MAGNETIC RESONANCE IN MEDICINE 26,328-341 ( 1992)
`
`Partial RF Echo Planar Imaging with the FAISE Method. I.
`Experimental and Theoretical Assessment of Artifact
`PHILIPPE s. MELKI, FERENC A. JOLESZ, AND ROBERT v. MULKERN*
`Drpartment of Radiology, Brigham and Women’s Hospital, Harvard Medical School,
`Boston. Mussachusetts 021 15; and * Depurtment of Radiology, Children :F Hospital,
`Hurvard Medical School, Boslon, Massachusetls 021 15
`
`Received July 10, 1991; revised October 10, 1991; accepted October 14, 199 1
`
`The fast acquisition interleaved spin-echo (FAISE) method is a partial RF echo-planar
`technique which utilizes a specific phase-encode reordering algorithm to manipulate image
`contrast (Melki et ul., J. Mugr7. Rrson. imaging 1:319, 1991 ). The technique can generate
`“spin-echo” like images up to 16 times faster than conventional spin-echo methods. How-
`ever, the presence of T2 decay throughout the variable k-space trajectories used to manipulate
`contrast ensures the presence of image artifacts, especially along the phase-encode
`7;
`direction. In this work, we experimentally and theoretically examine the type and extent
`of artifacts associated with the FAISE technique. We demonstrate the existence of well-
`defined minima of phase-encode ghost noise for selected k-space trajectories, examine the
`extent of blumng and edge enhancement artifacts, demonstrate the influence of matrix
`size and number of echoes per train on phase-encode artifact, and show how proper choice
`of FAISE sequence parameters can lead to proton density brain images which are practically
`indistinguishable from conventional spin-echo proton density images. A comparison of
`contrast between FAISE and standard spin-echo methods is presented in a companion
`article referred to as 11. 8 1992 Academic Press, Inc.
`
`INTRODUCTION
`In MRI examinations, spin-echo sequences provide the primary images for lesion
`detection. The ability of partial RF echo-planar techniques to produce “spin-echo-
`like” images with standard contrast control in reduced acquisition times has been
`recently reported ( I ). The implications are that partial R F echo-planar techniques
`may complement or even replace conventional spin-echo sequences. However, it is
`important to appreciate the extent to which overall image quality is maintained with
`such methods and to understand the nature and extent of any associated image artifacts.
`This work aims to evaluate the typical image artifacts associated with a specific phase-
`encode reordering algorithm whose primary purpose is to offer T2 contrast manipu-
`lation. The theoretical and experimental analyses presented may be generalized to
`study alternative phase-encode reordering schemes. The specific correlation between
`contrast obtained with fast acquisition interleaved spin echo (FAISE ) and that obtained
`with equivalent spin-echo sequences is demonstrated in a follow-up work.
`
`MATERIALS AND METHODS
`Experiments were performed on 1.5-T GE SIGNA systems (General Electric Corp.,
`Milwaukee, WI ) equipped with actively shielded gradient coils and utilizing digital
`RF tranceiver systems.
`
`0740-3194192 $5.00
`Copynght B 1992 by Academic Press, Inc.
`All nghts of reproduction in any form reserved
`
`328
`
`General Electric Co. 1032 - Page 1
`
`

`
`THE FAISE METHOD: ARTIFACT ASSESSMENT
`
`329
`
`The FAISE sequence combines rapid acquisition relaxation-enhanced (RARE) se-
`quences (2-6) with a specific phase-encode reordering algorithm ( 1 ). Figure 1 depicts
`a generic RARE sequence consisting of one or more Carr-Purcell-Meiboom-Gill
`(CPMG) echo trains ( 7). Each echo is individually phase encoded, read out, and
`phase unwound prior to the next refocusing pulse. With a 16-echo train applied a
`total of 16 times ( 16 “shots”), 256 phase-encode steps may be acquired 16 times
`faster than with conventional spin-echo sequences. With echo spacings of 15 ms,
`acquisition windows of 8.2 ms and a TR of 2 s, eight slices may be generated in only
`34 s. Alternatively, the number of echoes per echo train N,, may be reduced (e.g., 12,
`8, 6, or 4) with the acquisition of Ny phase-encode steps ultimately requiring a total
`of N, = N,/N, shots.
`The specific phase-encode reordering algorithm associated with the FAISE method
`( 1 ) is illustrated in Fig. 2. For simplicity, only a four-echo, four-shot sequence gathering
`a total of 16 phase-encode values is illustrated. The vertical axes in the three diagrams
`represent T2-related signal intensity decay with echo number (Arabic numerals) within
`each of the four shots (Roman numerals). The horizontal axes in Fig. 2 are the phase-
`to K,,, . Figure 2a illustrates how the
`encode axes in k-space (8) ranging from - K,,,
`least T2-weighted images are obtained. Note that early echoes are phase encoded with
`the smallest magnitude k-values. Figure 2b depicts how the most heavily T2-weighted
`images are generated. Figures 2c and 2d illustrate how intermediate T2 weightings can
`be obtained.
`In the schemes illustrated in Fig. 2, the fabric of k-space is threaded with each shot,
`skipping Ny/(2N,) phase-encode steps between successive echoes. A range of 4 the
`total number of phase-encode steps are subsequently covered with each shot. In the
`case of intermediate T2 weighting (Figs. 2c and 2d), each shot covers sections of both
`the positive and negative phase-encode axis. It may also be seen from Fig. 2 that the
`critical parameter governing the T2 decay shape of a given k-space traverse is the
`
`21-
`
`23-
`
`I
`I
`
`I
`I
`
`I
`I
`
`I
`
`RF/
`Slgnal
`
`I
`I
`
`I
`
`I
`
`I
`
`I
`
`I
`I
`
`I
`I
`I
`I
`I
`
`I
`I
`
`I
`
`I
`I
`
`I
`I
`I
`I
`I
`
`I
`I
`
`I
`
`/ /
`
`,y
`
`2-t-
`
`I
`I
`
`I
`I
`
`I
`I
`
`I
`
`I
`
`FIG. 1 . Schematic diagram of the RARE pulse sequence
`
`General Electric Co. 1032 - Page 2
`
`

`
`330
`
`MELKI, JOLESZ, AND MULKERN
`
`a) IPS = 0
`
`- kmax
`
`t
`
`0
`
`-
`kmax
`
`‘P2 + +
`
`-
`kmax
`
`- kmax
`d) I P S = 2
`
`0
`t
`
`- kmax
`
`0
`
`kmax
`
`FIG. 2. K-space traverses as performed with the FAISE method.
`
`number of phase-encode values by which the k-space trajectory is shifted from the
`initial order depicted in Fig. la. We refer to this parameter as IPS, for initial phases
`skipped. The IPS values for each of the schemes in Fig. 2 are labeled in the figure.
`The algorithm illustrated in Fig. 2 is placed in concrete terms by mathematically
`associating each echo n , collected during a given shot m , a specific phase-encode
`gradient amplitude k( m,n) which will correspond to the acquisition of a given hori-
`zontal line in k-space. The phase-encode k-space increment /3 is the product of the
`proton gyromagnetic ratio y, the phase-encode gradient increment 6G, and the length
`of the phase-encode step T,,
`
`The parameter /3 is related to the field of view (FOV) through
`/3 = 4.rrNy/( FOV( N y - 1 )),
`for a square pulse
`/3 = 7r2Ny/(FOV(NY - I ) ) ,
`for a half-sinusoid pulse.
`
`General Electric Co. 1032 - Page 3
`
`

`
`THE FAISE METHOD: ARTIFACT ASSESSMENT
`
`33 1
`
`[31
`
`Each k-space value is defined by
`k(m, n ) = [ A ( m , n ) - ( N y + 1)/21P,
`where A ( m, n ) is given by
`A ( m , n ) = [ { N y / 2 - nNJ2 + /TI - 1 + 1PS)mod Ny] + 1
`for 1 < m G NJ2
`A ( m , n ) = [(Ny/2 + ( n - 2)Ns/2 + m - 1 + 1PS)mod Ny] + 1
`for N s / 2 < rit < N,.
`[4b]
`These relations define the correlation between each RF echo within each shot, as
`indexed by its m and n variables, and its corresponding phase-encode k-space value.
`The expressions are written in general enough form to accomodate any number of
`echoes N, used to acquire Ny phase-encode steps with N, shots. The Modulus function
`(mod) operates on the terms in curly brackets. The operator chosen variable IPS may
`range from 0 to Ny/2 and its choice ultimately dictates the T2 contrast. We now define
`a pseudo-echo time pTE, in terms of a set of specific IPS parameters through the
`relations
`
`[4a]
`
`1 < n < N,,
`
`for IPS = 0,
`pTE = 27
`for IPS = N,/4 + ( 1 2 - 1)Ns/2,
`pTE = n27
`151
`for IPS = Ny/2,
`pTE = Ne27
`where n, N,, N,, and Ny have been defined and 27 is the echo spacing within the
`CPMG trains.
`As mentioned above, T2-related signal loss between echo readouts leads to image
`artifacts along the phase-encode direction. To model these artifacts (6), a “double
`box with gap” one-dimensional profile of the form
`
`0
`1
`p ( y ) = O
`1
`0
`
`f o r y < - a
`f 0 r - a G y G - E
`f o r - t < y < t
`for c < y G a
`f o r y > a ,
`
`I61
`
`where a and e are in cm, was simulated with the expression
`
`where B ( m , n ) = A ( m , n ) - ( 1 + Ny/2). The function F i s given by
`F = [exp(ipB(m, n ) n ) - exp(ipB(m, n ) f ) ] e x p ( - n 2 ~ / T ~ ~ )
`+ [exp(iPB(m, n ) ~ ) - exp(-iPB(m, n ) a ) ] e x p ( - n 2 ~ / T ~ ~ )
`[ 8 ]
`and TZL and TZR are the T, values of the left- and right-hand boxes, respectively. This
`expression is used to simulate single-box profiles by setting t = 0 and equating the T2
`values of the left- and right-hand boxes.
`
`General Electric Co. 1032 - Page 4
`
`

`
`332
`
`MELKI, JOLESZ, AND MULKERN
`
`For each value of y ’ along the phase-encode direction (ranging from - 7r to T), Eq.
`[ 71 is used to calculate the complex signal intensity (and its magnitude) from the
`double-box density function in a manner similar to that performed in practice by an
`N, echo, N, shot FAISE sequence. The expression includes the effects of the T2 decay
`process as well as the various traverses of k-space that are available with IPS manip-
`ulations.
`Simulated profiles were obtained by implementing Eq. [ 71 on a VAX 8600 main-
`frame computer (Digital, Inc. Marlboro, MA). The simulations were performed with
`parameters closely matching those used to make FAISE images of NiC12 doped water
`phantoms. Phase-encode profiles obtained experimentally from the phantoms were
`compared with the corresponding simulations. The T2 values of individual phantoms
`were measured from CPMG image data sets and used as inputs for specific simulations.
`
`RESULTS
`Figures 3a-3d depict FAISE images of three NiC12 doped phantoms with T2 values
`of 44, 105, and 955 ms, respectively. The images shown were obtained with IPS values
`of 0, 4, 8, and 12 (Figs. 3a-3d, respectively). A 16-ech0, 16-shot sequence with a 2-s
`TR was used to acquire the 256 X 256 image matrices in 34 s. The windowing in Fig.
`3 has been adjusted to emphasize the ghosting artifacts along the phase-encode direc-
`
`FIG. 3. (a-d) FAISE images of T, phantoms acquired with a 16-echo, 16-shot sequence using different
`IPS values (clockwise from top left: IPS of 0,4, 12, and 8). From top to bottom, each image depicts three
`NiCI2 phantoms with T2 values of 44, 1 10, and 955 ms. Other imaging parameters used were a 15-ms echo
`spacing, a LO-mm-thick slice, and a 24-cm field of view.
`
`General Electric Co. 1032 - Page 5
`
`

`
`THE FAISE METHOD ARTIFACT ASSESSMENT
`
`333
`
`tion. Note that the overall ghosting appears much less in the IPS = 0 image (Fig. 3a)
`than in the IPS = 8 image (Fig. 3c). Note also that the shortest T2 phantom exhibits
`the most significant ghosting artifacts.
`A complete set of images with IPS values ranging from 0 to 128 was obtained for
`the T2 phantoms of Fig. 3 using the 16-echo, 16-shot sequence. The mean values of
`the phase-encode noise outside of the 44-ms T2 phantom and the 105-ms T2 phantom
`are plotted in Fig. 4a as a function of IPS. A decaying oscillatory pattern with a period
`of 8 IPS values is apparent. The mean value for the noise along the frequency-encode
`gradient axis is also plotted in Fig. 4a as a function of IPS and shows no such behavior.
`Computer simulations of the single-box phase-encode profiles were performed for a
`
`a
`
`04
`0
`
`b
`
`48
`
`d
`
`0 T2 = 105 insec
`
`0 T2 = 44 msec
`
`a Frequency Noise
`
`10
`
`32
`
`48
`
`64
`
`80
`
`96
`
`112
`
`1
`
`18
`
`IPS
`
`0 T2 = 105 msec
`
`-2 4
`0
`
`16
`
`32
`
`48
`
`64
`IPS
`FIG. 4. (a, b) Variation of the mean value of the phase-encode noise Np as a function of IPS. (a) Each
`measurement of Np was obtained adjacent to two different T , phantoms, imaged with a 16-echo, 16-shot
`FAISE sequence using a TR of 2 s and a I 5-ms echo spacing. (b) The corresponding computer simulations
`of the phase-encode noise values outside of single-box phantoms as a function of IPS.
`
`80
`
`96 112
`
`128
`
`General Electric Co. 1032 - Page 6
`
`

`
`3 34
`MELKI, JOLESZ, AND MULKERN
`16-echo, 16-shot sequence. Simulation parameters were a = 2.5 cm, p = 0.25 cm-’,
`27 = 15 ms, and input T2 values of 44 and 105 ms, all in close accord with the actual
`imaging parameters and phantom T2 values of the expenmental results in Figs. 3 and
`4a. The simulated profiles were calculated for IPS values ranging from 0 to 128. From
`each simulated profile, an estimate of the phase-encode noise was obtained by summing
`the magnitude calculated signal intensities outside of the box (specifically for y ’ values
`ranging from 0.75 to 3.1 radians). The results are plotted in Fig. 4b and clearly re-
`produce the decaying oscillatory behavior observed in Fig. 4a. These results demonstrate
`that it is possible to minimize phase-encode noise through appropriate selection of IPS.
`Figure 5a plots experimental phase-encode profiles of three phantoms with T, values
`of 24, 6 I , and 107 ms. Images from which profiles were extracted were obtained with
`a 16-echo, 8-shot FAISE sequence using an IPS value of 0. Figure 5b displays simulated
`profiles obtained using model parameters closely matching those used for the exper-
`imental profiles of Fig. 5a ( a = 2.2 cm; @ = 0.25 cm-’ ; IPS = 0; T?S of 24, 6 I , and
`
`a
`
`0 7 2 = 61 rnsec
`
`60
`
`b
`
`160
`110
`I’ho5e enrode rliiection
`
`A-A
`
`0 - 0
`
`I 0-0
`
`T2 = 107 rnsec
`T2 = 61 rnsec
`T~ = 2 1 rnsec
`
`c . _
`
`2 200
`
`-0 200
`- 1 200
`0 8 0 0
`F t 1 ci i e r n c t-w c (i 1 re^ t 1 on
`FIG. 5. (a, b) Effect of T2 on the phase-encodc profile of an IPS = 0 FAISE image. ( a ) Experimental
`profiles along the phase-encode direction ofdifferent T, phantoms. The corresponding computer simulations
`are depicted in (5b).
`
`1 800
`
`General Electric Co. 1032 - Page 7
`
`

`
`THE FAISE METHOD: ARTIFACT ASSESSMENT
`
`335
`
`107 ms with 27 = 15 ms). Both plots indicate how shortening the T2 for IPS = 0
`sequences increases blurring at the edges of the phantom. Note that IPS values of 0
`correspond to the least T2 weighting (shortest pseudo-echo times) that is obtained
`purely through phase-encode reordering and so represent the most “proton density”
`type of weighting possible with FAISE.
`Higher IPS values lead to more T, weighting (longer pseudo-echo times). Figure
`6a depicts phase-encode profiles of the 24-ms T2 phantom obtained from images
`acquired with IPS values of 12 and 20, IPS values which correspond to phase-encode
`noise minima for the 16-echo, 16-shot FAISE sequence ( pTE values of 30 and 45
`ms) . These profiles display edge enhancement features which become more prominent
`with increasing IPS. In Fig. 6b computer simulations of the phase-encode profiles for
`a 24-ms T, phantom obtained with parameters closely matching those used for the
`experimental profiles of Fig. 6a are presented. The simulations reproduce the edge
`enhancement features observed in the experimental profiles at these IPS values. It
`
`9 7 5 t
`
`a
`
`x
`+- ._
`Ln
`C
`Q,
`I -
`C
`0
`C
`CX m .-
`
`4 7 5
`
`A pTE = 30 msec
`
`pTE = 4 5 msec A
`
`-25
`25
`
`b
`
`75
`
`125
`
`I75
`
`225
`
`Phase encode direction
`
`0 pTE = 30 msec
`
`x
`Y ._
`
`I
`
`I
`
`-?5
`2.200
`
`~~
`
`~
`
`0.000
`1 100
`- 1 100
`1’ tho sc Encode Direct i rl n
`FIG. 6. (a, b) Phase-encode profiles ofa short T2 square phantom ( T, of 24 rns) obtained with IPS values
`of 12 and 20 with a 16-echo, 16-shot sequence. (a) Obtained from experiments. ( b ) The corresponding
`computer simulations which have been scaled vertically for comparison.
`
`2
`
`00
`
`General Electric Co. 1032 - Page 8
`
`

`
`336
`
`MELKI, JOLESZ, AND MULKERN
`
`should also be noted that this feature was consistently observed at higher IPS values and
`for IPS values other than those which yield well-defined minima of phase-encode noise.
`Figs. 7a-7c depict 256 X 256 images of four T2 phantoms ( T2 values of 24,40,6 1,
`and 107 ms) as acquired with a 16-echo, 16-shot sequence; an 8-echo, 32-shot sequence;
`and a 4-echo, 64-shot sequence; respectively. Figures 7d-7f depict the same phantoms
`imaged with 128 X 256 matrices. The smaller matrices were obtained by halving the
`number of shots in each of the sequences used to obtain the images in Figs. 7a-7c.
`The images are windowed so as to demonstrate the effects of matrix size and number
`of echoes on phase-encode artifact. Images acquired with a reduced number of echoes
`and/ or with larger phase-encode resolutions clearly demonstrate improved image
`quality. Figures 7a and 7e which were both acquired with 16 shots have similar image
`artifact. Likewise, Figures 7b and 7f were both acquired using 32-shot sequences and
`have comparable ghost patterns and edge blurring effects. Limitations in matrix size
`format have prevented us from further experimental analyses. However, simulated
`profiles of FAISE sequences using the same number of shots ( N , ) but acquiring a
`variable number of echoes ( N , ) , were performed. A 24-ms T2 phantom “imaged”
`with pTEs of 15 and 45 ms using 16-shot FAISE sequences acquiring 128, 256, and
`512 k-space lines were considered ( N , of 8, 16, and 32, respectively). The simulated
`
`FIG. 7. (a-f) 128 X 256 and 256 X 256 FAISE images obtained with 16, 8, and 4 echo sequences. The
`T2 values of the phantoms (from top to bottom) were 330, 107, 61, and 24 ms. Imaging parameters were a
`TR of 2000 ms, a pTE of 15 ms (IPS = 0), a 15-ms echo spacing, a 5-mm slice thickness, a 24-cm field of
`view, and one signal average.
`
`General Electric Co. 1032 - Page 9
`
`

`
`THE FAISE METHOD: ARTIFACT ASSESSMENT
`
`337
`
`profiles obtained from these 16-shot sequences (for each pTE separately) were all
`identical (results not shown) indicating that the number of shots is a key determinant
`of specific phase-encode image artifacts.
`Figure 8 illustrates how the number of echoes influences the overall image quality
`of normal head FAISE images for a fixed matrix size. Figures 8a-8c present proton
`density FAISE images all acquired with a TR of 2000 ms and a pTE of 15 ms (IPS
`of 0), but with a 16-echo, 16-shot sequence; an 8-echo, 32-shot sequence; and a 4-
`echo, 64-shot sequence; respectively. Figure 8d is the first echo image (TE = 15 ms)
`obtained from a 16-echo CPMG sequence using the same imaging parameters as the
`FAISE sequences (TR of 2000 ms, 5-mm slice thickness, and a 22-cm FOV). Note
`how closely the 4-echo, 64-shot FAISE image resembles the conventional spin-echo
`proton density image whereas the use of 8 or 16 echoes results in characteristic artifacts
`including overall blurring and edge enhancement at CSF/ brain tissue interfaces ( I ) .
`In order to appreciate the extent of artifacts taking place at the interface between two
`different mediums, double-box simulations were performed for typical image weightings
`for both FAISE and SE sequences. Figures 9a-9c depict simulated double-box profiles
`( a = 6 cm, c = 0 cm). The T2 of the left-hand box was 100 ms and the T2 of the
`right-hand box was 1000 ms. The simulations were performed with an IPS of 0 (pTE
`= 15 ms) for a 16-echo, 16-shot sequence; a 4-echo, 64-shot sequence; and a I-echo,
`256-shot SE sequence. Characteristic edge interface distortions are noted with the 16-
`shot sequence (Fig. 9a). These artifacts are minimized by reducing the number of
`echoes with the 64-shot sequence (Fig. 9b) generating profiles practically indistin-
`guishable from those of the SE sequence (Fig. 9c). Similar simulations were performed
`for longer pTE/TE values (75 ms) and the differences in profiles between a 16-echo,
`16-shot sequence; a 4-echo, 64-shot sequence; and a I-echo, 256-shot SE sequence
`were minimal (results not shown).
`
`DISCUSSION
`
`The FAISE method ( 1 ) introduces a specific phase-encode reordering algorithm
`into RF refocused echo-planar techniques (2-6). The proper choice of phase-encode
`reordering, number of echoes per train, and echo spacing all lead to rapid k-space
`coverage without sacrificing the full range of contrast options associated with conven-
`tional spin-echo images. The price for the increased speed of partial RF echo-planar
`techniques over conventional spin-echo methods lies largely in image degradation
`caused by T2 relaxation between phase-encoding steps (6). In this study, we examined
`the image artifacts associated with FAISE. Though the results obtained apply specifically
`to the phase-encode reordering algorithm formulated by Eq. [ 41, they may be gen-
`eralized to encompass arbitrary reordering schemes.
`Figure 3 demonstrates how the phase-encode ghosting patterns are a distinct function
`of IPS and are more pronounced as the T2 of the material is decreased. This ghosting
`artifact was quantified by measuring mean values for phase-encode noise. The results,
`presented in Fig. 4, reveal that specific IPS values may be chosen to minimize ghosting
`effects and that overall noise values decrease with increasing IPS. The agreement be-
`tween experimental measurements and theoretical calculations of phase-encode noise
`with IPS number (Figs. 4a and 4b) indicates that the factors responsible for the noise
`
`General Electric Co. 1032 - Page 10
`
`

`
`338
`
`MELKI, JOLESZ, AND MULKERN
`
`FIG, 8. (a-d) Comparison between FAlSE head images acquired with different number of echoes but
`fixed matrix size. (a-c) 256 X 256 FAISE images acquired with 16 echo, 16 shots (a), 8 echo, 32 shots ( b ) ,
`and 4 echo, 64 shots (c). (d) The first echo image obtained from a 16-echo CPMG sequence (TE = 15 ms,
`256 X 256 matrix). Each image was obtained with a pTE/TE of 15 ms, a TR of 2000 ms, a 5-rnm slice
`thickness, a 24-cm FOV, and one signal average.
`
`General Electric Co. 1032 - Page 11
`
`

`
`THE FAISE METHOD: ARTIFACT ASSESSMENT
`
`339
`
`.~ c c x
`-
`-
`
`oi .-
`m
`
`20 30'
`
`l o t
`
`.
`
`7
`
`i
`
`7
`
`
`
`16 echo 16 shot '-
`pTE = 15 m s
`
`-104
`-3.200
`
`- 1.600
`
`0.000
`
`1.600
`
`3.200
`
`7
`
`FT Variable
`
`40
`
`30
`
`10
`
`-
`
`4 echo 64 shot
`
`-3.200
`
`-1 600
`
`0.000
`
`1600
`
`3 200
`
`FT Variable
`
`-- f
`.-+
`
`30
`
`20
`--
`
`--
`l o
`
`0
`-*
`
`- l o +
`
`I.c
`
`SE
`TE = 15 rns
`
`'-
`
`FIG. 9. (a-c) Double-box profile simulations mimicking brain tissue/CSF interfaces for proton density
`acquisitions using a 16-echo, 16-shot sequence (a), a 64-shot, 4-echo sequence (b), and an SE sequence ( 1
`echo. 256 shots). Note the interface artifacts for the 16-echo, 16-shot sequence.
`
`General Electric Co. 1032 - Page 12
`
`

`
`340
`
`MELKI, JOLESZ, AND MULJSERN
`
`are well accounted for when T2 decay processes are incorporated into the FAISE k-
`space trajectories.
`For a 16-echo, 16-shot sequence, minimal phase-encode noise is obtained for IPS
`values of 4 + 8N and maximal noise appears at IPS values of 8N ( N a n integer). The
`periodicity with IPS of the maxima and minima noise values is reduced to 4 when a
`16-echo, 8-shot sequence ( 128 matrix) is considered (experiments and simulations
`not shown). In general, this periodicity is half the number of shots. The illustrations
`of the k-space trajectories in Fig. 2 provide a means for understanding this periodicity.
`These trajectories are represented by decaying step functions. Each step is composed
`of NJ2 echoes with equal T2 weighting. The minimum noise IPS values correspond
`to those trajectories which place the origin of k-space at the center of a step rather
`than at the edge of a step (see Fig. 2c as opposed to Fig. 2d). Subsequently, for the
`contrast studies presented in 11, we restrict our attention to pseudo-echo times which
`utilize the IPS values indicated in Eq. [ 51.
`Aside from overall noise considerations, specific edge artifacts can be identified on
`FAISE images which are also related to the T2 decay during the k-space trajectory.
`Chief among these are a blurring artifact and an edge enhancement artifact. The
`blurring artifact and its dependence on T2 for IPS = 0 values is demonstrated exper-
`imentally in Fig. 5a and confirmed in the simulations of Fig. 5b. Both the signal
`intensity and loss of edge definition have a predictable dependence on T 2 , a result
`previously reported for “single-shot” RARE sequences (5, 6). Edge enhancement
`artifacts are encountered when IPS values greater than 0 are used. This is demonstrated
`experimentally in Fig. 6a with a short T2 phantom and confirmed by the simulations
`of Fig. 6b.
`Some general considerations concerning matrix size and number of echoes per
`CPMG train with regard to phase-encode noise and overall artifacts can be made from
`the images and simulations displayed in Figs. 7-9. First, for a given phase-encode
`resolution, overall image quality increases with decreasing the number of echoes per
`train, the limit being the conventional spin-echo ( N , = 1 ) sequence. Second, when
`the number of echoes per CPMG train are fixed, then reducing the number of phase-
`encode steps results in image degradation (Fig. 7). More generally it is found that the
`number of shots governs the extent of T2-related artifact. Indeed, simulations show
`that phase-encode profiles are identical when the number of shots and IPS are fixed
`and the number of echoes is varied. These considerations lead to practical compromises
`between data acquisition times-as dictated by the number of shots-and
`image qual-
`ity. For T2-weighted images, the use of 16-shot sequences provides excellent compar-
`isons with conventional long TR, long TE images ( 1 ) . However, this compromise
`between the number of shots and the overall image quality is more critical for proton
`density (short pTE) image weightings than for more heavily T2-weighted imaging.
`Indeed, to acquire a proton density FAISE brain image which can hardly be distin-
`guished from its conventional SE counterpart (Figs. 8c and 8d) , an increased number
`of shots (e.g., 64) must be considered. Short pTE images acquired with 8 or 16 echo
`trains suffer from both blurring and edge enhancement artifacts to a degree that prob-
`ably makes them unfit for diagnostic use. The bright/dark borders of the ventricles
`in the short pTE 16-echo FAISE image is a consequence of the edge enhancement
`artifact at CSF/ brain tissue interfaces, a feature confirmed by the double-box profile
`
`General Electric Co. 1032 - Page 13
`
`

`
`THE FAISE METHOD ARTIFACT ASSESSMENT
`
`34 1
`
`simulations of Fig. 9a. The extent of this artifact is substantially reduced for the 4-echo,
`64-shot simulation, which hardly differs from the SE simulation (Figs. Yb and Yc).
`Finally, we point out that the periodic phase-encode artifacts due to the stepwise
`T2 attenuation of signals might be minimized somewhat if echo times for the signal
`readouts were incremented between shots. However, fundamental T, decay mecha-
`nisms are intrinsic to the sequence and would continue to play a dominating role in
`image artifact. Reducing the echo spacing can reduce these fundamental artifacts,
`though this would require reduced echo readout times with increased bandwidth and
`subsequent loss in signal to noise.
`
`CONCLUSIONS
`Fundamental artifacts associated with the FAISE methods are demonstrated to be
`directly related to the T2 decay process occuring during data collection. These artifacts
`are characterized by ghosting effects and are responsible for increased phase-encode
`noise and deteriorations in edge definition. Within the framework of the specific phase-
`encode reordering scheme utilized, well-defined IPS values corresponding to a select
`set of pTEs may be chosen which minimize phase-encode noise. For a given image
`weighting, the number of shots governs the extent of overall image artifacts and should
`remain above a certain acceptable limit while varying the number of echo and/or the
`phase-encode matrix resolution. As demonstrated for proton density brain images, a
`judicious choice of both IPS and the number of echoes per train can lead to the
`generation of FAISE images which are practically indistinguishable from conventional
`spin-echo images, though acquired in substantially reduced acquisition times.
`
`ACKNOWLEDGMENTS
`
`This work was supported in part by l’institut National de la Sante et de la Recherche Medical (INSERM,
`France) and by NIH grant PO1 CA 41 167. One of us (P.S.M.) is grateful for the financial support obtained
`from the INSERM institut while on leave from the Unite 316 de I’INSERM (C.H.U. de Tours, France).
`We would like to thank John Listerud (Dept. of Radiology, Hospital of the University of Pennsylvania,
`Philadelphia) for helpful discussions we had during the course of this work.
`
`REFERENCES
`1. P. S. MELKI, R. V. MULKERN, L. P. PANYCH, AND F. A. JOLESZ, J. Magn. Reson. Imaging 1, 3 19
`(1991).
`2. J. HENNIG, A. NAURETH, AND H. FRIEDBURG, Magn. Reson. Med. 3,823 ( 1986).
`3. J. HENNIG, H. FRIEDBURG, AND B. STROBEL, J. Comput. Assist. Tomogr. 10, 375 ( 1986).
`4. J. HENNIG, Mugn. Reson. Imaging 78, 397 ( 1988).
`5. R. V. MULKERN, S. T. S. WONG, C. WINALSKI, AND F. A. JOLESZ, M a p Reson. Imaging 8, 5 5 1
`(1990).
`6. R. V. MULKERN, P. S . MELKI, P. JAKAB, N. HIGUCHI, AND F. A. JOLESZ, Med. Phys. 18, 1032 ( 1991 ).
`7. S. MEIBOOM AND D. GILL, Rev. Sci. Instrum. 29, 688 ( 1958).
`8. D. B. TWIEG, Med. Phys. 10,610 (1983).
`
`General Electric Co. 1032 - Page 14