Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
20,?~3~
15~M03463
A M~THOD FOR COMBINING ACQUIRED NMR
DA~A TO SUPPRESS ~CT~OU AR'IFACTS
Backaround o~ the Invention
I This invention xelates to nuclear magne~ic resonance
¦ imaging methods. ~ore specifically, this invention relates
I to a method for controlling image artifacts caused by -~
substantially periodic NMR signal variations due, for
example, to subject motion in the course of an NMR scan.
NMR has been developed to obtain images of anatomical
features of human patients. Such imageQ depict nuclear 3p n
distribution (typically, protons assoclated with water and ;
!~ 15 tissue), spin-lattice relaxation time T1, and/or spin-spin
relaxation time T2 and are of medical diagnostic value. ~MR
' data for constructlng images can be collected using one of
; many available techniques, such as multiple angle projection
reconstruction and Fourier transform (FT). Typically, such
techniques comprise a pulse sequence made up of a plurality
of sequentially implemented views. Each view may include one
or more NMR experiments, each of which comprises at least an -
RF excitation pulse and a magnetic field gradient pulse to
encode spatial information into the resulting NM~ signal. As
. 25 is well-known, the NMR signal may be a free induction decay
(FID~ or, preferably, a spin-echo cignal.
The preferred embodiments of the invention will be
described in detail with reference to a variant of the well
" ~
2 ~ 3
15NM03463
~nown FT technique, which is frequently re~erred to as ~spin-
warp~. It will be recognized, however, that the method of
the invention is not limited to FT imaging methods, ~ut ~ay
be advantageously practiced in conjunction wi~h other
S techniques, such as multiple angle projection reconstruction -
disclosed in U.S. Patent No. 4,471,306, and another variant
of the FT tech~ique disclosed in U.S. eatent No. 4,070,611.
The spin-warp technique is discussed in an article entitled ; :
"Spin Warp NMR Imaging and Applications to Human Whole Body ~ -
Imaging~ by W. A. Edelstein et al., ~ L~ Y:~ 91
Biolo~y, Vol. 25, pp. 751-756 (1980).
Briefly, the spin-warp technique employs a variable
amplitude phase encoding magnetic fleld gradient pulse prior
to the acquisition of NMR spin-echo sign~ls to phase encode
lS spatial information in ~he direction of this gradient. In a
two-dimensional implementation ~2DFT), for example, spatial ~
information is encoded in one direction by applying a phase~ ;
encoding gradient ~Gy) along that direction, and the
observing a spin-echo signal in the presence of a second
magnetlc field gradient ~Gx) in a direction orthogonal to the
phase-encoding directlon. The Gx gradient present during the
spin-echo encodes spatial information in the orthogonal
dlrection. In a typical 2DFT pulse sequence, the magnitude
of the phase-encoding gradient pulse Gy is incremented ~Gy)
25 monotonically in the sequence of views that are acquired to ~-
produce a set of NMR data from which an entire image can be
recon~tructed.
Ob~ect motion during the acquisition o NMR image data
produces both blurring and ~'ghosts~' in the phase-encoded
direction. Ghosts are particularly apparent when the motion
is periodic, or nearly so. For most physiological motion,
including cardlac and respiratory motion, each view of the
NMR signal is acquired in a period short enough that the
object may be considered stationary during the acquisition
,. . . , ~
, . . . . .
20~3~
15NM03463
window. Blurring and ghosts, therefore, are due primarily to --
the inconsistent appearance of the object from view to view,
and in partic~lar, due to changes in the amplitude and/or
phase of the NMR signal due to the motion.
Both blurring and ghosts can be reduced if the data
acquisition is synchronized with the functional cycle of the -~
object. This method is known as gated NMR scanning, and its
objective is to acquire NMR data at the same point during
successive functional cycles so that the object "looks" the
same in each view. The drawback of gating is that NMR data
may be acquired only durins a small fraction of the object's
functional cycle~ and even when the shortest acceptable pulse
sequence is employed, the gating technique can significantly
lengthen the data acquisition time.
One proposed method for eliminating ghost artifacts is
disclosed in U.S. Patent No. 4,567,893, issued on February 4,
l9a6, and is assigned to the same assignee as the present
invention. In this prior applica~ion, it is recognized that
the distance in the image between the ghosts and the object
being imaged is maximized when the NMR pulse sequence
repetition time is an odd multiple of one-fourth of the
duration of the periodic signal variatlon (if two phase-
alternated RF excitation pulses per view are used, as
disclo~ed and claimed in commonly assigned U.S. Patent No.
4,443,760, issued April 17, 1984). It i~ recognized that
this ratio can be used to alleviate ghosts due to respiratory
motion. While thi~ method, indeed, improves image quality,
it doe~ impose a constraint on the N~R pulse sequence
repetition time and it often results in a longer total scan
time. It also assumes that the motion is periodic. Its
effectivene~s is diminished when the subject's breathing is
irregular because the ghosts are blurred and can overlap t~.e
image.
~rJ`'~, ~
2 ~
-.:
;- .:
15NM03463 -
4 ;
Another method for reducing the undesirable effects due
to periodic signal variations is disclosed in U.S. Patent ~o. ~-
4,706,026 issued November 10, 1987 and entitled "A Method For
Reducing Image Artifacts Due To Periodic Variations In NMR
S Imaging." In one embodiment of this method, an assumpt-o~ ~s
made about the signal variation period (e.g. due, for
example, to patient respiration~ and the view order is
altered from the usual monotonically increasing phase- ~
encoding gradient to a preselected order. This involves -
10 establishing the order in which either the gradient ;~
parameters, i.e. the amplitude of the phase-encoding gradien
pulses (in the spin-warp method~ or the direction of the
read-out gradient pulses (in the multiple angle projection
reconstruction methodl are implemented. For a given signal
variation period, a view order is chosen so as to make the
NMR signal variation as a function of the phase-encoding
amplitude (or gradient direction) be at a desired frequency.
In one embod~ment, the view order is selected such that the
variation period appears to be equal to the total NMR scan ;
time (low frequency) so that the ghost artifacts are brough
as close to the object being imaged as possible. In another
embodiment (high frequency), the view order is chosen to make
the variation period appear to be as short as possible so as
to push the ghost artifacts as far from the object as
possible.
Thi-~ prior method is effective in reducing artifacts,
and is in some respects ideal if the variation is rather
regular and at a known frequency. On the other hand, the
method is not very robust if the assumption made about the
motion temporal period does not hold (~.g., because the
patient's breathing pattern changec or is irregular). If
this occurs, the method loses some of its effectiveness
because the focusing of the ghosts, either as close to the ~ ~-
object or as far from the object as possible, becomes ~-
s`:~
:. . ,-- : . . . , : : .
2 ~ 3 A~ ~ '
.
'
15NM03463
blurred. A solution to this problem is disclosed in U.S.
Patent No. 4,663,591 entitled "A Method For Reducing Image
Art facts Due To Periodic Signal variations in NMR Imaging ~t
In this method, the non-monotonic view order is determined as -~
the scan is executed and is responsive to changes in the
period so as to produce a desired relationship ~low frequency
or high frequency) between the signal variations and the
gradient param2ter.
While the above described methods reduce motion -
artifacts, they do rely on some regularity, or
predictability, in the cyclic motion which is not always
present. For example, irregularities may occur during a
breathing cycle which introduces errors into the data
acquired for one or more views of the scan. One method for
reducing such spurlous errors is to perform two scans, or
acquire twice the necessary data during a single scan and
then average the information acquired for each view to
produce an image of increased quality.
Summary of the Invention
The present invention relates to an improved method and
system for combining redundant NMR data for the purpose of
reducing artifacts caused by patient motion. More
specifically, the invention includes means for acquiring a
plurality of sets of NMR image data, acquiring with each set
of image data an associated set of motion data that indicates
displacement of the subject during each view of acquired
image data; means for generating a smooth reference curve for
each set of acquir~d motion data; meanQ for determining the
deviation of the acquired motion data from its reference
curve; and combining the NMR image data for each view in the
two sets of NMR image data a~ a function of the deviation of
its associated motion data. Rather than merely averaging ~he
15NM03463
redundant data for each view, the p~esent invention
contemplates giving more weight to data which is more likely
to provide reduced motion artifacts in the image, based o~
the deviation of its associated motion data from the
reference curve.
A general ob~ect of the invention is to reduce motio~ -
artifacts by intelligently combining redundant NMR image
data. The associated motion data provides an indication of
the integrity of the acquired NMR image data in each se~.
~ather than merely averaging the redundant data, therefore,
more weight is given to the NMR data with higher integrity.
The result is a reduction in motion artifacts in the image
which is reconstructed from the intelligently combined data.
Another object of the invention is to improve the
quality of images which are reconstructed from NMR data
acquired in a scan having a non-monotonic view order. The
strategies which use a non-monotonic view order presume t~.er
respiration ~ollows a smooth periodic cycle. To the extent
that this is not true, the strategy does not work and motion
artifacts are produced in the image. ~y acquiring redundant
NMR data and combining it according to the teachings of the
present invention, however, NMR data is produced which
enhances the artifact suppressing mechanism of non-monotonic
view ordered scans.
Tho foregoing and other objects and advantages of the
invention will appea~ from the following description. In the
description, reference is made to the accompanying drawings
which form a part hereof, and ~n which there is shown by way
of illustration a preferred embodiment of the invention.
Such embodiment doeq not neces~arily represent the full sco?e
of the invention, however, and re erence i-~ made therefore ;o
the claims herein for interpreting the scope of the
invention.
2 ~-~
15N~103463
Br~f Des~ri~tion of ;he Dr~in~
FLg. ~ is an electrical block diagram of an NMR system
which employs the present invention;
Fig. 2 is an exemplary imaging pulse sequence which s
S executed by the system of Fig. 1;
Fig. 3 is a preferred em~odiment of the motion phase
pulse sequence which is interleaved with the imaging pulse
sequence o~ Fig. 2;
Fig. 4 is a pictorial representation of the data
structures which are produced when practicing the preferred
embodiment of the invention; and ~ -
Fig. 5 is a flow chart of a program executed by the NMR
system of Fig. 1 to carry out the present invention.
Description of the Preferred Embodiment
Fig. 1 is a simplified block diagram of an NMR imaging
system which employs the preferred embodiment of the
invention. The system includes a pulse control module 112
which provide3 properly timed pulse waveform signals, under
the control of a host computer 114, to magnetic field
gradient power 3upplies collectively deslgnated at 116.
These power supplieQ 116 energize gradient coils which form
part of a gradient coil assembly generally indicated by block
118. The assembly contains coils which~produce the Gx, Gy and
Gz magnetic field gradients directed in the x, y, and z
direct~ons, respectively, of the Cartesian coordinate system.
The use of the Gx, Gy and Gz gradients in NM~ imaging
applications will be described hereinafter with reference tO
Fig. 2.
Continuing with reference to Fig. 1, the pulse control
module 112 provldes activating pulses to an RF synthesizer
120 which is part of an RE transceiver, portions of which are
2~,3~ ~
15NM03463
enclosed by dash-line block 122. The pulse control ~odule
112 also supplies signals to a modulator 124 which modulates
the out?ut of the RF frequency synthesizer 120. The
modulated RF signals are applied to an RF coil assembly 126
through an RF power amplifier 128 and a transmit/receive
switch 130. The RF signals are used to excite nuclear splr.s
in a sample object (not shown) which is to be imaged.
The NMR signals from the excited nuclear spins are
sensed by the RF coil assembly 126 and applied through the
transmit/receive switch 130 to an RF preamplifier 132. The
amplified NMR signals are applied to a quadrature phase
detector 134, and the detected signals are digitized by A/D
converter 136 and applied to computer 114 for storage and
processing.
Reference is made to Fig. 2 which depicts two views of 2
conventional imaging pulse sequence of the type known as ~o-
dimensional Fourier transforms (2DFT), and which is also
referred to as two-dimensional "spin-warp". This pulse
sequence is useful in obtaining, in a well known manner,
imaging NMR data to reconstruct images of an object being
investigated. The two views are indicated at ~A" and "B" and
they are identical with the exception of the phase-encoding
gradient field Gy~ ~ach view is a pulse sequence which
utilizes phase-alternated RF excitation pulses which, as
disclosed and claimed in the above-identified U.S. Patent No.
4,443,760, produce phase-alternated NMR signals Sl~t) and
S1'(t) to cancel certain baseline error~ in the NMR system.
ReSerring now to View A in Fig. 2, there is shown in -;
interval 1 (indicated along the horizontal axis) a selective
90 RF excitation pulse applied in the presence of a positive
Gz magnetic field gradient pulse. Pulse control module 112
(Fig. 1) provide3 the needed control signals to the frequenc-
~synthesizer 120 and modulator 124 so that the resulting
excitation pulse is of the correct phase and frequency to
2J ~ . d ~
15N~103463
excite nuclear spins only in a predetermined region of the
object being imaged. Typically, the excitation pulse can be
; amplitude modulated by a ~sin x)/x function. The frequency
of the synthesizer 120 is dependent on the strength of t~.e
S applied polarizing magnetic field and the particular NMR
species being imaged in accordance with the well known La~ or
equation. The pulse control module 112 also applies
activating signals to the gradient power supplies 116 tO
generate, in this case, the Gz gradient pulse.
; 10 Continuing with reference to Fig. 2, Gx, Gy and Gz
gradient pulses are applied simultaneously in interval 2.
The Gz gradient in interval 2 is a rephasing pulse typically
selected such that the time integral of the gradient wavefor~
over interval 2 is approximately equal to -'~ of the ti.~e
i 15 integral of the Gz gradient waveform over interval 1. The
function of the negative Gz pulse is to rephase the nuclear
; spins excited in interval 1. The Gy gradient pulse is a
phase-encoding pulse selected to have a different amplitude
in each of Views A, B, . . ., etc., to encode spatial
information in the direction of the gradient. The number O r
different Gy gradient amplitudeq is typlcally selected to
equal at least the number of pixel resolution elements the
reconstructed im2ge will have in the phase-encoding ~Y)
direction. Typically, 128, 256, or 512 different gradient
amplitudes Gy are selected and in the typical NMR system, ~..e
Gy values are incremented a fixed amount from one view to the
æ next until the NMR scan is complete.
The Gx gradient pulse in in~erval 2 is a dephasing pulse
needed to dephase the excited nuclear spins by a
30 predetermined amoùnt to delay the time of occurrence of a
spin-echo signal Sl(t) in interval 4. The spin-echo signal
is produced typically ~y the application of a 180 RF pulse
in interval 3. As is known, the 180 RF pulse is ~ pulse
which reverses the direction of spin dephasing so as to
? æ
., j ~ ~ . j
,''`~' . ~ , . . :
' `' . . ' ' `
.,.,~, . ' ` , . .
2~2~
. , .
15NM03463
produce the spin-echo signal. The spin-echo signal is
sampled in interval 4 in the presence of a sradient pulse Gx
to enc~de spatial infor~ation in the direction (x) of th s
~radient.
S As indicated above, baseline error components are
eliminated by using an addltional NMR measurement in eac:~
view. This second measurement is substantially identical tO
the first with the exception that the RF excitation pulse i-.
interval S of View A is selected to be 180 out of phase (as
suggested by the minus sign) relative to the excitation pulse
in interval l of view A. As a result, the spin-echo signal
Sl'(t) in interval 8 is 180 out of phase with the spin-echo
signal Sl(t) in interval 4. If the signal Sl'(t) is
subtracted from Sl(t), only those components of the signals
with reversed sign in the signal S1'(t) are retained. The
baseline error components thus cancel.
The process described above with reference to View A s
repeated for View B and so on for all amplitudes of the
phase-encoding Gy gradient. The NMR image data which is
collected during this scan is stored in the host computer 114
~ where it is processed to produce image data suitable for
i controlling a CRT display.
As the above described conventional NMR scan i
performed, NMR data is acquired from all physical locations ;~
25 within the plane, or slice, of the object being imaged. If ~;
an accurate image is to be reconstructed, then both the
object and the measurement conditions must be stable, or
fixed, during the time needed to complete the entire NMR
scan. The present invention deals with the very prac~ical
situations in which this is noe the case, but instead, the
measurement conditions change in some cyclic, or nearly
cyclic, manner.
One such situation occur~ when an image is to be ~ -
produced throu~h the abdomen of a human subject. In this
.
.,~. ..
' ! .. . . , :
2 ~
15NM03463
11
case, much of the material beLng imaged is in motion due to
the s~bject's breathinq, and the time needed to acquire the
NMR data for an entire image will often transcend many
respiration cycles. If NMR data is acquired continuously
S throughout the respiration cycles, the subject will be
disposed differently from view to view and the reconstructed
image will contain many motion artifacts.
Co-pending U.S. patent application serial no. 427,401
filed on October 10, 19a9 and entitled ~'A Method For
Monitoring Respiration ~ith Acquired NMR Data~ describes a
system in which the motion of the patient dl~e to respiration
is monitored using a special NMR pulse se~uence which is
interleaved with each view of the acquired image data. The
disclosure of this co-pending application is expressly
incorporated into this application. Referring again to Fig.
1, the motion NMR pulse sequence described in this co-pendir.g
application is executed by the pulse control module 112 jus~
prior to the execution of each view in the image scan. The
N~R signal which results is received by the computer 114 and
is analyzed as described in the above cited co-pending U.S.
patent application to produce a displacement value ~Y which
indicates the position of the patient's anterior abdominal
wall with respect to a reference position. This displacement
value ~Y is output by the computer 114 to a respiration
proce~sor 188 which converts it in real time to a value wnich
indicateq the current phase of the patient respiration cycle.
As descrlbed in the above cited U.S. Patent Nos. 4,663,591;
4,706,026 and 4,720,678, this phase value which is computed
by processor 188 is applied to the pulse control module 112
to select the orde~ in which the amplitudes of the phase-
encoding gradient pulses Gy are applied durlng the scan. In
other word~, the phase values produced prior to each view a_e
used by the pulse control module 112 to control the
particular non-mono~onic view order which is being used to
.~ ,, . .. . ~ . .
15N~103463
12
suppress artifacts caused by patient respiratiOn. Therefore,
to t~e extent that a phase value deviates from the ~rue, or
expeceed, respiration cycle of the patient, it can be
anticipated that the integrity of the image data which s
acquired during the subsequent view may suffer. More
particularly, the motion artifact suppression strategy may ~e
less effective and the data acquired durinq the subsequent
view may produce more artifacts in the reconstructed image.
By using the present invention, motion artifacts can be
further suppressed. As explained above, during a complete
scan at least one set of image data from a digitized NMR
signal is obtained at each phase encoding value, ox ~view
number". Associated with each such set of image data is a
displacement value ~Y which indicates the measured
respiration phase when the NMR image data set was acquired.
In addition, it is also possible to obtain more than one set
of NMR image data and an associated displacement value ~Y for ;
each view number. For example, the pulse sequence of Fig. 2
can be executed twice for each phase encoding value during
. ..
the scan, or a second scan can be executed. In either case,
redundant NMR image data is acquired and may be combined to
reduce noise and motion artifacts in accordance with the
present invention.
. . - .
It is, of course, common practice to combine redundant
NMR image data to reduce random noise. Such a com~ination is
accompllshed by taking the average value of the corresponding
data element~ in each array o~ acquired image data. For
example, if two values for each element have been acquired,
their values are added and the result i~ divided by two.
Such averaging provides a ~ reduction in random noise. This
well known method weights each value equally in arriving at
the average.
The present invention combines redundant NMR data to
improve image quality, but it weights the acquired values
~ . : , . . .
15NM03463
13
being combined as a function of their measured integrity. .~s
will now be explained in detail, the integrity of the NMR
data s measured by uslng the associated displacement values
~Y.
S Referring particularly to Fig. 3, the displacement
values ~Y for each view in two successive scans are plotted.
A "low frequency" sort as described in the above cited Paten.
Nos. 4,663,591; 4,706,026 and 4,720,678 was used to acauire
the associated image data, and as a result, the measured
displacement values should follow a slowly changing smooth
cur~e as indicated at 200 and 201. It is readily apparent
that the displacement values deviate from these smooth
curves, and it is a teaching of the present invention that
the extent of this deviation is an indication of the
propensity for motion artifact generation when the associated
image data is used to produce an image. Such deviations,
therefore, provide a quantitative measure of the integrity o
the image data. It should be noted that these deviations are
not necessarily measurement errors, but may instead be caused
by spurious variations in the patient's breathing pattern
which has disrupted the motion artifact suppression strategy.
Accordingly, the present invention combines the redundant NMR
image data to reduce both noise and motion artifacts by -
weighting the data as a function of the deviations in their
2S corresponding displacement values.
The invention is carried out by thç computer 114 under ~-
the direction of a program which is executed after the data
sets have been acquired. The operatlon of this program will
now be described with reference to the flow chart in Fig. 5.
As indicated in Fig. 4, the acquired data includes a two-
dimensional array of image data 205 and an associated one-
dimensional array 206 of displacement data such as that
illustrated in Fig. 3. Similar arrays 207 and 208 are stored
for the second scan, or second set of acquired NMR data. The ;~
2~23~
I5N~103463
14
lmage data and the associa~ed displacement data have been
sor~ed into sequential v~ew number order, with each row of
image data array 205 and 207 being associated with a
corresponding element in its respective displacement ar ay
206 and 208.
As indicated by process block S0 in Fig. 5, the fi-st
step of the processing is to fit a quadratic equation to ~he
values in the displacement arrays 206 and 208 using a well
known least-squares technique. This effectively establishes
the smooth curves 200 and 201 (Fig. 3) which serve as the
I reference from which deviation values can be computed. These
j are identified as reference functions rl(ky) and r2t~y). As -~
t indicated by process block 51, the scan with the smallest
I variance in displacement values from their reference curve is
.
~ 15 then computed as follows~
: ~...~
N
!' variance - (l/N) ~ (m(ky) - r(ky))2 (l)
~ky ~
~!
where: N - the total number of views in the scan, - ~-
m(ky) - values in the displacement array, and
r(ky) - values of the reference function.
::~
Absolute devia~ions between ~he displacement values
m(ky) and their reference values r(ky) are then calculated
subject to a lower threshold deviation e ~ . 003 in the
preferred embodiment)~ as indicated by process block S2.
That is, one-dimensional deviation arrays 209 and 210 are
25 produced by finding the absolute value of the difference .
m(ky) - r(ky). If the difference does not exceed the
threshold value ( e ), then the corresponding image data
elements are weighted equally (i.e., .5).
. .
.
2~<~
15N~S03463
Using the deviation values D1(ky) and D2(ky) in the
respective arrays 209 and 210, the image data in the arrays
205 and 207 is now combined. More specifically, each row of
data S1(ky) in the first image data array 204 is combined with
its corresponding row of data S2~ky) in the second image data
array 207 to form a row of data S(ky) in a combined image
data array 211 as indicated by process blocks 53 and 54.
S(~y) - Sl(ky) (W~(ky)) + S2~ky) (W2(ky)) (2)
where: W1(ky) = D2(ky)/(Dl(ky) + D2(ky)) (3)
W2(ky) - Dl~ky)/(Dl~ky) + D2(ky))
As indicated above, the weighting factors Wl and W2 are set
equal to .S in cases where both deviationq are less than the
threshold amount (e).
The resulting image data array 211 is then processed as
indicated at block 55 using the conventional reconstruction
technique (2DFT in the preferred embodiment) to produce a
two-dimensional display data array 212 which indicates the
intensity of each display pixel.
It should be apparent to those skilled in the art that
while two sets of data are acquired in the preferred
embodiment of the invention, the invention is also applicable
when three or more sets of data are acquired. Regardless of
the number of set~ acquired, the present invention can be
employed to select the best data or to weight the averaging
of the data to minimize motion artifacts.
