Download Image Quality Recovery in Cardiac CT Angiography

Image Quality R e cove ry i n Ca rd i a c CT A n gi ogra p h y :
How to edit an ECG trace to improve image quality
Barbara L. Kruger, R.T.(R)(CT), Fremont D. Lee, R.T.(R), Natalie N. Braun, B.S., Michael R. Bruesewitz, R.T.(R), Andrew N. Primak, Ph.D., Eric E. Williamson, M.D., Cynthia H. McCollough, Ph.D.
CT Clinical Innovation Center, Department of Radiology, Mayo Clinic, Rochester, MN
Selecting the “Best” Reconstruction Phase
Background
• Cardiac CT imaging offers a robust, non-invasive
Figure 1: The patient’s ECG is recorded by the scanner for use in retrospectively-gated
method of visualizing the coronary arteries, as well
image reconstructions. The important information conveyed to the operator is labeled
as the morphology and function of the heart.
here.
• Patient preparation is vital in order to avoid image
Bullet Point Showing
Heart Rate (bpm)
Relative Phase (%) or
artifacts.
Detected R-wave
for each R-R Interval
Absolute Phase (ms)
Reconstructed
- Proper placement and connection of the ECG leads
Phase
are needed to provide a stable, low-noise ECG signal. On the evaluated system (SOMATOM
Definition, Siemens Medical Solutions), several
important pieces of information are conveyed to
the operator using the ECG data strip, as shown in
Figure 1.
- Prior to scanning, it is essential to coach the patient
regarding correct breathing techniques and the
Purple Region: Scanner
White Region: Scanner
Operating at Prescribed
Operating at Reduced
Scan Time
Pink Line:
need to refrain from moving or swallowing during
Tube Current
Tube Current
in Seconds
X-ray is On
the scan.
• Artifacts can be classified into two categories: cardiac
artifacts and non-cardiac artifacts. Displacement artifacts occur when contiguous sections of anatomy are displaced with
respect to each other, making the tissue appear discontinuous.
- Cardiac sources of artifacts include heart rate variability or arrhythmias and errors in the detection of the R-wave due to
a poor quality ECG signal.
- Non-cardiac sources of artifacts include patient breathing, patient movement (voluntary and involuntary), and implanted metal medical devices.
Cardiac Displacement Artifacts
Variations in patient heart rate due to arrhythmia or ectopic beats can cause displacement artifacts. Additionally, missed or
extraneous R-waves (as detected by the scanner) can result in an incorrect phase being used for image reconstruction. This
principle is illustrated in Figure 2.
Figure 2: (a) In cardiac CT, different sections of anatomy along the Z axis are imaged during successive cardiac cycles (represented by the colored bars in each of the
four R-R intervals shown). Displacement artifacts occur when one R-R interval is shorter or longer than adjacent intervals (cycle with green bar). The reconstructed data
is then at a different phase and discontiguous with the adjacent anatomy. (b) Patient exam demonstrating this principle.
A
B
Manual selection of the best phase for reconstruction is a cumbersome process.
Retrospective ECG-gated image reconstruction allows the reconstruction of images at
any point in the cardiac cycle, providing the
opportunity to seek out and use the phase
with the least motion. This is referred to as the
“best phase”. However, the best phase can
vary from patient to patient, even for similar
heart rates. Even for low heart rates, occasionally a systolic phase (e.g. 30-40%) yields better
image quality than a diastolic phase (e.g. 6080%), although most often end-diastole is better for heart rates < 70 bpm and end-systole is
better for heart rates > 70 bpm. Thus, the
operator must determine for each patient
which is the optimal reconstruction phase.
Further, it is not uncommon that the right
coronary artery looks better at a different
phase (typically earlier) than the left anterior
descending and left circumflex coronary arteries, due to the complex motion of the heart.
Adding New Syncs
Figure 6: Examples of high quality ECG strips where the heart rate was
relatively constant and the R-waves easily identified. (a) Mean heart rate of
51 bpm. The reconstruction phase is most appropriate in the quiescent
end-diastolic phase. (b) Mean heart rate of 109 bpm. The reconstruction
phase is most appropriate at end-systole due to the decreased length of
diastole. Note that 3 R-waves were detected by the system (blue arrows)
and marked (blue bullets) slightly earlier than the actual R-wave (red
arrows). If image artifacts are apparent, these bullets should be manually
moved to over the respective R-waves. The apparent heart rate variation
would be decreased by proper placement of the R-wave “bullet” identifiers.
A
Best Diastole
B
Best Systole
Premature R-Wave
70%
70%
Premature
R-Wave
To determine the best phase, the operator typically selects one anatomic level of interest
(e.g. right coronary and circumflex arteries
both seen, at about mid heart) and performs a
reconstruction at that level throughout the
entire cardiac cycle, typically in 5% increments (e.g. 0, 5, 10% … 95%). This “preview”
series is then viewed and the best phase
selected. This manual process is time consuming, subjective, and is based on only one
anatomic level of the heart.
Figure 7: (a) A relative recon-
0.7*TRR,1
R
0.7*TRR,2
R
R
ECG
TRR,1
TRR,2
+300 ms
R
TRR,3
Time
+300 ms
+300 ms
R
R
R
ECG
TRR,2
TRR,1
TRR,3
R
Time
-300 ms
-300 ms
-300 ms
R
R
R
TRR,1
TRR,2
TRR,3
Time
Table 1: Suggested reconstruction phases as a function of heart rate.
The Best Diastole™ and Best Systole™ images are automatically selected
using a motion assessment algorithm.
Heart Rate (bpm)
Manufacturer Recommended Reconstruction Phase
<75
Only Best Diastole™
75<HR<85
Both Best Diastole™ and Systole™
Non-cardiac Artifacts
• Artifacts caused by non-cardiac motion include those caused by breathing, swallowing, or internal motion (bearing
down, Valsalva maneuver). Motion external to the heart may cause displacement artifacts in the cardiac anatomy, which
unfortunately, cannot be corrected with reconstructions at alternate phases of the cardiac cycle or by ECG editing. The
presence or absence of non-cardiac motion can be determined by:
- Looking at the sternum or diaphragm for displacement artifacts outside of the heart region using a sagittal or coronal
view (Figure 3).
- Examining the lung tissue for motion blurring using a lung window setting (Figure 4).
• Other sources of non-cardiac artifacts may include metal from surgical clips, artificial valves, pacemakers, defibrillators,
or other medical implants, which can cause streak artifacts (Figure 5).
In general, as heart rate increases, the dura>85
Only Best Diastole
tion of diastole decreases, such that endsystolic images are superior for heart rates
above 75-85 bpm. Figure 6 illustrates an ECG strip where (a) Best Diastole™ and (b) Best Systole™ were
most appropriate. Table 1 provides manufacturer recommendations for choosing which “best phase” to
reconstruct.
™
Figure 3: Discontinuities of the sternum or skin
demonstrate that extra-cardiac motion has
occurred. The interpreting physician should review
the images to determine whether the extra-cardiac
motion sufficiently compromises cardiac image
quality, such that repeating the exam is warranted.
These discontinuities can be identified most easily
using the sagittal plane.
Figure 4: Lung motion can be recognized by the blurring in
the lung parenchyma, which is best seen using the lung window setting. Some pericardial motion blurring is common.
Diaphragmatic motion, however, can cause generalized motion
of the lungs and heart that results in blurring of the mid and
peripheral lung tissue and displacement artifacts in the heart.
Figure 5: Streak artifacts due to
the presence of metallic cardiac
pacing leads. Similar artifacts are
caused by mediastinal clips post
cardiac by-pass graft surgery.
Figure 8: (a) Displacement artifacts (arrows) due to the system mistaking noise on the ECG signal as an R-wave. The incorrect syncs (R-waves) are
disabled, resulting in complete recovery of image quality (b).
To remedy this situation, the extraneous sync
can be either deleted or disabled, where disabling a sync instructs the reconstruction
algorithm to ignore that sync. To disable a
sync, the operator must right click on the
desired R-wave bullet and select “Disable
Sync” (Figure 8a). It is better to disable syncs
than to delete them because, once a sync is
deleted, it can only be restored by selecting
the original ECG box. This restores the edited
ECG to the originally recorded ECG and any
other modifications made to the syncs are
lost. In contrast, disabled syncs can be easily
enabled by right clicking on the blue bullet
and selecting “Enable Sync”.
After the extraneous sync is disabled, the
reconstruction window associated with it is
turned a darker shade of blue to denote that
images were not made there. The new reconstructed images demonstrate that the displacement artifact has been eliminated (Figure 8b).
• Using the described principles of ECG sync editing, we developed and evaluated a systematic approach
to evaluating and resolving cardiac artifacts. The raw projection and ECG data were saved for 40 retrospectively ECG-gated patient exams where cardiac artifacts were noted by the technologist performing
the examination. Thirty-three of these exams were selected from 242 contiguous, outpatient, ECG-gated
examinations having one of the
following indications: pulmonary
Figure 11: Flowchart of systematic approach to image quality recovery.
vein stenosis, coronary artery CT
angiography, or ECG-gated thorax.
Average
Heart
Rate
Heart Rate (bpm)
Thus, 14% (33/242) of cardiacgated exams required some evalu<75
>85
75-85
ation for the presence of artifacts.
An additional 7 cases were conReconstruct in
Reconstruct in
tributed from a hospital-based
Relative Systolic
Relative Diastolic
scanner, where cumulative exam
Severe
counts were unavailable.
Sternum
No
Done
Yes
Artifact?
Yes
Done*
Discontinuities?
• To assist operators in performing
the necessary ECG editing, a sysNo
tematic approach was developed
Severe
by two experienced cardiac CT
One Recon Box
Lung Window
Yes
Done*
No
Per R-R Interval?
Motion?
technologists, with assistance from
a medical physicist and cardiovascular radiologist. The proposed
No
Yes
approach is shown in Figure 11.
• To evaluate the efficacy of the sugAdd or Disable
Move Syncs
No
Syncs as Needed
gested approach, two experienced
cardiac CT technologists restored
the data to the scanner where the
Absolute reverse
data were acquired (SOMATOM
-400 ms
Bullet Points Positioned
Recon With
Yes
Done*
Above R-wave Peaks?
Absolute Phases
Definition, Siemens Medical
Absolute forward
400 ms
Solutions) and reconstructed the
images using the original ECG
data in order to document the
originally observed artifact. They
then followed the approach shown in Figure 11. Multi-planar reformatted images were reconstructed to
evaluate the cardiac region in an oblique sagittal projection and to document the effect of each step in the
suggested approach.
Figure 12:
• Clinically, once a satisfactory
reconstruction is obtained, the
A
First Action that Yielded Acceptable Image Quality
operator would not perform addi40
Action that Yielded Best Image Quality
tional ECG manipulation or image
35
reconstructions. For the purpose of
30
25
this evaluation, each step of the
20
proposed workflow was per15
formed, regardless of the success
10
of the previous steps. The first
5
action in the process that yielded
0
an acceptable image was recorded,
40%
70%
Disable
Add Sync Move Sync + 400 ms
- 400 ms
Unedited
Unedited
Sync
as was the action that yielded the
best image quality (Figure 12a).
B
Finally, the preferred reconstruc30
tion phase was recorded for each
< 70 bpm
25
exam and evaluated as a function
>70 bpm
20
of mean heart rate (Figure 12b).
15
Motion of the sternum and lung
10
parenchyma was noted on 6/40
5
(15%) and 8/40 (20%) of exams,
0
respectively. In 3/40 exams (7.5%),
70%
40%
+ 400 ms
- 400 ms
gross patient motion (breathing)
Preferred Reconstruction Phase
prevented image quality recovery.
*If artifact is not resolved, see “When All Else Fails.”
ECG
An automated technique is available on the
evaluated system that reconstructs the entire
heart in 2% intervals and computes a figure of
merit for each phase that assesses the relative
motion at each phase. The algorithm then
selects and reconstructs at full resolution the
best systolic and/or diastolic phases for the
operator (referred to as Best Diastole™ and
Best Systole™).
When the system detects an R-wave, a blue
bullet is placed on the ECG strip to denote the
synchronization point (R-wave) for that cardiac cycle. These are referred to on the user
interface as “syncs”. Occasionally, noise on
the ECG signal is considered by the system to
be an R-wave. These “extra syncs” do not represent actual R-waves and an extra image will
be reconstructed that is not really at the
desired reconstruction phase. This results in
cardiac displacement artifacts, as seen in
Figure 8a.
A
0.7*TRR,3
R
a
struction phase is expressed as
a percentage of the R-R interval.
For a 1000 ms (60 bpm) R-R
interval, a 70% reconstruction
b
phase will occur at 700 ms after
the detected R-wave, which is
approximately end-diastole. If
the heart rate increases to 90
c
bpm (667 ms R-R interval), the
70% phase will correspond to
467 ms after the R-wave, which
will be close to end-systole. To avoid this inconsistency in reconstructed
phase when the heart rate varies considerably, the reconstruction phase
can be prescribed in absolute terms, as the time (ms) either after (b) a
detected R-wave (absolute forward phase) or before (c) the next detected
R-wave (absolute reverse phase).
Eliminating Undesirable Syncs
Displacement
Artifact
Figure 9: (a) Image quality is severely degraded (arrows) when an insufficient number of R-waves is detected, forcing the system to interpolate data
beyond acceptable limits. (b) By inserting syncs at the undetected R-waves,
the image quality is completely recovered.
B
70%
Image
Formation
A Systematic Approach to Image Quality Recovery
Insert Sync
The reconstruction phase can be prescribed in either relative terms (percent of the R-R interval) or
absolute terms (a specific time interval, in ms, before or after an R-wave), as illustrated in Figure 7.
70%
Occasionally there is insufficient
data to create acceptable images
due to the large distance
between the detected R-waves
(Figure 9a). Adding new syncs at
appropriate locations can alleviate this artifact (Figure 9b). To
add new syncs, the operator
must right click on the numerical
value of the heart rate per R-R
interval on the ECG strip and
select “Insert Sync” (Fig. 9a).
Moving Existing Syncs
After inserting new syncs, it is
often necessary to move the
location of the R-wave bullets to
be positioned directly above the
R-wave peak.
Additionally, image reconstruction at different positions in the
cardiac cycle (different phases)
is possible by moving the Rwave bullet. For relative (%)
reconstructions, moving a sync
will affect the reconstruction
position both behind and in
front of the moved R-wave. For
absolute (ms) reconstructions,
only the prescribed reconstruction is affected (for forward
phase this would be after the
moved R-wave, or for reverse
phase this would be before the
moved R-wave).
Figure 10: (a) Image quality is degraded by a momentary loss of a good
ECG signal. (b) By first inserting a sync and then moving it to an appropriate location, image quality is completely restored.
A
Percent of Exams (of 40)
• Learn how variations in patient heart rate or errors in R-wave detection can cause displacement artifacts in
retrospectively-gated cardiac CT.
• Provide specific strategies for editing the patient ECG trace in order to decrease displacement artifacts.
• Show examples of the suggested processes and the resultant improvement in image quality.
Percent of Exams (of 40)
Purpose
B
A
Delete Sync
Disable Sync
B
In Figure 10a, degradation of a
portion of the image occurs due
to insufficient data in the corresponding R-R interval. This
occurred when the ECG signal
was briefly degraded, inverting
the R-wave, which the autodetection algorithm then
missed. Image quality is
restored (Figure 10b) by
1) adding one sync and
2) moving it over the position of
one of the inverted R-waves.
Step 1: Insert sync
Step 2: Move sync
When All Else Fails
• If none of the suggested ECG editing strategies resolve the artifact, or if different phases are required to
adequately visualize different arteries, some additional actions can be performed:
- Detailed ECG information may be viewed to assist the operator in knowing precisely where to disable, add or move a sync. This feature is enabled using the HeartView configuration menu.
- Additional relative or absolute reconstruction phases may be tried.
- The physician may use images from different phases for evaluating different coronary arteries. For
example, a systolic phase may be needed to adequately visualize the right coronary artery but may
not be acceptable for the left coronary arteries, which require a diastolic phase. In this case, reconstruct both phases so that the interpreting physician may use the best phase for each specific coronary
artery.
• Unfortunately, there are exams that simply cannot be made diagnostic, typically due to substantial thoracic/diaphramatic motion, an extremely poor quality ECG signal, or a very irregular heart rate.
Improved temporal resolution and/or heart rate control can reduce the frequency of this situation, as
can diligent patient education/coaching before the scan with regard to the importance of holding still
and maintaining a non-strained breathold. In the event of a non-recoverable exam, the responsible
physician should be consulted for a decision regarding repeating the exam.
References
• Leschka S, Scheffel H, Desbiolles L, et al. Image Quality and Reconstruction Intervals of Dual-Source CT Coronary
Angiography: Recommendations for ECG-Pulsing Windowing. Invest Radiol 2007; 42(8):543-549.
• Ohnesorge BM, Becker CR, Flohr TG, Reiser MF. Multi-slice CT in Cardiac Imaging. Springer-Verlag, Berlin, 2002.
© 2007 Mayo Foundation for Medical Education and Research