Gupta R. Dual-energy computed tomography: Physical principles, approaches to scanning, usage, and implementation: Part 2. Neuroimaging Clin N Am 2017; 27: 385–400. [DOI] [PubMed] [Google Scholar]
- 4.Kim KS, Lee JM, Kim SH. Image fusion in dual energy computed tomography for detection of hypervascular liver hepatocellular carcinoma: phantom and preliminary studies. Invest Radiol 2010; 45: 149–157. [DOI] [PubMed] [Google Scholar]
- 5.Park J, Choi YH, Cheon JE, et al. Advanced virtual monochromatic reconstruction of dual-energy unenhanced brain computed tomography in children: Comparison of image quality against standard mono-energetic images and conventional polychromatic computed tomography. Pediatr Radiol 2017; 47: 1648–1658. [DOI] [PubMed] [Google Scholar]
- 6.Kamiya K, Kunimatsu A, Mori H, et al. Preliminary report on virtual monochromatic spectral imaging with fast kVp switching dual energy head CT: Comparable image quality to that of 120-kVp CT without increasing the radiation dose. Jpn J Radiol 2013; 31: 293–298. [DOI] [PubMed] [Google Scholar]
- 7.Hwang WD, Mossa-Basha M, Andre JB, et al. Qualitative comparison of noncontrast head dual-energy computed tomography using rapid voltage switching technique and conventional computed tomography. J Comput Assist Tomogr 2016; 40: 320–325. [DOI] [PubMed] [Google Scholar]
- 8.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: 318–325. [DOI] [PubMed] [Google Scholar]
- 9.Matsumoto K, Jinzaki M, Tanami Y, et al. Virtual monochromatic spectral imaging with fast kilovoltage switching: Improved image quality as compared with that obtained with conventional 120-kVp CT. Radiology 2011; 259: 257–262. [DOI] [PubMed] [Google Scholar]
- 10.Weinman JP, Mirsky DM, Jensen AM, et al. Dual energy head CT to maintain image quality while reducing dose in pediatric patients. Clin Imaging 2019; 55: 83–88. [DOI] [PubMed] [Google Scholar]
- 11.Neuhaus V, Abdullayev N, Grosse Hokamp N, et al. Improvement of image quality in unenhanced dual-layer CT of the head using virtual monoenergetic images compared with polyenergetic single-energy CT. Invest Radiol 2017; 52: 470–476. [DOI] [PubMed] [Google Scholar]
- 12.Bongers MN, Schabel C, Krauss B, et al. Noise-optimized virtual monoenergetic images and iodine maps for the detection of venous thrombosis in second generation dual-energy CT (DECT): An ex vivo phantom study. Euro Radiol 2015; 25: 1655–1664. [DOI] [PubMed] [Google Scholar]
- 13.Mullins ME, Lev MH, Bove P, et al. Comparison of image quality between conventional and low-dose nonenhanced head CT AJNR Am J Neuroradiol 2004; 25: 533–538. [PMC free article] [PubMed] [Google Scholar]
- 14.Wong J, Xu T, Husain A, et al. Effect of area X-ray beam equalization on image quality and dose in digital mammography. Phys Med Biol 2004; 49: 3539–3557. [DOI] [PubMed] [Google Scholar]
- 15.Hallgren KA. Computing inter-rater reliability for observational data: An overview and tutorial. Tutor Quant Methods Psychol 2012; 8: 23–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Cicchetti DV. Guidelines, criteria, and rules of thumb for evaluating normed and standardized assessment instruments in psychology. Psychol Assess 1994; 6: 284–290. [Google Scholar]
- 17.Lin XZ, Miao F, Li JY, et al. High-definition CT gemstone spectral imaging of the brain: Initial results of selecting optimal monochromatic image for beam-hardening artifacts and image noise reduction. J Comput Assist Tomogr 2011; 35: 294–297. [DOI] [PubMed] [Google Scholar]
- 18.Tijssen MP, Hofman PA, Stadler AA, et al. The role of dual energy CT in differentiating between brain haemorrhage and contrast medium after mechanical revascularisation in acute ischaemic stroke. Eur Radiol 2014; 24: 834–840. [DOI] [PubMed] [Google Scholar]
- 19.Paul J, Bauer RW, Maentele W, et al. Image fusion in dual energy computed tomography for detection of various anatomic structures-effect on contrast enhancement, contrast-to-noise ratio, signal-to-noise ratio and image quality. Eur J Radiol 2011; 80: 612–619. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental material, sj-pdf-1-neu-10.1177_1971400920920785 for Comparing image quality of single- and dual-energy computed tomography of the brain by Doris Dodig, Slavica Kovačić, Zrinka Matana Kaštelan, Iva Žuža, Filip Benić, Jurković Slaven, Damir Miletić and Zoran Rumboldt in The Neuroradiology Journal
Supplemental material, sj-pdf-2-neu-10.1177_1971400920920785 for Comparing image quality of single- and dual-energy computed tomography of the brain by Doris Dodig, Slavica Kovačić, Zrinka Matana Kaštelan, Iva Žuža, Filip Benić, Jurković Slaven, Damir Miletić and Zoran Rumboldt in The Neuroradiology Journal