|
Software Patent Abstract
Methods and software for self-gating a set of images. In exemplary
embodiments, a fundamental heart frequency of the patient can be
measured without the use of an ECG signal. In one method, the fundamental
heart frequency can be determined by analyzing the size of the heart
in the images. In another method, the fundamental heart frequency
can be determined by applying a Fourier Transform. The measured
fundamental heart frequency can thereafter be used to select slice
images from the image scan for creation of a sagittal or coronal
projection image. In exemplary embodiments, the resultant projection
image can be used for coronary calcium detection and scoring.
Software Patent Claims
What is claimed is:
1. A method of self-gating a set of images, the method comprising:
acquiring a set of overlapping slice images of a patient's heart;
generating a projection with the set of slice images; marking a
region of the projection; analyzing the marked region to calculate
a heart frequency and phase of the patient's heart motion, wherein
the analyzing comprises: summing an intensity value along a slice
direction of the marked region of the projection, and Fourier transforming
the intensity value to generate Fourier components in frequency
space; selecting groups of slice images from the set of slice images,
based on their relative position in the calculated heart motion
frequency and phase; and generating a plurality of groups of slices
that correspond to different phases of heart motion.
2. The method of claim 1 comprising measuring a size of the heart
in a marked region of the set of slice images.
3. The method of claim 2 wherein measuring comprises applying a
derivative filter to measure a local intensity signal derived from
the marked region of the slice images.
4. The method of claim 1 comprising: obtaining a projection image
from each set of selected slice images, each representing a phase
of the heart motion; displaying the group of projections of the
patient's heart; and highlighting the projection image of the patient's
heart in which the marked region of the heart has its largest size.
5. The method of claim 4 wherein highlighting comprises displaying
the projection image of the patient's heart in which the patient's
marked region of the heart has its largest size as a larger image.
6. The method of claim 4 wherein highlighting comprises marking
by a ranking number the projection of the patient's heart in which
the patient's marked area of the heart has its largest size.
7. The method of claim 1 comprising: selecting a set of slices
of the patient's heart on the basis of a preferred projection of
the set of slices; and calcium scoring the selected set of slices
of the patient's heart.
8. The method of claim 1 comprising: selecting a set of slices
of the patient's heart on the basis of a preferred projection of
the set of slices; and 3-D rendering the set of slices of the patient's
heart.
9. The method of claim 1 comprising computing a frequency power
spectrum from the Fourier components.
10. The method of claim 9 comprising finding a maximum value of
the frequency power spectrum of the heart motion.
11. The method of claim 10 comprising smoothing the frequency power
spectrum of the heart motion.
12. The method of claim 10 comprising taking a middle point of
a half-height interval of the maximum value of the frequency power
spectrum of the heart motion.
13. The method of claim 1 comprising verifying that the marked
region contains enough information to compute the frequency and
phase.
14. The method of claim 13 where verifying comprises computing
that at least 3 seconds of data were included in the marked region.
15. The method of claim 13 where verifying comprises computing
that the marked region of the heart does not extend further than
half of the field of view.
16. The method of claim 1 comprising ranking the groups of slices.
17. The method of claim 16 comprising displaying the projection
images of the selected slice sets in order based on their ranking.
18. A method of Fourier gating an image dataset, the method comprising:
obtaining a plurality of overlapping slice images of a patient's
heart; generating at least one of a coronal and sagittal projection
with the set of slice images; marking a region of the projection;
calculating an intensity signal along a direction of the slice images
for the projection in the marked region; Fourier transforming the
intensity signal to find a fundamental frequency of a patient's
heart cycle; analyzing the intensity signal with a derivative filter
to locate slice images that were obtained during a diastole of the
patient's heart cycle; using the intensity signal analysis to establish
the phase of the fundamental frequency obtained from the Fourier
transformation of the heart motion; extending the selection process
outside the marked region by obtaining the frequency of the heart
motion from the Fourier transformation and the phase from the intensity
signal; and selecting slices that correspond to the patient's diastole.
Mobile Phone Patent Description
BACKGROUND OF THE INVENTION
The present invention relates generally to medical imaging. More
specifically, the present invention relates to gating of an image
scan to improve calcium scoring of a patient's heart and coronary
arteries.
CT scanning of the heart is an increasingly common procedure to
obtain information about the presence of calcification in the coronary
arteries. Unfortunately, two body motions can interfere with the
quality of the images obtained by the CT scanner: the heart motion
and the patient's breathing motion. A normal heart scan takes about
20 seconds and to reduce the effect of the breathing motion, the
patient is generally asked to hold their breath to eliminate the
breath motion. The heart motion, on the other hand, cannot be readily
eliminated and can lead to blurring, introduction of artifacts into
the images, and misregistration.
A common procedure to reduce the heart motion is gating. As is
described in U.S. Pat. Nos. 6,370,217 B1 and 6,243,437 to Hu et
al., the motion of the heart is fastest during systole and relatively
motionless during diastole. Prospective gating methodologies use
an electrocardiograph signal (ECG) to predict the time of the diastole
such that the CT scanner can be activated to obtain an image during
the relatively motionless diastole period. A major issue with prospective
gating in subjects with irregular heart beats is that the trigger
can only be set to acquire data after the R-wave. If the following
beat is short, the data acquisition may overlap the next systolic
period. Retrospective gating, on the other hand, uses the electrocardiograph
signal to retrospectively find motionless points in the heart cycle
to select the image slice. In retrospective gating, the ECG signal
information can be used, in retrospect, to select the slice images
that were acquired during the diastole. The heart moves through
a cycle in somewhat under a second, and a scanners generally take
from a quarter second to a half second to acquire the information
for each slice, thus it is possible to select from a number of slices
for each cardiac cycle.
There are two major issues with retrospective gating. The first
is that while reconstruction at finer intervals than the whole acquisition
cycle does not increase the radiation dose to the subject to produce
the extra images, the overlap of the scanned volume and the fact
that the scanner's x-ray tube is continuously on (instead of being
turned off during the parts of the cardiac cycle that are not of
interest) increase the radiation dosage. The second problem is that
gating from an ECG signal requires the placement of electrodes on
the subject and testing to confirm that their placement is adequate.
In a busy screening or diagnostic practice the added steps can decrease
utilization and negatively affect the economics of the imaging operation.
There are various shortcomings in existing software for retrospective
gating. When the operator is performing the selection of slices,
there is no real time feedback as to the adequacy of the selection.
Information as to the length of the cardiac cycle during the study,
convenient ways to ascertain whether it changed during the study,
and measurement of any one cycle are also not readily available.
Except for manually adjusting each slice (there can be 350-500 slices
in a study), there is no way to account for changes in the cardiac
cycle. All of these contribute to decreasing the certainty with
which a particular coronary calcium score is known, and to increasing
the variability of the resulting calcium scores.
Consequently, what is needed are improved methods and software
for generating a reconstructed projection image of the patient's
heart which more fully utilizes the information content of the acquisition
cycle, so that less of the increased dose is wasted or thrown out.
Additionally, what is also needed are methods and software that
can gate an image scan without the use of an ECG signal.
BRIEF SUMMARY OF THE INVENTION
The present invention provides methods and software for improving
the imaging of a patient's heart. In a particular use, the present
invention improves calcium scoring and 3-D rendering of a patient's
heart by gating a set of images without the use of a gating signal.
Advantageously, the methods of the present invention use information
present in the slice images themselves to select slices for calcium
scoring and 3-D rendering.
Typically, the images are analyzed to calculate a fundamental heart
frequency, and projection images are generated by selecting slice
images that were obtained during the same specific point (typically
diastole) of the patient's cardiac cycle. The selected images can
thereafter be calcium scored, if desired.
In one aspect, the present invention selects slice images from
the set of slices based on the size of the heart. A set of overlapping
images of the volume of the patient is acquired. Selection of the
images can be done successively by depopulating the slice set (based
on the size of the heart in the image) until the necessary number
of slice images are selected, enough to cover the heart without
gaps, which depends on slice thickness and heart size. In one exemplary
embodiment, depopulating the image scan can be carried out by pairwise
comparison. Once the slice images are selected, the coronal/sagittal
projection can be generated and the selected images of the heart
can be calcium scored or 3-D rendered.
In another aspect, the present invention comprises generating a
plurality of sagittal or coronal projection images of the patient's
heart. Each projection image will include groups of slice images
of the patient's heart that were taken during the same phase of
the patient's heart frequency. Consequently, a projection image
can be displayed of the patient's heart during each of the phases
of the patient's heart beat (e.g. systole, diastole, and the like.)
Thereafter, a user can determine which slice sets are best for calcium
scoring or 3-D rendering based on the projection images.
In an exemplary method, a set of overlapping slice images of a
patient's heart is acquired. A coronal or sagittal projection with
the set of slice images is generated and a region of the projection
is marked. The marked region is analyzed to calculate a heart frequency
and phase of the patient's heart motion. Groups of slice images
are selected from the set of slice images, based on their relative
position in the calculated heart motion frequency and phase. Thereafter,
a plurality of groups of slices are generated that correspond to
different phases of the heart motion.
In some embodiments, the marked region is analyzed by applying
at least one of a Fourier transformation and a derivative filter
to an intensity signal that is derived from the slice images. The
Fourier transform can be used to derive a fundamental heart frequency,
while the derivative filter can be used to measure the phase of
each of the slice images so as to allow the user to determine which
slices correspond to the patient's diastole.
In some configurations, the methods and software of the present
invention can apply a quality measure to the plurality of groups
of slices to rank the images. Typically, the images will be ranked
on the size of the marked region of the heart in the projection
of the slices, since the heart is largest (and clearest) when the
heart is in diastole.
In another aspect, the present invention comprises determining
a fundamental heart frequency of the patient by applying a Fourier
transformation to an intensity signal of the image slices. A plurality
of overlapping slice images of a patient's heart can be obtained.
A coronal or sagittal projection is generated with the set of slice
images. The invention of this application is not limited to the
use of coronal or sagittal projections. Other projections may be
chosen, such as those of the heart's short or long axis. A region
of the projection image is marked and an intensity signal of the
marked overlapping slice region is calculated along each line in
the projection image corresponding to a slice. The intensity signal
can be Fourier transformed to find a fundamental frequency of the
patient's heart cycle. The intensity signal can be analyzed with
a derivative filter to locate slice images that were obtained during
the diastolic portion of the patient's heart cycle. The intensity
signal analysis can be further used to establish a phase of the
fundamental frequency obtained from the Fourier transformation of
the heart motion. The selection process can be extended outside
the marked region by obtaining the frequency of the heart motion
from the Fourier transformation and the phase from the intensity
signal, and slices can be selected that correspond to the patient's
diastole. Optionally, the selected slices can thereafter be calcium
scored and/or 3-D rendered.
In yet another aspect, the present invention provides a method
of Fourier gating an image dataset. The method comprises obtaining
a plurality of overlapping slice images of a patient's heart. A
coronal or sagittal projection is generated with the set of slice
images. A region of the projection image is marked and an intensity
signal of the marked overlapping slice region is calculated along
each line in the projection image corresponding to a slice. The
intensity signal can be Fourier transformed to find a fundamental
frequency of the patient's heart cycle. A principal component of
the Fourier spectrum is obtained within an allowed frequency window.
Data sets of slices are formed in which the datasets are separated
by a time interval that substantially corresponds to the time interval
corresponding to the principal component. A projection image formed
from the data sets is presented to the operator to select a set
for further processing. Optionally, the selected slices can thereafter
be calcium scored and/or 3-D rendered.
For a further understanding of the nature and advantages of the
invention, reference should be made to the following description
taken in conjunction with the accompanying drawings and claims.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 schematically illustrates a simplified retrospective gating
method of the present invention with the optional steps in dotted
lines;
FIG. 2 illustrates one exemplary graphical user interface displaying
information regarding the duration of an R-R cycle of a patient;
FIG. 3 is an enlarged portion of the R-R cycle information of FIG.
2;
FIG. 4 illustrates a graphical user interface having the view tab
and view screen displayed and slice images selected during a diastole;
FIG. 5 illustrates a graphical user interface displaying slice
images selected during a systole;
FIG. 6 is a graphical user interface displaying a stretched image
and an overlaid ECG signal;
FIG. 7 schematically illustrates a simplified method of self gating
a set of image slices;
FIG. 8 schematically illustrates another simplified method of self
gating as et of image slices with the optional steps in dotted lines;
FIGS. 9A and 9B are coronal and sagittal projections of a patient's
heart, respectively;
FIG. 10 illustrates a freehand editing of a region of interest;
FIG. 11 illustrates a straight line editing of a region of interest;
FIG. 12 is an example of an intensity profile;
FIG. 13 is a smoothed output of a local intensity signal with markers
indicating maxima obtained from a derivative filter;
FIG. 14 is an example of a power spectrum;
FIG. 15 illustrates a graphical user interface in which an ECG
has not been loaded and a user can self-gate the image scan;
FIG. 16 is an exemplary data flow diagram of self gating; and
FIG. 17 is a graphical user interface of a self gating preview.
DETAILED DESCRIPTION OF THE INVENTION
The present invention provides methods and graphical user interfaces
for self gating and retrospectively gating a set of image slices
(referred to herein as an image scan).
While the remaining discussion focuses on the gating of an image
scan from a CT scanner for use in coronary calcium measurements,
it should be appreciated that the methods and devices of the present
invention are not limited to such imaging modalities and uses. For
example, instead of analyzing the image scan for measuring coronary
calcium, the image scan can be used for 3-D reconstructions of the
heart, such as those used for CT angiography, or for heart function
studies, including dynamic studies.
In some exemplary embodiments, the present invention uses a patient's
measured ECG signals taken during the acquisition of the image scan
to gate the image scan. The ECG signal is a repetitive pattern that
reflects the electrical activity of the patient's heart. An ECG
signal has a plurality of cardiac cycles (sometimes referred to
as R-R cycles), with each cardiac cycle beginning with an R-wave
(e.g., highest amplitude peak) during systole and ending with a
relatively motionless diastolic phase. Blurring of the images is
most likely to occur when imaging during systole. Consequently,
it is preferable to use image slices taken during diastole so as
to reduce the amount of artifacts found in the selected image(s).
Unfortunately, the R-R interval can vary through the image scan
and the cardiac cycle will not always occur during regular intervals
(e.g., irregular heartbeat). For example, in many imaging sessions,
the subject is asked to hold their breath so as to reduce introduction
of artifacts due to the breathing motion. The patient's holding
of their breath, however, may cause a change in the heart cycle.
Additionally, patient's who have irregular heart beats may not be
effectively imaged by selecting a specific point in time during
the cardiac cycle. While some studies have proclaimed that it is
best to select a particular time with respect to the R-wave, (some
preferring a certain number of milliseconds before or after the
R-wave), the selection of an absolute time does not allow for compensation
for irregular heartbeats or changing times between successive R-waves.
FIG. 1 schematically illustrates one method 20 of the present invention.
First, a set of slice images of a volume of tissue of the patient
is obtained and a coronal or sagittal reconstruction of the slice
images can be generated. (Step 22). Acquisition of the image scan
can be carried out by any conventional or proprietary CT scanner
(e.g., moving, stationary, single detector, multiple detector, helical,
and the like). A helical scan of the heart can include approximately
350-500 overlapping images, while a non-overlapping scan usually
includes around 40-50 slices. It should be appreciated however,
that it may be possible to use magnetic resonance image (MRI) scanners,
ultrasound scanners, or other slice imaging devices to obtain the
image scan used in the methods of the present invention.
The electrical activity of the patient's heart can be measured
by attaching one or more electrocardiograph leads to the patient
to monitor the patient's ECG signal during the acquisition of the
image scan. The electrical activity can be analyzed to derive information
regarding the duration of each of the R-R cycles of the ECG signal
(Step 24). The R-R cycle information allows the user to determine
if there are any substantial variations in the duration of the R-R
cycles over the acquisition period of the slice images. Such information
allows the user to make appropriate adjustments to their selection
of the slice images used for generating the coronal/sagittal projection,
for calcium scoring, or for 3-D rendering.
The ECG information can be analyzed automatically by software or
manually by the user to determine the duration of each cardiac cycle
(illustrated in FIGS. 2 and 3 as "Duration of R-Cycle").
Based on the calculated R-R cycle information, the user can choose
an appropriate global selection criteria of choosing the slice images
from the image scan (Step 26). In exemplary embodiments, the selection
criteria for choosing the slice image(s) includes (1) an absolute
time period before or after the R-wave or (2) a percentage of the
cycle (e.g., 65% of the heart cycle) before or after the R-wave.
In exemplary embodiments, the user will be allowed to separately
choose the selection criteria (e.g., percentage or absolute time)
and a timing selection (e.g., before or after the R-wave).
It should be appreciated however, that while the preferred selection
criteria are an absolute time before or after the R-wave or a percentage
of the cardiac cycle before or after the R-wave, that other selection
criteria may be used to select the slice images.
In some embodiments, when the user selects a percentage of the
cycle as the selection criteria, the software of the present invention
can also display a complementary time value that corresponds to
the selected percentage. Similarly, if a user chooses an absolute
time period, the software of the present invention can display the
corresponding percentage. This significantly decreases operator
load. For instance, for most of the heart cycles, the heart rate
may be quite constant. The operator can set a preferred time, say,
450 msec before the R-wave, and the program will show what percentage
of the cycle this is. The operator can then select for a portion
of the scan when the heart rate slows near the end, a percentage
that is already available from the software. In another case the
heart rate may have increased, and the end of the scan may overlap
the following R-wave. The operator can then select the complementary
time after the R-wave to better center the selected image.
In exemplary embodiments, a graphical illustration of the duration
of the R-R cycle can be displayed on a user interface to illustrate
the duration of the R-R cycle. Advantageously, the graph of the
R-R cycle will guide the user toward the patient's irregular heart
beats and show if there are any substantial variations in the length
of the R-R cycle that may effect the selection of the slices.
One example of a graphical illustration is illustrated in FIG.
3. Graph 92 shows that the duration of the R-R cycle is substantially
the same length during the entire acquisition period. Graph 94 (in
dotted lines) shows that the duration of the R-R cycle changes over
the acquisition period.
In the instance in which graph 92 is relatively consistent over
time, the user can apply an absolute time period (before or after
the R-wave) to select the slice images for inclusion in the reconstruction
projection image. Since the R-R cycles are substantially the same
throughout the acquisition period, the absolute time period should
generally fit each of the R-R cycles.
If the duration of the R-R cycle changes over the acquisition period
(shown as a dotted line 94), the user would likely use the "percentage
of cardiac cycle" (before or after the R-wave) selection criteria
to select the slices since the absolute time does not compensate
for irregular or changing times of the R-R cycle. While an absolute
time period, (for example 450 msec before the R-wave) may be appropriate
for a first portion of the ECG signal, because the R-R cycle decreases
over time, the chosen absolute time period would likely be inappropriate
for the latter, shorter R-R cycles since the selected slice would
likely overlap over a portion of the high amplitude R-wave. Thus,
such a slice would likely introduce artifacts into the resultant
projection image and reduce the accuracy of the calcium scoring
of the image slice.
Additionally or alternatively, to graphically illustrate the duration
of the R-R cycle, the methods of the present invention can also
numerically display the duration of the R-R cycle for specific intervals
of the acquisition period of the ECG. For example, as illustrated
in FIG. 3, the acquisition period may be broken up into a plurality
of intervals. In one exemplary embodiment, the first interval 104
is the first 10 cycles of ECG, the second interval 106 is the middle
10 cycles of the ECG, and the third interval 108 is the last 10
cycles of the ECG. It should be appreciated however, that the ECG
can be separate into any number of different ECG intervals, and
the present invention should not be limited to the illustrated three
intervals.
By quantitatively providing the average length of the R-R cycle
for the different intervals, the user will be able to accurately
determine which selection criteria to employ. For example, if the
R-R cycle duration varies by more than a certain percentage or time
length (typically about 70 msec or about 10% of the R-R cycle),
the user will likely want to employ the percentage selection criteria.
But if the R-R cycle duration difference is less than the certain
percentage or time length, the user will likely want to employ the
absolute difference criteria, as described above.
Alternatively, instead of choosing a global absolute time period
for all of the cycles, it may be possible to apply a separate selection
criteria to each of the intervals of the ECG. Thus, if two of the
intervals are consistent and the third interval is changing in duration
or at a lower duration than the first two intervals, it may be beneficial
to apply an absolute time selection criteria to the first two intervals
and a shorter absolute time duration or a percentage of cycle to
the third interval. For example, for the illustrated example in
FIG. 3, as first attempt, the user can select a slice image that
is 450 msec before the R-wave for first 10 cycles, 450 msec before
the R-wave for middle 10 cycles, and 400 msec before the R-wave
for last 10 cycles. In this manner, an optimal selection can be
achieved, the possibilities being limited by the acquisition process,
and not the gating software.
After the appropriate selection method is chosen and applied, the
selected slices will be combined to generate a corrected coronal/sagittal
projection. (Step 28). In some embodiments a bilinear algorithm
is used to generate the correct aspect ratio coronal/sagittal projection.
It should be appreciated however, that other conventional interpolation
and scaling algorithms can be used.
The combination of functionalities and flexibility in choosing
the slices allow for convenient and at the same time highly specific
selection of slices on the basis of timing with respect to the ECG
signal. Because the selection of the slices can be displayed to
the user in real time (described below), the user can rapidly assess
the adequacy of the timing selection of the slice images.
If the projection images are deemed to be acceptable, the selected
slices can be calcium scored or 3-D rendered, if desired. (Step
32). Because there is an overlap of the slices during scanning,
and because the x-ray tube is on during the full cardiac cycle instead
of just during the acquisition of the desired time interval within
the cycle (as in prospective gating) there is an increase in delivered
radiation dose to the patient. Such a dosage increase in unavoidable,
but retrospectively it is possible to obtain information from the
additional radiation dose. After acquisition, the reconstruction
software can generate additional slices at finer intervals than
those determined by table motion and scanner rotation speed, typically
ten times finer. In methods which analyze the slice images for calcium
scoring, the calcium will be very bright in the images. Using a
maximum intensity projection algorithm, the selected slice and its
two immediate neighbors can be analyzed to select the brightest
pixel in each of the slices. The slice that has the brightest pixel
can then be chosen for inclusion in the calcium scoring study. Thus,
the process of the present invention effectively utilizes three
out of ten images (e.g., the "selected" slice and its
two neighbors) instead of just one out of every ten images.
Because a CT image is obtained from hundreds of individual projections
and processed through back-projection algorithms, inconsistencies
in some projections due to heart motion or motion of a point in
the heart that in some way aliases with the acquisition process
can produce a significant artifact even at a time where the heart
is relatively quiescent. Optionally, if the selected slice images
chosen by the above method are not all deemed appropriate because
of such a problem, the user can manually scroll through the selected
slice images and choose other "non-selected" slice images
to replace the undesired "selected" slice images. (Step
30). One method of deselecting slices from the image scan is described
below, in relation to one exemplary graphical user interface of
the present invention.
FIGS. 2-6 illustrate some exemplary graphical user interfaces and
methods for gating an image scan. It should be appreciated however,
that the graphical user interfaces described and illustrated herein
are meant only to be examples, and should not be used to limit the
scope of the present invention.
FIG. 2 schematically illustrates one exemplary graphical user interface
(GUI) 40 of the present invention. GUI 40 is generally displayed
on a user output device such as a computer monitor. GUI includes
a first screen portion 42 for displaying a selected image, a second
screen portion 44 for displaying an ECG that was taken during the
image scan, and a third screen portion 46 for displaying a coronal
and/or a sagittal image projection of the selected slices. Typically,
third screen portion 46 will display a first projection image 48
that is composed of all of the slices of the image scan and/or a
second projection image 50 that is composed only of the selected
images slices. As will be described in detail below, GUI can further
include a fourth screen portion 52 that can be toggled between a
variety of views to allow a user to select and display various functions,
menus, and information. GUI can also include a menu toolbar 53 so
as to allow a user to select and toggle between the different functionalities
and plug-ins of the software of the present invention.
In preferred embodiments the GUI 40 of the present invention can
simultaneously display on a single screen a selected slice image,
at least a portion of the ECG signal, and the sagittal/coronal reconstruction
projection image that is composed of the selected slices. Such an
interface 40 allows the user to view in real-time, the effect that
the choice or change of image slices has on the quality and resolution
of the composite projection image. Thus, if the selected slices
do not improve the quality of the coronal or sagittal reconstruction
projection image, the user can de-select the slice(s) to improve
the image quality, and hence improve the calcium scoring or 3-D
rendering of the patient's heart.
As shown in FIG. 2 in exemplary embodiments first screen portion
42 can display a selected slice image in window 54 and previous
and next slice images in windows 56, 58, respectively. Slice image
window 54 can include a header 60 that indicates the slice number,
zoom level, and the like. The adjacent slice image windows 56, 58
can include a header that indicates "Previous Slice" or
"Next Slice." It should be appreciated however, that a
variety of headers can be used to indicate other information, if
desired. Image windows 56, 58 can include a scroll bar 61 that allows
a user to scroll through (review) the slice. In some exemplary embodiments,
image windows 56, 58 that display the non-selected slices are smaller
in size than image window 54. It should be appreciated however,
that if desired, image windows 56, 58 can be the same size or larger
than image window 54 if desired.
First screen portion 42 can also include user actuatable buttons
62, 63 that allows the user to toggle through the other individual
slice images of the image scan. If user actuates button 63, the
image slice that was originally displayed in window 58 will be displayed
in window 54, the image slice that was originally displayed in window
54 will be moved to image window 56, the image originally displayed
in image window 56 will not be displayed, and a previously undisplayed
slice image will be shown in window 58. Likewise, if a user actuates
button 62, the image slice that was originally displayed in window
56 will be displayed in window 54, the image slice that was originally
displayed in window 54 will be moved to image window 58, the image
originally displayed in image window 58 will not be displayed, and
a previously undisplayed slice image will be displayed in window
56.
As shown in FIG. 4, if the slice image displayed in window 54 is
not a "selected slice," first screen portion 42 can include
a "Select Slice" button 64 that allows the user to select
a previously "unselected" slice that is displayed in window
54 for inclusion into the projection image displayed in third screen
portion 46. Similarly, if a slice displayed in window 54 is a slice
that is already selected or included in projection image 50, first
portion 42 can include a "Deselect" button 65 that, when
actuated, can remove the slice from inclusion in the reconstruction
projection image. (FIG. 2)
If through any of the process described therein, there are gaps
in the image data, before saving or calcium scoring the gated image,
the user will be warned of the gaps and asked if the gaps should
be filled. If the user chooses to fill the gap, the software can
automatically fill the gap by selecting a slice image that is substantially
in the middle of the gap.
As shown in FIG. 4, in exemplary embodiments, first portion 42
can also include a "Previously Selected Slice" button
66 and a "Next Selected Slice" button 68 that allows the
user to jump to the next or previous selected slice in the image
scan. In exemplary embodiments, the next selected slice will be
a slice that corresponds to a similar time point during the R-R
cycle, as described above.
Windows 48, 50, 54, 56, 58 can be zoomed in and out, panned to
adjust the size of the image displayed. The zooming and panning
can be done synchronously for all of the windows, or the zooming
of each window can be performed independent of each other.
Referring again to FIG. 2, second screen portion 44 of GUI 40 can
include an ECG field 70 that displays a patient's ECG signal that
was taken during the imaging of the patient's heart. In most embodiments,
only a portion of the entire ECG reading will be displayed on the
screen. Thus, a scroll bar 72 and a zoom bar 74 can allow the user
to scroll through the ECG and/or to zoom in and out of the ECG.
The ECG field can be highlighted, typically through a difference
in colors or shading from a background of the ECG field, to indicate
which slices are chosen relative to the ECG for inclusion into the
projection image 50. For ease of reference, the selected slice image
that is displayed in window 54 will generally have a different shading
from the ECG field background and the highlighting of the other
selected slices. In one exemplary embodiment, the slice displayed
in window 54 will be identified in the ECG field by a light red
band 76, and the other selected slices will be identified by a blue
band 78.
In some embodiments, if the user wishes to manually measure the
time interval of an R-R cycle(s), the user can measure the time
interval between two arbitrary or chosen points within the ECG setting
one boundary delimiter by clicking into the ECG and dragging the
free boundary delimiter with a mouse, or other input device, to
the second point on the ECG. A field below the ECG can then display
the time length between the two selected points (not shown).
As seen further in FIGS. 2, 4, and 5, information regarding the
number of selected slices, position of the current slice in the
ECG (in milliseconds), and the position of the current slice in
millimeters, can be placed below the ECG field to provide information
to the user about the selected slice and ECG.
As the user scrolls through images in the first portion 42, the
user can merely click on the image window 54 to center the ECG cardiac
cycle within the ECG field so that the user can simultaneously view
the selected image slice and its corresponding cardiac cycle. Alternatively
clicking on a portion of a stretched (or normal) reconstruction
projection will display such a slice in window 54 and center the
corresponding ECG signal in the ECG field. Moreover, the user can
use scrollbar 72 below the ECG field to scroll through the R-R cycles
until the selected R-R cycle is displayed within the ECG field.
As noted, the selected R-R cycle will be highlighted a different
color from the other selected R-R cycle slices. Also, clicking on
the ECG display will select a slice with its center closest to the
point where the user had placed a cursor. The slice will be highlighted
on the ECG to display the location of the slice relative to the
ECG.
Third screen portion 46 can be configured and sized to display
one or more reconstruction projection images. In exemplary embodiments,
third screen portion 46 can display a coronal and a sagittal projection
image of the slices. Alternatively, third screen portion 46 can
display only a projection image that is composed of only the selected
slices. If desired, in order to provide a visual impression of the
image quality of the projection image with only the selected slices
50, a projection image having all of the slice image of the image
scan 48 can be shown adjacent image 50. Additionally, the third
screen portion may only show the coronal/sagittal projection image
having only the selected slices.
Third portion 46 can include a line 80 across the reconstruction
projection image to indicate the position of the slice image that
is displayed in window 54.
In exemplary embodiments, fourth screen section 52 can be toggled
between an "ECG" screen 82 (FIGS. 2 and 3) and a "View"
screen 84 (FIGS. 4 and 5). Once the View tab 83 is activated, a
View screen 84 will be displayed. View screen 84 includes buttons
86, 88 that allow the user to change the view of the reconstruction
image 50 between a coronal (or MPR3) and a sagittal (or MPR2) projection.
Fourth screen portion 52 can include an ECG tab 90 which when clicked
or otherwise selected by the user will display ECG screen 82 so
as to display information about the average length of the R-cycle
for the patient for certain intervals of the ECG. In some embodiments,
the ECG screen will have a graph which illustrates the duration
of the patient's R-R cycle. Such a graph can graphically illustrate
the duration of the R-R cycles, typically in milliseconds. Thus,
if the R-R cycle is seen to be decreasing or increasing over time,
the user can modify the method in which the slice images are selected.
For example, as shown in FIGS. 2 and 3, the graph 92 shows that
the R-R cycle stays relatively constant through 30 measured R-R
cycles. For such information datasets, selecting an absolute time
before or after the R-wave will likely be sufficient to select the
appropriate slice images for inclusion into the projection reconstruction.
If, however, the patient had graph 94, which shows a change in the
duration of R-R cycle over time (e.g., a slope in the graph), it
would probably be beneficial to use a percentage of cardiac cycle
as the selection criteria for the slices.
Once the user decides on a selection criteria, the user can activate
View tab 83 to bring up View screen 84. View Screen will include
fields that 96 allow the user to enter their desired selection criteria.
View Screen 84 can also include an Apply Values button 98 that applies
the slice selection criteria for the R-R cycles, a Deselect all
Slices 100, Center the ECG image in the ECG field 102 and described
more fully below.
If the Deselect All Slices button 100 is activated, the slices
that were selected for inclusion in the reconstruction projection
image will all be deselected and the user will be allowed to reselect
the slice images for the reconstruction projection, using the slice
selection criteria input into the specified field. Activation of
the Center button 102 will center the ECG cardiac cycle within the
ECG field for the image slice that is displayed in window 54.
As illustrated in FIG. 6, in order to display a stretched image
of the coronal and/or sagittal reconstruction projection, the user
can activate a input box 105 in the fourth screen section. A stretched
image allows examination as to whether a particular slice fits well
with respect to its neighbors, or whether another slice may fit
better.
Referring again to FIGS. 4 and 6, checking of a box on the interface
will provide a stretched coronal or sagittal projection of the reconstruction
of slices in third screen section 46. Checking of box 105 will make
a "Display/Overlay ECG" box 107 active to allow the user
to overlay an ECG signal over the stretched reconstruction projection.
If desired, the user can overlay the ECG over the stretched image
so as to allow the user to determine if a slice fits well with respect
to its neighbors in a particular cycle, or whether another slice
may fit better.
The stretched view is needed because the spatial resolution of
the computer screen/eye combination is not sufficient to adequately
view the image with the necessary detail. Zooming the image would
require too large a space on the screen for the in-plane dimension,
so that the image is zoomed only along the slice axis and thus appears
stretched.
When displaying a stretched image with an overlaid ECG, fourth
screen portion 52 can include a "Match" button 113. As
shown in FIG. 6, activation of the "Match" function will
scale and zoom the stretched view of the ECG in window 109 to match
the portion of the ECG displayed in window 111, the two ECGs being
displayed synchronized. In addition, with the click of a button
on the input device, the software of the present invention can also
center the ECG and the stretched view on the current slice, in case
it is scrolled out of the field of view.
If the user desires to replace a slice image from the stretched
view, the user can scroll through the slices displayed in window
42 until the highlight marker 76 in the ECG field 70 is over the
desired portion of the cardiac cycle within field 70. Thereafter,
the user can activate the select slice button 64 to include the
slice in the stretched view.
Referring again to FIG. 4, the user can choose to toggle between
a coronal projection and a sagittal projection to alter the view
of the projection image by activating the input field 86, 88. In
other embodiments, it may be possible to activate both of fields
86, 88 so as to simultaneously display the coronal and sagittal
projections.
The method of using the graphical user interfaces of the present
invention will now be described. The software of the present invention
can be a stand alone software package or it can be in the form of
a plug-in into a software package, such as a calcium scoring package.
First, the user can load an image scan, or a collection of slices
acquired during imaging into the software. The image scan can be
a saved image scan, or alternatively, the image scan can come directly
from a CT scanner attached to the computer running the software
of the present invention.
If available, ECG information that corresponds to the image dataset
can also be downloaded into the software. If an ECG information
is not available, the software can use the self gating methods described
below, to gate the images. If an ECG is loaded into the software,
the ECG will be displayed in ECG field 44 and a composite sagittal/coronal
image of all of the slices of the image scan will be displayed in
window 46. In some embodiments, a center slice of the image set
and its two neighbors can be displayed in windows 54, 56, and 58.
As can be seen in FIG. 2, the composite image with all of the slices
will generally have a jagged outline due to the movement of the
heart. To improve the selection of the slices included in the sagittal/coronal
projection, the user can click on the ECG tab 90 to display the
R-R cycle information.
After analyzing the R-R cycle information for any changes in the
duration of the R-R cycle during the acquisition period, the user
can choose from a plurality of selection criteria, typically either
an absolute time period or percentage of cycle period. The user
can select the View tab 83 and enter the selected criteria in the
appropriate field 96. In some embodiments, if the user selects an
absolute time selection criteria for a slice, the program will automatically
calculate a corresponding percentage of cycle that corresponds to
the absolute time entered by the user for that slice (Window 54).
Similarly, if the user selects a percentage of cycle as the selection
criteria, the software will automatically calculate and display
a corresponding absolute time relative to the R-wave.
Once the user has entered the selection criteria, the user can
activate the Apply Values button 98 to select the slices for inclusion
into the projection image. As shown in FIG. 2, once the selection
criteria value is applied, the user will be provided with a coronal/sagittal
projection using only the selected slices in window 50 that is adjacent
the coronal/sagittal projection using all of the selected slices.
The ECG will also be highlighted 76, 78 to illustrate which slices
are chosen and the position of the slices relative to the ECG.
FIG. 4 illustrates an coronal image 50 which was selected during
the diastole. In contrast, FIG. 5 illustrates the coronal projection
image 50' that was selected during systole. As can be seen in the
images, the coronal projection image of the heart during systole
is noticeably blurrier.
If the user desires to re-select the selection criteria, the user
can again click on the View tab 83 and enter a new selection criteria
(e.g., a new time or percentage value) until an acceptable coronal/sagittal
projection image is generated. Advantageously, because the coronal/sagittal
image is updated in real-time when the new slices are selected,
the user can tell, in real-time, the effect of the choice of the
images on the quality of the coronal/sagittal projection image.
Once the user has found an acceptable "global" selection
criteria, the user can manually scroll through the slice images
to select or deselect individual slice images of the image scan
to improve the choice of the individual slice images. For example,
as shown in FIG. 2, to scroll through the selected slices, the user
can activate the Prev. Selected Slice button 66 and Next Selected
Slice Button 68. Such buttons will display in window 54 the Selected
slice and in windows 56, 58 the slices adjacent the selected slice.
If the user wants to keep the slice displayed in window 54, the
user can move to the next slice image by pressing either button
66 or 68. If however, the user wants to select another slice, the
user can activate the Deselect button 65 and scroll through the
adjacent slices by activating button 62, 63. Once the user finds
a slice that is acceptable, the user can activate the Select button
64 (FIGS. 4 and 5). The user can repeat this process until all of
the slices have been selected. Thereafter, the user can save the
image scan (e.g., the selected slices, the sagittal/coronal projection,
selection criteria, and the like), and the image of the heart with
the selected slices can be calcium scored and/or 3-D rendered. The
calcium scoring can be carried out by a separate software program,
or it can be carried out by the same program that gated the image
scan. Some exemplary computer systems for displaying the GUI of
the present invention, calcium scoring methods, and software are
more fully described in co-pending U.S. patent application Ser.
No. 10/096,356, filed Mar. 11, 2002 and U.S. patent application
Ser. No. 10/126,463, filed Apr. 18, 2002, entitled "Methods
& Software for Improving Coronary Calcium Scoring Consistency,"
the complete disclosures of which are incorporated herein by reference.
In another aspect, the present invention provides methods and software
for gating an image scan without the use of a gating signal. Because
the gating signal (e.g., ECG signal) requires the purchase and use
of additional expensive hardware and software packages, and requires
added time for placing the electrodes on the patient and confirming
the adequacy of the signal being obtained, it is often desirable
to be able to perform an image scan without the use of a gating
signal. Exemplary self-gating methods of the present invention use
information derived from the image slices themselves to infer the
heart motion without the use of an ECG signal.
In one self-gating method, the image slices are selected through
detection of the size of the heart or pixel intensity in each of
the slice images. In another self-gating method, image slices are
chosen through deriving an average heart rate from the variability
of the signal in the image data and selecting the images based on
the calculated frequency information. In some configurations, a
size of the heart is used in conjunction with the frequency measurement
to perform the slice selection.
During the quiescent time (e.g., diastole), the heart will be imaged
in relatively motionless and fully expanded size. In contrast, when
the heart is in systole, the heart will be contracted. By selecting
the images in which the image of the heart volume is largest, the
set of images will be selected when the heart is in diastole.
In a first self-gating method illustrated schematically in FIG.
7, the software of the present invention selects slice images from
the set of slices based on the size of the heart. The first step
of the first method of self-gating the images is to acquire the
set of overlapping images of the volume of the patient (Step 192).
Selection of the images can be done successively by depopulating
the slice set (based on the size of the heart in the image) until
the necessary number of slice images are selected, enough to cover
the heart without gaps, which depends on slice thickness and heart
size. In one exemplary embodiment, depopulating the image scan can
be carried out by pairwise comparison. (Step 194). Once the slice
images are selected, the coronal/sagittal projection can be generated
and the image of the heart can be calcium scored or 3-D rendered.
(Step 196).
If in some of the methods of this invention there are gaps in the
image data, before saving or calcium scoring the gated image, the
user will be warned of the gaps and asked if the gaps should be
filled. If the user chooses to fill the gap, the software can automatically
fill the gap by selecting a slice image that is substantially in
the middle of the gap.
By drawing on a sagittal or coronal view a region of interest (ROI)
encompassing one side of the heart, one can determine the state
of the heart muscle by noting the total signal along the line representing
the slice, or noting how many pixels have the signal of muscle rather
than the much lower signal of fat of the lung. When comparing a
slice to its immediate neighbors, the slice with the most expansion
will provide a line with a higher total signal, or with more pixels
above a specified threshold, than a slice belonging to a point in
time with less expansion. For pairwise comparison, each slice is
compared to one neighbor, and the one with most expansion kept.
This process can stop when a gap would be generated by further depopulation
of the slices.
FIG. 8 schematically illustrates another simplified method of self-gating
in which the images are selected by finding the fundamental frequency
of the heart from the images themselves. The exemplary method 200
comprise obtaining a set of overlapping slice images of a volume
of the patient, typically of the patient's heart. (Step 202). As
described above in relation to the retrospective gating, the set
of images can be obtained with a CT scanner, or an equivalent imaging
technology.
The set of images can be run through an algorithm to generate a
coronal/sagittal projection of the volume of tissue of the patient.
(Step 204). FIGS. 9A and 9B illustrate a coronal and sagittal projection
of an image of a patient's heart. The images have jagged edges due
to the motion of the heart.
The user can then highlight one or more region of the heart in
the coronal/sagittal projection (Step 206). Generally, the user
can select some region around the jagged outline of the heart. More
distinctive outlines around the heart will give better results.
In exemplary embodiments, the regions of the heart can be marked
with a freehand region (FIG. 10), a straight line region (FIG. 11).
Due to the scanner rotation time, the outline of the heart is generally
only selected on one side of the heart outline. It should be appreciated
however, that other conventional marking methodologies can be used
to mark a region of the image, including the use of automated boundary-finding
algorithms.
A pixel intensity signal of the images can be generated by summing
the pixel intensities (HU) within the selected region for each slice
in a direction that is perpendicular to the slice direction. (Step
208). The result of the summation is a signal graph, as illustrated
in FIG. 12. The signal graph will produce a plurality of maximas
and minimas, wherein each of the maximas generally corresponds to
a maximum inflation of the heart. The position along the horizontal
axis corresponds to the slice number. The intensity is 0 for slices
not included in the selection. The intensities can be corresponded
to the slices, each of which is acquired at a certain point in time,
usually every 100 ms. It should be appreciated however, that the
point in time in which the image acquisition is performed will vary
depending on the patient's heart rate or other factors, and such
parameters can be set accordingly.
The signal can be analyzed to extract the time information from
the images. (Step 210). Extracting the time information from the
signal can be carried out through an analysis of the frequency spectrum
and/or through analysis of local intensities of the signal.
In one exemplary method of extracting the time information, a Fourier
transformation is applied to analyze the frequency components of
the signal. In the Fourier transformation, the amplitude profile
is viewed as a function of frequency. Each function can be represented
through its Fourier components--by combining a number of sine and
cosine functions of different frequencies. The signal intensity
profile of the slices will provide a repetitive maxima and minima.
The sinusoid (e.g. sine or cosine function) of the same frequency
as the repetitive pattern will have a large contribution. The goal
is to find this principal component, the sinusoid of the corresponding
frequency.
The result of the Fourier analysis will be a series of complex
numbers. Each number corresponds to a sinusoid of a certain frequency.
The formula is:
.function..times..times..function..times.e.times..pi..times..times.
##EQU00001## where m is the slice number, M is the total number
of slices and k is the coordinate in frequency space (or k-space)
and k/M is the frequency which corresponds to value F(k),
From the Fourier analysis, information about the magnitude and
phase of the sinusoid can be obtained. The magnitude indicates the
strength of any one frequency component, including the principal
component. For each component there is corresponding phase information
which contains information about where that component begins. While
the phase can theoretically be obtained from this phase information,
in practice, the phase is changing very fast as a function of frequency,
and the measurement is not reliable.
To find the frequency which is the most dominant portion of the
function, only the information about the "energy" for
each frequency component is necessary. The energy of the sinusoid
can be read from the power spectrum, in which power is defined by:
Power(k)=Re(F(k)).sup.2+Im(F(k)).sup.2 where Re is the real part
of the complex number F(k) and Im is the imaginary part of the complex
number F(k). The result will be a sequence which contains only real
values. One example of a power spectrum is illustrated in FIG. 14.
After the power spectrum is computed, the Fourier series can be
smoothed with a Gaussian filter to reduce spurious peaks. Because
the task of finding the heart beat is circumscribed by physiologic
restrictions, the present invention can restrict the search for
the maximum frequency to a range of approximately 1/2000 ms and
1/500 ms, which corresponds to an interval of 500 ms to 2 seconds
between two heart beats.
Thereafter, the absolute maximum value in the power series and
frequency can be determined. Additionally, the lower and higher
frequencies next to the maximum frequency where the value is half
of the maximum value can be measured (noted as the half-height interval
in FIG. 14). If the maximum frequency and the half-height interval
are found, the frequency which is directly in the middle of the
interval defined by the half-heights is used as the "maximum."
If, however, the half-height interval can not be determined, the
absolute maximum can be used as the "maximum." From the
maximum frequency, the fundamental frequency (e.g., the heart beat)
of the heart can be determined.
From the Fourier transformation, the software can determine the
fundamental frequency of the heart and generate images of the heart
in different phases of the heart cycle. As will be described below,
the user can display a plurality of projection images of the heart,
in which each of the images corresponds to a different phase of
the heart cycle.
Because it is difficult to extract the phase information present
in the Fourier spectrum, the Fourier transformation does not inform
the user as to which slices represent the diastolic phase, systolic
phase, and the like. Moreover, such a transformation does not account
for irregular heartbeats or a changing of the heartbeat over the
image acquisition period. In order to determine which slices correspond
to the diastole, the software of the present invention can analyze
the slice images to find the biggest heart volume image (e.g., the
diastole) in which the heart motion is the least.
To determine the phase of each of the slices, (e.g., to determine
which slices correspond to diastole), a local intensity signal of
the slice images can be run through a derivative filter to produce
a graph such as FIG. 13. Generally, this method can be used in conjunction
with the results from Fourier analysis, as described above, to find
the size of the heart in each of the slice images. With the frequency
derived from Fourier analysis and phase from the local maxima, slice
selection can be extended beyond the ROI of Step 206. It should
be appreciated however, that it may be possible to use the local
intensity profile as an independent algorithm. In such embodiments,
the user would need to cover all slices with the selected region
of Step 206.
In such an analysis, as illustrated in FIG. 13 each local peak
220 in the intensity signal corresponds to the maximum inflation
of the heart. The peaks can be located through a differential analysis
with the differential filter in which each peak (i.e., local maximum)
has a first derivative of zero and a second derivative that produces
a zero crossing response.
From the filtered data, the zero-crossings can be located. A crossing
from a negative number to a positive and back to a negative corresponds
to a maximum. Crossing from a positive to a negative and back to
a positive corresponds to a minimum. It should be appreciated however,
that the signs of the zero-crossings are dependent on the sign of
the second derivative filter, which as described above was fixed
to be negative-positive-negative. From the zero crossing intervals,
the location of the maximum intensity values are found and the slices
in which the heart is in diastole are chosen.
Post-processing of the maxima found above can proceed in several
passes over the slice selection. As an initial step, the distance
between two adjacent selected slices will be checked to determine
if the slices are too close together. In one configuration, the
slices will be deemed to be too close if they are within one third
of the heart-rate frequency found by the Fourier transformation,
this being a reasonable limit for how much the heart rate may change
during the study. It should be appreciated however, that in other
configurations, a smaller or larger frequency distance can be used.
If the slices are deemed to be too close, the slice that has the
lower intensity value will be removed from the image set of selected
slices.
Next, for each selected slice, the algorithm can resample the images
to verify that at least two of the slices' four neighbors are within
30% of the heart rate measured by the Fourier analysis so as to
avoid irregular spacing. If the slice is outside of the 30% range,
the slice will be deleted from the set of selected slices. It should
be appreciated however, that it may be possible to use a criteria
different criteria (i.e., smaller or larger than 30% of the heart
rate), if desired.
Thereafter, the algorithm can resample the images to check the
spacing between the remaining slices to see if there are any gaps
that are bigger than the slice thickness (which is combination of
the thickness of the slice for a stationary scan and the broadening
introduced by the travel of the patient bed during the helical scan).
If there is such a gap, the gap can be filled in with a slice of
maximum intensity in FIG. 13 in the location of the gap. It should
be noted, that it is preferable to have the slices be spaced so
as not to leave gaps not covered by the slice thickness, as noted
above. If there are any slices between two slices that are within
the heart rate found by the Fourier analysis, the slices are deleted
from the set of selected slices.
Generally, the derivative filter algorithm will only cover the
selected region of the scan that was marked by the user. Thus, if
the user did not select the entire image additional slices need
to be selected. If slices need to be added, a pseudo-selection of
slices can be generated on each end of the selection region. The
generated slices will be spaced by the frequency found by the Fourier
analysis. A cross-correlation at various offsets can be performed
to obtain the best estimate of the phase for extension of the frequency
information. The offset that returns the biggest correlation value
is used to extend the dataset to complete the image. The same cross-correlation
algorithm can be applied to the pseudo selection slices, as described
above.
The computed heart rate can then be used to generate multiple slice
subsets from the original set of slice images, in which each of
the slice subsets correspond to a different phase of the heart cycle.
(Step 212). The present invention can use software to efficiently
select as many sets as there are redundancy, and present them to
the user for selection. Multiple selections can be generated from
the frequency but at different phases of the cardiac cycle to give
the user the choice to select one. There are (1/frequency*1/time
between slices) different offsets from the first slice in the original
image set. The software program selects the i.sup.th slice as offset+i*(1/frequency*1/time
between slices) so as to result in 1/frequency*1/time between slices
subsets of the original scan. The user can choose the desired set
from these.
Having the fundamental heart rate, however, is not sufficient for
the best selection of slices since the fundamental heart rate does
not explicitly define which slices correspond to the diastolic phase.
Thus, to select the images that were obtained during diastole, the
heart frequency information can be used along with the information
obtained with the derivative filters to obtain time and phase information
to generate an image in which the heart is at its largest volume
(e.g., diastole). (Step 214).
Additionally or alternatively, the plurality of images of the heart
can be ranked by applying a quality measure so as to rank the images
based on heart size. (Step 216). One quality measure algorithm comprises
summing all of the pixel intensity values over a certain threshold
value. The intensity value is normalized by the total number of
pixels in the image to provide the average intensity value of the
image. Thereafter, each of the average intensity values of each
of the images are compared to rank the images relative to each other.
Another quality measure algorithm counts the number of pixels above
a threshold value. The number of pixels above the threshold is normalized
by the number of pixels in the image to provide a fraction. The
fraction can identify the percentage of the image that the heart
occupies in the image. Generally, the higher the fraction, the better
the selection. Thereafter, the fractions of each of the generated
images are compared and ranked relative to each other. It should
be appreciated however, that other quality measure algorithms can
be used to rank the images of the heart.
Thereafter, the images of the heart can be displayed on a computer
output display in order of rank so as to allow the user to select
the phase most appropriate for the scoring of each vessel within
the heart. (Step 218). Alternatively, it is possible for the software
to automatically display only the image with the highest rank.
In some methods, the software of the present invention can be used
to auto correlate between image pairs and computes the quality of
the correlation. Times of slow motion produce better correlations
than when the motion is rapid. A repetitive pattern can be established
from which the quiescent times are selected to create the gated
image set. Advantageously, the same graphical user interface of
FIG. 2 can be used to gate the image scan.
FIG. 15 illustrates a graphical user interface that can be used
to self-gate the set of image slices without the use of an ECG signal.
As shown in FIG. 15, menu toolbar 53 can include additional buttons
"Edit," "Clear Selection," and "Self Gate"
that allows the user to self gate the image scan. The software allows
the user to delineate the regions of the heart where the heart motion
can be visually observed.
With the "Edit" button 110, the user can enter an editing/drawing
mode in which the user can draw a boundary around a region of the
image of the heart and mark it. The region can be selected by at
least two different manners. A first manner is through a straight
line selection, in which the user selects a first end point of a
straight line and a second end point of the line to define the region.
In selecting the region, the region must have a minimum length
across the slice direction and a maximum length within the direction
of one slice. If the selected region is too small to obtain enough
information for analysis (e.g., less than about three seconds) or
too wide so that the signal is lost because of scanner rotation
(e.g., more than approximately half of the image width), the software
of the present invention can provide the user with an error message
to prompt the user to select a different region and to prevent the
computation of a heart-rate from unsuitable data.
In exemplary embodiments, a left click of the mouse defines the
first endpoint, and a right click of the mouse defines the second
point. The line drawn by the user will be used to define a diagonal
of a rectangular region. In a second manner, the user can use a
freehand selection, which allows the user to select a region of
arbitrary shape. In one embodiment, the user can depress a "Control"
key on a keyboard of a computer system and move the mouse to draw
the region of interest into the arbitrary shape. Releasing the control
key closes the region. It should be appreciated however, that the
above methods of drawing the region are merely examples, and other
conventional methods of drawing/selecting the region can be used.
The region can be a portion of one border of the patient's heart.
Advantageously, drawing the region around multiple portions of the
border of the heart allows the user to see and track differentially
the motion of the heart through the different portions of the heart
cycle. Thus, the user can view the different chambers of the heart
as it moves through the R-R cycle.
The "Clear Selection" button 112 can delete a region
that was previously marked by the user. The "Self Gate"
button 114 starts the self gating procedure that is described herein.
Referring again to FIGS. 10-12, to self gate an image scan, the
user marks selected section(s) in the sagittal view or the coronal
view on one side of the heart where the motion can be seen. Motion
of the heart will be shown by the jagged edge of the heart in the
sagittal image and coronal image. Straight line boundaries 118 (FIG.
11) or freehand boundaries 120 (FIG. 10) can be drawn on one or
more portions of an edge of the heart. If desired, the user can
select multiple regions.
Once the regions have been selected, the user can click on the
Self Gate button 114. The self gate software can them compute the
average frequency of the heart beat using the information of the
selected region and generate a number of selection. FIG. 16 schematically
illustrates an exemplary data flow of the present invention.
FIG. 17 shows one preview screen graphical interface for displaying
the multiple selections. One selection represents the biggest heart
volume based on the frequency information of the selected region
and the intensity values in the marked regions. The other selections
are based only on the frequency information. Each selection corresponds
to a different phase of the measured heart frequency.
As shown in FIG. 17, the preview screen 129 includes a main preview
window 130 having the current selection. The default selection is
derived from the selection that includes the intensity and frequency
information. A smaller image 132 of the current selection can also
be displayed alongside the right portion of the graphical user interface
in one of the small preview windows and can be framed by colored
frame 134. Clicking or otherwise selecting on another image alongside
the small preview windows (e.g., the right portion of the graphical
user interface) will display the selected image on the main preview
window. The topmost image 132 shows the selection inferred by intensity
and frequency information derived from the image scan. The remaining
images show selections that use only the computed heart frequency
at different offsets (e.g., different phases of the heart's motion).
As seen in FIG. 17, many of the images at the different phase of
the heart has noticeable blurring due to the motion of the heart.
Nonetheless, providing a plurality of images of the heart in different
phases allows the user to visually determine which heart image is
best.
The user can select different projection of the current preview
by activating the Axial button 136, Coronal button 138, or Sagittal
button 140. The slider 142 can be activated by the user to scroll
through the slice projections, if desired. When the user finds a
projection image that is acceptable, the user can click on the OK
button 144, which will apply the current selection and return to
the main screen of the graphical user interface (FIG. 2).
As described above, once the selected image set is deemed acceptable,
the image set can be calcium scored and saved. Before saving the
slices as a new DICOM series, the selection of images can be checked
for gaps. If there are gaps, the number of gaps can be reported
to the user with their size range. The user can then select to ignore
the gaps or can elect to fill in the gaps that are bigger than a
specified threshold, which the user can specify.
One method of filling in the gaps is with a slice closest to the
middle point between the selected slices. Of course, these gaps
may also be filled through low order interpolation algorithms such
as nearest neighbor, and in increasing order, linear, cubic and
so on, or Fourier interpolation.
In another aspect, the present invention provides improved methods
and software for calcium scoring the images. Retrospective gating
often causes mismatches between the scanner rotation and the heart
rate. Consequently, the selected images may not always be equally
spaced such that there are gaps between the images. Most calcium
scoring algorithms, however, are based on algorithms that require
a fixed spacing between the slice images.
Unfortunately, conventional linear or other low order interpolation
schemes that can be used to generate equally spaced slice images
from the selected merely blur the images, which degrades the calcium
scoring of the images. The present invention provides a Fourier
Interpolation that can rescale the dimensions of the image slices
that does not introduce blurring or degrade the resolution. A more
complete description of Fourier Interpolation can be found in U.S.
Pat. Nos. 4,908,573 and 5,036,281 and in Kramer D. M., Li A, Simovsky
I, Hawryszko C, Hale J and Kaufman L., "Applications of Voxel
Shifting in Magnetic Resonance Imaging," Invest Radiol 25:1305,
1990, the complete disclosures of which are incorporated herein
by reference.
While all the above is a complete description of the preferred
embodiments of the inventions, various alternatives, modifications,
and equivalents may be used. Although the foregoing invention has
been described in detail for purposes of clarity of understanding,
it will be obvious that certain modifications may be practiced within
the scope of the appended claims.
|