Kory Byrns MD, Girish Fatterpekar MD, Jessica Hu MD,
James Babb PhD, Adam Davis MD
eP-106
Control # 1679

Adam Davis:
 Siemens Medical Product Development.
 Olea Medical Product Development.

All other authors report no disclosures.

Dual Energy CT (DECT) capitalizes on differences in absorption
due to the photoelectric effect to identify and differentiate tissue
types. Two different energy spectra are applied utilizing
simultaneous acquisition with 80kVp and 150kVp via two tubes.

The most common neuroradiology applications include:
 Differentiation of hemorrhage from iodine
 Virtual non-contrast imaging
 Bone subtraction for CT angiography
 Calcification volumetrics and atherosclerotic plaque analysis
 Metal artifact reduction.

This is termed material decomposition and has been the major
focus of DECT research and clinical practice.

Images are produced by blending the output from the two different
energy tubes in a user determined ratio (80kvp vs 150kvp).

Previous work evaluated the subjective advantage of various low
and high energy blending ratios of DECT technique for non-contrast
CT of the head image quality (1). Typically a greater low energy
contribution enhances tissue differentiation while a greater high
energy contribution enhances structural detail (1,2,3).

However, no single ratio has been shown to be uniformly superior
for overall image quality (1). Consequently a balanced ratio of 0.5 is
routinely used at our institution.

Despite widespread use of the technique, direct comparison
between blended DECT acquisition and single energy CT (SECT)
technique for head CT image quality has not been previously
performed.
We investigated whether DECT scans provide subjective
image quality equivalence to traditional SECT technique.

This is a HIPAA compliant institutional review board approved study.

Non-contrast head CT studies performed on the same scanner
(SOMATOM Force, Siemens Medical, Forchheim) at two locations
within the same institution over a two-month period were
prospectively randomized to either SECT (120kVp) or DE (80kVp150kVp) technique

Other acquisition parameters (gantry rotation time, pitch, FOV, CTDI)
were identical.

A total of 572 studies were obtained over two months and reviewed.

100 studies free from artifact or pathology were identified

A total of 40 paired cases (20 with each technique) were compiled.

Balanced low and high-energy kV blend (0.5) images were created
utilizing available post processing software (syngoVia, Siemens
Medical, Erlangen). All images were reconstructed at same slice
thickness (5mm) with no interval spacing.

Yoked pairs of SECT and DECT studies were created with age
difference < 5 years.

Three blinded practicing neuroradiologists (GF, JH, AD) reviewed
each pair utilizing the same PACS workstation (iSite, Philips Medical)
and monitor with identical parameters (window/level 80/35).

Each observer selected which study from each pair was superior for
each of six anatomic criteria as well as overall image quality.

If neither study was felt to be superior, this was recorded.

Criteria were chosen to highlight clinically important areas of tissue
differentiation that traditionally are more difficult to visualize and
which may be strongly influenced by exam technique.
Evaluation Criteria
1 Adequate visualization of the pons
2 Cerebellar gray-white differentiation at the level of the brachium pontis
3 Visualization of the basal ganglia
4 Visualization of the insular cortical ribbon at the level of the insular cistern
5 Gray-white differentiation one axial section above the superior margin of the
lateral ventricles
6 Visualization of the cortex (as distinct from the sulci and calvarium) one axial
section above the superior margin of the lateral ventricles
7 Overall technical quality of the study

Patient age was not significantly different for the two groups.

There was a minimally higher CTDI for DE technique.

Although the protocol design called for equivalent CTDI, difference
is attributable to variation in mAs selected by technologist or
automated exposure.
Single energy
Dual energy
Age
(years)
40.2 ± 12.8
39.3 ± 13.1
CDTI
(mGy)
45.72 ± 0.57
p = 0.66
47.35 ± 0.01
p < 0.001
Table 1: Demographics and Technique
Reader 1
Reader 2
Reader 3
Image Quality Criteria
DE ratio
p-value
DE ratio
p-value
DE ratio
p-value
Visualization of the pons
89.5%
0.001
75.0%
0.077
64.7%
0.332
Cerebellar gray-white differentiation
84.2%
0.004
75.0%
0.077
81.3%
0.021
Visualization of the basal ganglia
72.2%
0.096
57.9%
0.648
78.6%
0.057
Visualization of insular cortex
72.2%
0.096
57.9%
0.648
86.7%
0.007
Cerebral grey-white differentiation
66.7%
0.238
65.0%
0.263
52.6%
1.000
Visualization of cerebral cortex
61.1%
0.481
NA
NA
33.3%
0.687
Overall quality
68.4%
0.167
65.0%
0.263
75.0%
0.077
Table 2: DE ratio
Percentage of pairs in which DECT was chosen
if a superior technique was identified.

DECT was chosen as superior in the combined results of all three
readers more frequently for all criteria except criterion 6
(visualization of cerebral cortex)
 Reader 2 rated all 20 cases as equivalent for this parameter,
reader 3 only chose a superior technique in 6/20 instances.

DECT studies were chosen significantly more frequently for
visualization of the pons by one reader (p = 0.001), the cerebellar
gray-white interface by two readers (p = 0.004, p = 0.021), and the
insular cortex by one reader (p = 0.007).

SECT studies were not statistically significantly superior for any of
the criteria.
Image Quality Criteria
Fleiss Kappa
95% CI
Interpretation
Visualization of the pons
0.3676
0.1729 to 0.5622
Fair agreement
Cerebellar gray-white differentiation
0.6206
0.4293 to 0.8119 Substantial agreement
Visualization of the basal ganglia
0.5501
0.3595 to 0.7407
Moderate agreement
Visualization of insular cortex
0.3672
0.1737 to 0.5607
Fair agreement
Cerebral grey-white differentiation
0.5533
0.3339 to 0.7726
Moderate agreement
Visualization of cerebral cortex
-0.1111
-0.3119 to 0.0897
Poor agreement
Overall quality
0.6000
0.3931 to 0.8069 Moderate agreement
Table 3: Fleiss Kappa Score of Inter-observer Agreement.
Guide to Kappa Score
<0
Poor agreement
0.01 – 0.20
Slight agreement
0.21 – 0.40
Fair agreement
0.41 – 0.60
Moderate agreement
0.61 – 0.80
Substantial agreement
0.81 – 1.00 Almost perfect agreement

The degree of inter-observer agreement was determined by
calculating a Fleiss kappa score for each of the criteria.

Criteria in which DECT was chosen more often demonstrated fair,
moderate, or substantial agreement.

The only criterion with poor agreement (visualization of the cerebral
cortex) was also the only one in which DECT was not chosen more
frequently. This indicates a marked variation between observers for
this one criterion and casts doubt as to the reliability of this result.

Visualization of the pons (criterion 1) is often compromised by artifact
introduced by the dense bone surrounding the posterior fossa

As demonstrated by this example SECT/DECT pair, the latter
technique may provide a more homogeneous rendition of the pontine
parenchyma, a difference that was significantly preferred by one
observer.
SECT
DECT

Similarly, dense bone surrounding the posterior fossa may hinder
differentiation of the gray-white interface of the cerebellum (criterion 2).

This example demonstrates a subtle, but noticeable, subjective
improvement in this distinction with DECT technique

DECT was significantly preferred by two observers—notably, this
criterion demonstrated substantial inter-observer agreement.
SECT
DECT

The ability to confidently identify the basal ganglia and insular cortex
as distinct from neighboring white matter (criteria 3 & 4, respectively)
is clinically important in detecting early, subtle signs of infarct.

Both techniques appear to resolve the basal ganglia well, although
visualization of the insular cortex was found to be significantly superior
by one observer.
SECT
DECT

Confident identification of the cortical ribbon
depends on adequate differentiation from both the
subjacent white matter and surrounding sulci.

CT technique is inherently limited due to beam
hardening effects of the nearby dense calvarium.

Although the preference was not statistically
significant, DECT was chosen more frequently as
superior for cerebral gray-white differentiation
(criterion 5) with moderate inter-observer
agreement.

Conversely, our observers did not indicate any
preference with regard to differentiation of cortex
from sulci and calvarial artifact (criterion 6).
SECT
DECT

Only 20 paired examinations were studied. This may have failed to
provide enough power to disclose more subtle statistically significant
differences that were inherent in the study.

Observers were able to pass on a comparison if they did not feel one
image was superior to the other. The lack of ‘forced choice’ greatly
reduced the statistical power of this study.

Only a balanced ratio (equivalent contribution of the 80kVp and 150kVp
tubes) was studied.

This neglects the inherent advantages of dual energy technique as
compared with single energy technique. Alternative blending ratios that
favor high or low kVp may be superior for tissue differentiation and/or
artifact reduction.

Repeat the current study with increased numbers of patients.

Use forced choice methods (i.e. require the observer to select one
technique as superior).

Use alternative low energy and high energy blending ratios to
investigate the proposed inherent advantages of DECT for tissue
differentiation and anatomic detail as compared with SECT.

DECT non-contrast head CT reconstructed with a balanced 0.5
blending ratio tended to be chosen as superior to SECT for
evaluating certain anatomical structures.

This preference was statistically significant for some readers
evaluating certain criterion, particularly those that necessitated
differentiation of gray and white matter.

Otherwise, there was no significant difference or inferiority when
compared to SECT technique

These results support routine use of DECT technique even for noncontrast exams.

This confers the additional benefit of having dual-energy data
available for further analysis tailored to the particular clinical
situation (e.g. calcium or iodine suppression or artifact reduction).
1.
Mak et al. Optimal 80kVp/140kVp Blend for Dual Source Dual Energy NonContrast CT of the Head. Electronic Exhibit. ASNR 2015, Chicago.
2.
Paul J, Bauer R et al. Image fusion in dual energy CT for detection of
various anatomic structures - Effect on contrast enhancement, contrast-tonoise radio, signal-to-noise ratio and image quality. European J of
Radiology 80 (2011) 612-619.
3.
Tawfik A.M, et al. Image quality and radiation dose of dual energy CT of
the head and neck compared with a standard 120kVp acquisition. AJNR
(2011) 32: 1994-1999.