Original Research
`
`Comparing the FAISE Method
`withConventiona1 Dual-Echo
`Sequences1
`
`Philippe S. Melki, MD Robert V. Mulkern, PhD
`Lawrence P. Panych, MS
`Ferenc A. Jolesz, MD
`
`The FAISE (fast-acquisition interleaved spin-
`echo) technique consists of a hybrid rapid-acqui-
`sition relaxation-enhanced (RARE] sequence
`combined with a specific phase-encode reorder-
`ing method. Implemented on a 1.6-T unit. this
`multirrection, high-rerolution technique permits
`convenient contrast manipulation similar to that
`of spin-echo imaging, with selection of a pseudo-
`echo-time parameter and a TR interval. With a
`TR of 2 seconds, eight 266 X 266 images are ob-
`tained in 34 second. with either T2 or proton-
`density weighting. A direct comparison between
`FAISE and spin echo for obtaining T2-weighted
`head images in healthy subjects indicates that
`FAISE and spin-echo images are qualitatively
`and quantitatively similar. Image artifacts are
`more pronounced on “proton-density” FAISE
`images than on the T2-weighted FAISE images.
`T1 contrast can be obtained with inversion re-
`covery and short TR FAISE images. Preliminary
`temperature mcamrements in dine phantoms
`do not indicate u c d v e temperature increases
`with extended FAISE acquisitions. However, ex-
`tensive studies of radio-frequency power depocli-
`tion effects should be performed if the FAISE
`technique i. to be fully exploited.
`
`Index terms: Head, MR studles. 10.1214. Physics Pulse se-
`quences Rapid imaging
`JMRI 1881; 1:319-326
`
`Abbrcviatiorm C/N = contrast-to-noise ratio. CPMG = Carr-
`Purcell-Meiboom-Gl11, FAISE = fast-acquisltion interleaved
`spin-echo. IPS = initial phases skipped. pTE = pseudo echo
`time. RARE = rapid-acquisition relaxatlon-enhanced. RF = ra-
`dio frequency. TI = Inversion tlme.
`
`’ From the Department of Radlology. Brigham and Women’s
`
`Hospital. Harvard Medical School. 75 Francis St. Boston. MA
`02115(P.S.M..L.P.P..F.A.J.):and theDepartmentofRadiolcgy.
`Children’s Hospital. Boston (R.V.M.). Recelved October 4, 1990:
`revlslon requested November 13; revislon recelved December
`26: accepted January 7 , 1991. Addmu reprint requmtm to
`P.S.M.
`0 SMRI, 1991
`
`THE DEVELOPMENT OF FAST IMAGING methods
`has been the focus of increased interest since the
`early stages of clinical magnetic resonance (MR)
`imaging. Besides the economic aspects associated
`with decreasing acquisition times, development of
`fast imaging methods is also motivated by the re-
`duction in artifact due to gross patient motion,
`three-dimensional acquisition and visualization of
`imaging data, and the possibility of observing dy-
`namic changes with MR imaging. Despite the rela-
`tively long data acquisition times, spin-echo se-
`quences remain the “gold standard” among MR
`pulse sequences for routine screening procedures.
`Any technique designed to replace the spin-echo
`method must realistically exploit the same basic
`contrast mechanisms and should produce images
`with the same overall quality.
`Our goal was to determine the extent to which
`the FAISE (fast-acquisition interleaved spin-echo)
`method can yield head images that compare with
`conventional dual-echo spin-echo images in terms
`of contrast-to-noise ratios (C/Ns), noise levels, and
`overall image quality.
`
`0 METHODS
`The FAISE method was implemented on a 1.5-T
`Signa system (GE Medical Systems, Milwaukee)
`equipped with actively shielded gradient coils, digi-
`tal transmitter and receiver, and a 4.x hardware/
`software configuration. A radio-frequency (RF)
`transmit-receive head coil operated in quadrature
`was used for all studies. The FAISE method uses
`the hybrid RARE (rapid-acquisition relaxation-en-
`hanced) sequence discussed by Hennig et a1 (1,2).
`RF-refocused echoes are individually phase-en-
`coded prior to readout and phase “unwound” after
`readout, as depicted in Figure 1. The FAISE se-
`quence consisted of a 16-echo CPMG train with an
`echo spacing of 15 msec and an acquisition window
`of 8.2 msec. In some cases, the acquisition window
`was reduced to 4.1 msec, permitting 11.5-msec
`echo spacings. A total of 16 “shots.” or spin excita-
`tions, of the CPMG sequence covers k space with
`data suitable for a 256 X 256 image matrix. With a
`2-sec TR between successive spin excitations, the
`
`319
`
`General Electric Co. 1008 - Page 1
`
`

`
`Gz ~L..JI
`1 n m
`
`I
`I
`
`G X
`
`n
`
`,$/
`
`I
`I
`
`m ,//’ m
`
`I
`I
`
`I
`I
`
`I
`
`I
`
`I
`
`I
`
`I
`I
`I
`I
`I
`I
`
`I
`I
`
`krnln
`
`I
`
`7
`
`Figure 1. The hybrid RARE pulse sequence for generat-
`ing individually phase-encoded echoes within a Carr-Pur-
`cell-Meiboom-Gill (CPMG) echo train.
`
`kmln
`
`I l l I V I I
`
`I
`
`krnax
`
`kmax
`
`total imaging time is 34 sec, a time period that in-
`cludes a baseline shot. A n echo spacing of 15 msec
`places the last echo of each train at 240 msec, per-
`mitting eight sections to be excited within the 2-sec
`TR period. Alternatively, the number of echoes col-
`lected after the excitation pulse may be reduced to
`eight or four. With a reduced echo train, an in-
`creased number of sections may be excited at the
`expense of imaging time [increased number of
`shots).
`Each separate echo train covers a discontinuous
`region of k space. The k-space traverses are succes-
`sively interleaved in the course of a full FAISE ac-
`quisition. The resulting T2 contrast within the 16-
`echo, 16-shot FAISE acquisition is controlled by al-
`tering the order in which phase encoding is applied
`to the RF spin echoes. Because overall image con-
`trast is dominated by those signals acquired with
`the smallest phase-encoding gradient amplitudes
`(3,4), the T2 contrast may be controlled by acquir-
`ing these data lines at a user-selected pseudo echo
`time [pTE). Figure 2 illustrates the interleaving pro-
`cess that permits contrast manipulation. For sim-
`plicity, a four-shot, four-echo FAISE sequence with
`only 16 total phase-encoding steps is considered.
`The vertical axes in the three plots represent sig-
`nal-intensity decay with echo number (Arabic nu-
`merals) within each of the four shots [Roman nu-
`merals). The horizontal axes in Figure 2 are the
`phase-encoding axes in k space. The top plot illus-
`trates how the least T2-weighted images, associat-
`ed with a short pTE value, are obtained. Note that
`within each shot, the early echoes are phase-en-
`coded with the smallest gradient amplitudes. An in-
`termediate T2 weighting is obtained with the
`phase-encoding interleaving scheme illustrated in
`the middle plot in Figure 2. In this pattern, both
`negative and positive phase-encoding gradients oc-
`cur within two of the shots (broken arrows, shots 111
`and IV). The bottom plot illustrates the correlation
`
`320 0 JMRl May/June 1991
`
`I l l
`
`I V I I
`
`I
`
`kmax
`kmin
`Figure 2. The FAISE method of phase reordering for ob-
`taining short pTE (proton-density) images (top), interme-
`diate T2-weighted images (middle), and heavily T2-
`weighted images (bottom]. The FAISE technique is illus-
`trated with a four-shot, four-echo sequence in which only
`16 phase-encoding steps are acquired. Each shot (echo
`train) is labeled with a Roman numeral and each echo
`with an Arabic numeral. The vertical axis in each plot
`represents the relative T2 weighting. With an echo spac-
`ing of 15 msec, the three plots can be associated with
`pTEs of 15.30, and 60 msec (top to bottom], because the
`first, second, and fourth echoes, respectively, have been
`acquired near the origin of k space.
`
`between the phase-encoding axis and echo number
`that yields the most heavily T2-weighted images,
`the late echoes being encoded with the smallest
`phase-encoding gradient amplitudes.
`For a complete description of the k-space tra-
`verse as covered with multiple-echo trains, a gener-
`al algorithm for hybrid RARE sequences has been
`developed. This algorithm calculates for each echo
`numberj. collected during echo train, or shot, i, a
`specific phase-encoding gradient amplitude Gy[t, j)
`for each line in k space. Each Gy[i, j ) amplitude,
`which ranges from a maximum negative phase-en-
`coding gradient amplitude Gymin to a maximum pos-
`itive amplitude Gymax. is related to an index A[i.J),
`ranging from 1 to Ly, where ty is the total number
`of phase-encoding steps. Each A(1. j) also corre-
`sponds to a k-space line index (3) and is used for
`proper signal reordering of the raw data matrix pri-
`or to image reconstruction. The formalism reads as
`
`General Electric Co. 1008 - Page 2
`
`

`
`follows:
`Gy(Lj) = Gym,, + [A(iJ) - 1I[Gymax - GyminI/[Ly - 11,
`(1)
`where for 1 I i 5 NJ2,
`A[L,j) = [(L,/2 -jNs/2 + 1 - 1 + IPS)mod[L,)] + 1,
`( 2 4
`and for Ns/2 < i I Ns,
`A[i, j ) = [[L,/2 + (j - 2)Ns/2 + t - 1 + IPS)mod(L,)] + 1.
`(2b)
`These relations completely define the correlation
`existing between each RF echo, indexed by the ( t , j )
`variables, and its corresponding k-space phase-en-
`coding gradient amplitude G,(i,j). They hold gener-
`ally for any number of echoes Ne and for an arbi-
`trary number of phase-encoding steps L,. where
`the number of shots Ns obeys the relation Ns = Ly/
`Ne. The modulus function [mod) operates on the
`terms in parentheses. The operator-chosen vari-
`able IPS (initial phases skipped) may range from 0
`to L,/2 and is the number of phase-encoding steps
`[from the 0 phase-encoding step) that are skipped
`in the initial shot (1 = 1). IPS is varied to manipulate
`T2 contrast. We now define the convenient pTE pa-
`rameter and relate it to the IPS value through the
`following set of relations:
`pTE = 27, for IPS = 0,
`pTE = 275 for (j - 1)Ns/2 < IPS 5jNs/2,
`where 27 is the echo spacing of the CPMG se-
`quence. Note, however, that for a 16-echo, 16-shot
`sequence, eight IPS values correspond to a single
`pTE. Therefore, IPS values are reported along with
`the more "user friendly" pTE value for the sake of
`precision.
`FAISE head images in healthy volunteers were
`obtained with various pTE values and a TR of
`2,000 msec between shots. Conventional spin-echo
`images with the same TR and TEs of 30 and 90
`msec (TR msec/TE msec = 2,000/30,90) were also
`collected. The total data acquisition time for the
`single-excitation, 256 X 256 matrix, spin-echo se-
`quence was 8 minutes 56 seconds. Signal intensi-
`ties were measured in selected regions and C/Ns
`were calculated for cerebrospinal fluid and white
`matter and for gray matter and white matter with
`use of the relation
`C/N = [Si- S,)/Np,
`(4)
`where S1 and S2 are the mean values of signal in-
`tensities from selected regions of interest (ROIs) in
`the two tissue types. Np is the mean value of the
`noise signal intensity measured in ROIs along the
`phase-encoding direction and outside the imaged
`object. Mean values and standard deviations of
`noise signal intensities (NR) in ROIs outside the im-
`aged object and along the readout direction were
`also measured and compared with Np. These mea-
`surements were performed in selected regions
`
`(34
`(3b)
`
`with more than 300 pixels.
`Two approaches were used to obtain T1 contrast
`with FAISE: inversion recovery and partial satura-
`tion. The inversion-recovery experiments were per-
`formed by application of a section-selective 180"
`pulse at an inversion time [TI) of 600 msec before
`each excitation pulse. Partial-saturation T 1 -
`weighted images were obtained by decreasing the
`TR between FAISE shots to 500 msec. The echo
`spacing of the CPMG trains was reduced to 1 1.5
`msec by reducing the acquisition window to 4.1
`msec, with a subsequent increase in bandwidth
`and reduction in signal-to-noise ratio. Both experi-
`ments were performed in a healthy volunteer.
`Experiments designed to measure RF-induced
`temperature changes in a phantom consisting of
`100 mL of 0.9 wt% saline in an insulated Styro-
`foam cup were performed. A thermocouple consist-
`ing of constantan-copper wire was inserted into the
`saline phantom. The thermocouple was connected
`to a two-wire thermocouple transmitter (model
`TX52; Omega Engineering, Stamford, Conn) that
`generates a standard 4-20-mA output current pro-
`portional to the millivolt thermocouple signal. Mea-
`surements were made both before and after appli-
`cation of a five-section 16-echo, 16-shot FAISE se-
`quence (TR = 2 sec) but with nine signal averages
`per phase encoding (total data acquisition time,
`290 sec). For comparison, thermocouple voltage
`changes were also recorded before and after a 290-
`sec application of a 20-section, four-echo sequence
`with the same number of 180" RF pulses per TR as
`the FAISE sequence. Voltage changes on the load
`resistor in the current loop were converted to tem-
`perature changes (6T in degrees centigrade) with a
`calibration plot previously obtained with a tem-
`perature-regulated water bath. The amplitude fluc-
`tuations of measured temperatures during typical
`baseline temperature measurements were on the
`order of 0.025OC. Four separate before/after tem-
`perature measurements were made for each se-
`quence, and the mean values and standard devi-
`ations of the temperature increases were calculat-
`ed.
`Specific absorption rates were estimated with the
`manufacturer-supplied software (GE Medical Sys-
`tems, Milwaukee) for typical FAISE acquisitions
`and patient weights ranging from 50 to 100 kg.
`
`0 RESULTS
`Sagittal FAISE head images (256 X 256 matrix)
`obtained with pTEs of 45.75, 105, and 165 msec
`(IPS values of 20.36, 52, and 84, respectively) are
`shown in Figure 3. The echo spacing was 15 msec
`and a 2,000-msec TR was used between shots. The
`images, obtained in a healthy volunteer, reveal how
`image contrast is influenced by the pTE value.
`Axial FAISE head images in a healthy volunteer,
`obtained through the ventricles and acquired with
`pTEs of 30 and 90 msec (IPS values of 12 and 44,
`respectively: a 256 X 256 matrix, 5-mm thick sec-
`tions, an echo spacing of 15 msec, and a TR of
`2,000 msec], are shown in Figure 4. The corre-
`sponding 256 X 256 spin-echo images, obtained
`with the same TR and with TEs of 30 and 90 msec.
`
`Volume 1 Number 3
`
`JMRl
`
`321
`
`General Electric Co. 1008 - Page 3
`
`

`
`a.
`
`b.
`
`d.
`C.
`Figure 3. Manipulation of T2 contrast by varying pTE by means of the IPS parameter in a FAISE sequence performed in
`a healthy volunteer (TR = 2,000 msec, 15-msec echo spacing, 256 X 256 matrix size, 5-mm section thickness). The pTEs
`were 45 (a), 75 (b), 105 (c), and 165 msec (a).
`
`are shown in Figure 5. The Table lists C/N values
`obtained from these spin-echo images and the
`FAISE images. Also tabulated are C/N results ob-
`tained from FAISE acquisitions with other pTE/IPS
`values (images not shown]. Figure 6 shows axial
`head images obtained through the circle of Willis,
`one with FAISE with a pTE of 90 msec (1 5-msec
`echo spacing, TR of 2,000 msec, and IPS of 44) and
`the other a corresponding second-echo image of a
`dual-echo spin-echo sequence (2,000/30, 90). Both
`images are clearly T2 weighted, but the FAISE im-
`age demonstrates increased signal intensity within
`the substantia nigra, smaller-diameter vascular lu-
`
`mina, and a persistence of subcutaneous fat signal
`not apparent on the spin-echo image.
`The T1 -weighted FAISE images obtained with in-
`version-recovery and partial-saturation techniques
`are shown in Figure 7. The inversion-recovery
`FAISE image was obtained in a healthy volunteer
`with a TI of 600 msec, a TR of 2,000 msec, and a
`pTE of 69 msec (IPS of 44, 34-sec acquisition). The
`partial-saturation image was acquired in 9 sec,
`with a TR of 500 msec and pTE of 34.5 msec (IPS of
`20). For both T1 acquisitions, the 4-msec acquisi-
`tion window was used to reduce the echo spacing
`within the CPMG echo train.
`
`322 JMRl May/June 1991
`
`General Electric Co. 1008 - Page 4
`
`

`
`NP
`
`NR
`
`NPINR
`
`16.9 f 8.3
`16.0 f 7.9
`
`13.9 f 6.6
`13.9 f 6.5
`
`1.22
`1.15
`
`4.73
`6.00
`
`-5.27
`16.25
`
`Contrast-to-Noise Ratios of Dual-Echo Sdn-EChO and FAISE Ima#les
`Sequence
`(GM-WM)/Np
`(CSF-WM)/Np
`Spin echo
`TE msec
`30
`90
`FAISE
`pTE msec (IPS)
`19.8f 10.0
`1.38
`14.3 f 6.6
`-3.46
`2.30
`15 (0)
`19.5 f 9.8
`1.79
`3.79
`1.35
`14.4 f 6.7
`30 (121
`1.20
`14.4 f 6.9
`17.3 f 8.5
`3.38
`4.90
`60 (281
`17.8 f 8.5
`9.15
`5.13
`14.0 f 6.5
`1.27
`75 (36)
`16.99
`7.23
`16.1 f 7.9
`1.15
`14.0 f 6.5
`90 (44)
`15.4 f 7.4
`26.48
`6.48
`150 (76)
`1.09
`14.1 f6.8
`14.6 f 7.1
`29.94
`5.42
`210(116)
`1.04
`14.1 f 6.7
`for gray matter-white matter (GM-WM) and cerebrospinal fluid-white matter (CSF-WM) from images in a healthy
`Note.-C/Ns
`volunteer obtained with a double-echo spin-echo sequence (2,000/30. 90) and with several pTE values of a FAISE sequence ( 15-
`msec echo spacing). Included are noise measurements (mean f standard deviation) made along the phase-encoding (NP] and read
`gradient (NR) axes and their ratio.
`
`a.
`
`b.
`
`Figure 4. FAlSE images
`in a healthy volunteer ac-
`quired with pTE values of
`30 (a) and 90 msec (b), an
`echo spacing of 1 5 msec, a
`TR of 2,000 msec, a 5-mm
`section thickness, and a n
`acquisition time of 34 sec.
`
`Figure 5. Spin-echo im-
`ages (2,000/30.90) corre-
`sponding to those in
`Figure 4.
`
`a.
`
`b.
`
`Volume1 Number3
`
`JMRl
`
`323
`
`General Electric Co. 1008 - Page 5
`
`

`
`Figure 6. (a) FAISE im-
`age with pTE of 90 msec
`and (b) spin-echo image
`with TE of 90 msec ob-
`tained through the circle of
`Willis with the same TR
`(2,000 msec) and 5-mm
`section thickness. Note the
`narrower appearance of
`the arteries and the persis-
`tence of subcutaneous fat
`signal on the FAlSE image.
`There is also less signal in-
`tensity in the substantia
`nigra on the spin-echo im-
`age.
`
`Figure 7. T1-weighted
`images (5-mm section
`thickness) obtained with
`(a) the inversion FAISE se-
`quence (TI = 600 msec, TR
`= 2.000 msec, pTE = 69
`msec, 34-sec acquisition
`time) and (b) a FAISE se-
`quence with a reduced TR
`interval (TR = 500 msec,
`pTE = 34.5 msec, 9-sec ac-
`quisition time).
`
`a.
`
`b.
`
`a.
`
`b.
`
`Mean values and standard deviations of the tem-
`perature increases in the 100-mL-saline phantom
`were 0.20"C f 0.06"C for the FAISE sequence and
`0.15"C f 0.08"C for the four-echo spin-echo se-
`quence (results obtained from four separate trials).
`
`0 DISCUSSION
`In his original work, Hennig demonstrated the
`clinical utility of RARE sequences for imaging
`structures with long T2 values and has promoted
`the technique for its myelographic and urographic
`effects (1,2). With the phase-reordering algorithm
`
`we have introduced, hybrid RARE sequences are
`used to traverse k space in a manner that allows
`straightforward T2 contrast manipulation (Eq [2a].
`[2b]). The covering of k space with multiple shots
`increases the overall data acquisition time when
`compared with single-shot RARE methods (1.2.41,
`which can have overall acquisition times on the or-
`der of 1 sec. However, the use of a reduced number
`of echoes, compared with 128 or 256, permits
`multisection acquisitions within standard TRs
`while maintaining small time-per-image values and
`allows short TR acquisition to be performed. If the
`
`324
`
`JMRl May/June1991
`
`General Electric Co. 1008 - Page 6
`
`

`
`length of the echo train is made short with respect
`to the cardiac cycle, the potential may exist for syn-
`chronizing acquisitions with cardiac motion. The
`image quality we have observed with FAISE se-
`quences, as currently implemented, has led us to
`compare FAISE images with conventional spin-
`echo images of brain.
`Figure 3 illustrates how changing the pTE value
`systematically changes the T2 weighting of the im-
`ages. The pTEs in Figure 3 range from 45 to 165
`msec. Note the brightening of cerebrospinal fluid
`spaces and the gradual decrease in signal intensity
`in the corpus callosum as the pTE increases. In Fig-
`ures 4b and 5b. the T2-weighted axial views of a
`normal head obtained with FAISE and spin-echo
`sequences are strikingly similar. White matter/gray
`matter differentiation, deep gray matter structures,
`and overall image details are well depicted on both
`sets of images, with similar contrast distributions.
`The qualitative similarity in T2 weighting is further
`supported by the C/N measurements shown in the
`Table, where FAISE images with pTEs of 75 and 90
`msec have C/N values closely corresponding to
`those obtained from the spin-echo image with a TE
`of 90 msec. However, when FAISE and spin-echo
`methods for obtaining proton-density images (Figs
`4a, 5a) are compared, differences can be noted. The
`FAISE axial view of the head, acquired with a pTE
`of 30 msec (Fig 4a), suffers from a blurring effect
`common in RARE sequences (1.4). This is well ap-
`preciated at the border of gray matter and white
`matter structures in Figure 4a. In addition, the pro-
`ton-density FAISE image exhibits edge-enhance-
`ment artifacts at the interface between brain and
`cerebrospinal fluid structures. Some gyri with ori-
`entations perpendicular to the phase-encoding di-
`rection (horizontal) show increased signal intensi-
`ty, although most of the parietal gyri oriented in
`this direction appear as dark lines (Fig 4a). We be-
`lieve that both of these effects are related to the T2
`decay process occurring during data collection. The
`Table reveals that for higher pTE values, both
`types of FAISE images have phase-encoding noise
`values similar to those obtained for the spin-echo
`sequences and the Np values approach the read
`gradient values NR.
`Several other features differentiate FAISE and
`spin-echo images, and these must be addressed.
`For instance, Figure 6 demonstrates higher signal
`intensity in the substantia nigra on the FAISE im-
`age as opposed to the corresponding spin-echo im-
`age. In addition, there is a persistence of subcuta-
`neous fat signal in heavily T2-weighted FAISE im-
`ages compared with the long TE spin-echo images.
`These features have also been observed when indi-
`vidual images are constructed from each echo of a
`CPMG sequence with the same echo spacing and
`TR as in the FAISE sequence (images not shown).
`We believe, therefore, that the contrast mecha-
`nisms influencing CPMG and FAISE images are
`closely related and that when this contrast differs
`from that of a conventional spin-echo image ac-
`quired at the same TE or pTE, the difference is re-
`lated to a T2 dependence on echo spacing.
`Several well-known mechanisms lead to a
`
`lengthening of T2 as the echo spacing is decreased
`(5-1 1). These include homonuclear coupling inter-
`actions (5.6). chemical exchange between sites
`with distinct chemical shifts (7.8). and proton diffu-
`sion in the presence of microscopic or macroscopic
`field gradients (9-1 1). The first of these effects ap-
`pears quite pronounced in the subcutaneous fat, a
`tissue in which spin-spin splittings and chemical
`shift effects are active in determining relaxation.
`The second effect is noticeable in the substantia ni-
`gra. in which paramagnetic iron-containing moi-
`eties (1 2) generate microscopic field gradients re-
`sponsible for a diffusion-related dephasing phe-
`nomenon. Both of these mechanisms, which
`depend on echo spacing, apparently lead to longer
`T2s at the short echo spacings used in either FAISE
`or CPMG sequences. Therefore, the short echo
`spacings used in FAISE may lead to decreased sen-
`sitivity to susceptibility effects than routine dual-
`echo sequences that use longer echo spacings. This
`feature may prove advantageous when imaging re-
`gions such as the temporal lobes, pontocerebellar
`angle, and spinal structures in contact with bone.
`It may prove disadvantageous, however, when he-
`morraghic lesions are imaged. However, further
`studies are required to assess all of the detailed
`contrast mechanisms influencing the FAISE im-
`ages.
`The vascular structures in the FAISE image in
`Figure 6 appear narrower than the corresponding
`structures in the spin-echo image. This may in part
`be due to the postulated decreased susceptibility
`sensitivity of FAISE sequences to paramagnetic
`blood products. However, flow-related phenomenon
`influencing routine spin-echo sequences, most
`probably influence FAISE images differently be-
`cause of the different acquisition method. A thor-
`ough analysis of the effects of flow on signal inten-
`sity in FAISE images is certainly required if fea-
`tures like the ghosting artifact seen along the
`phase-encoding direction adjacent to the lower por-
`tion of the superior longitudinal sinus in Figure 6a
`are to be understood.
`Figure 7 demonstrates the ability to obtain T1
`contrast with FAISE. Ideally, short pTEs are re-
`quired if T2 weighting is to be minimized. However,
`the pTEs of 69 and 34 msec (Figure 7a and 7b, re-
`spectively) were used to reduce blurring effects as-
`sociated with the shortest pTEs. Both images repro-
`duce the typical T 1 -weighted contrast associated
`with inversion-recovery spin-echo and partial-sat-
`uration spin-echo sequences. Partial-saturation
`FAISE images, which are acquired in less than 10
`sec, could be used to perform contrast enhance-
`ment studies.
`If one assumes that image quality and tissue con-
`trast are not compromised with use of FAISE in
`place of spin echo, it is important to ask what effi-
`ciency advantage FAISE might have over dual-echo
`spin-echo sequences. Considering only 256 X 256
`image matrices, a 16-shot, 16-echo FAISE se-
`quence with 15-msec echo spacing yields eight sec-
`tions with a single (operator chosen) T2 weighting
`in 34 sec. This results in a theoretical time-per-im-
`age value of 4.25 sec. With our standard dual-echo
`
`Volume 1 Number3
`
`JMRl
`
`325
`
`General Electric Co. 1008 - Page 7
`
`

`
`spin-echo sequence (2.000/30.80), a total of 18
`sections are acquired (with two separate weightings
`per section) in 8 minutes 56 seconds. This leads to
`a time-per-image value of 14.8 sec. The decrease in
`the time-per-image of FAISE relative to that of the
`dual-echo sequence is. therefore, greater than a
`factor of three. To cover the same volume and pro-
`vide the same T2 weighting as the dual-echo se-
`quence, the FAISE sequence described above would
`require six serial acquisitions. To increase the
`number of sections acquired with FAISE, longer
`TRs or echo trains consisting of fewer echoes may
`be used at the expense of imaging time. Only two
`serial acquisitions would then be required to obtain
`two different T2 weightings over the required vol-
`umes.
`Finally, there is the issue of increased RF power
`deposition associated with FAISE. The general
`guidelines underlying RF power deposition levels
`are based on the restriction that tissue temperature
`increases should not exceed 1 "C (1 3). To test
`whether the high density of 180" refocusing pulses
`used in our FAISE method could potentially lead to
`such temperature increases, the saline phantom
`experiments were performed. The temperature in-
`creases observed in four separate trials with both
`the four-echo multiplanar sequence and the 16-
`echo FAlSE sequence were between 0.1 and 0.2"C.
`The temperature increases are small enough to
`suggest that irradiated regions of perfused tissue,
`with its inherent cooling capacity, should not be
`damaged. However, RF power deposition-induced
`temperature increases in saline phantoms are only
`crude indicators of potential temperature increases
`in actual applications to humans, where the load-
`ing of the RF coil and geometrically influenced eddy
`currents in the body are different. Estimated spe-
`cific absorption rate values for the FAISE sequence
`ranged from 0.1 16 to 0.130 W/kg for the eight-sec-
`tion FAISE acquisition (1 6-echo, 16-shot, TR =
`2,000 msec, 27 = 15 msec) performed with the head
`coil configuration. These estimated values are well
`below the Food and Drug Administration (FDA) lim-
`it of 3.2 W/kg (14). In other applications, however,
`it may be found that specific absorption rate values
`for maximum-volume FAISE sequences exceed
`FDA guidelines. In this case, it is possible to gener-
`ate echo trains in which the flip angles of the refo-
`cusing pulses are less than 180" (1 5), thus lowering
`RF power deposition levels to within FDA-approved
`limits at the expense of signal-to-noise ratio. Alter-
`natively, the number of sections can be reduced to
`remain within the guidelines, permitting fast, safe
`acquisitions over limited volumes.
`In conclusion, the FAISE sequence demonstrated
`here decreases theoretical time-per-image values
`by more than a factor of three compared with dual-
`echo spin-echo sequences while maintaining the
`overall contrast and quality of the latter for T2-
`weighted images. The proton-density images ob-
`tained with FAISE are not entirely equivalent to
`proton-density spin-echo images. However, it is
`possible that further reductions in echo spacing,
`shorter echo trains, and/or the application of k-
`
`326
`
`JMRl May/June1991
`
`space corrections in the postprocessing stage (1 6)
`could substantially improve proton-density FAlSE
`images. If the additional RF power deposition can
`be shown to involve minimal risk, the acquisition of
`larger matrix sizes (eg, 5 12 X 5 12) in reasonable
`time periods becomes feasible within a multisection
`two-dimensional format. Alternatively, the exten-
`sion of the two-dimensional Fourier transform
`FAISE techniques to three-dimensional Fourier
`transform methods (4) could permit the acquisition
`of a three-dimensional spin-echo-like data set in
`imaging times of less than 5 minutes. 0
`
`Acknowledgments: Dr Melki is grateful for financial sup-
`port from the Institut National de la Sante et de la Recherche
`Medicale (INSERM) for pursuing this work while on leave from
`the Unite 31 6 de I'INSERM (Systeme nerveux du foetus a I'en-
`fant) in the Centre Hospitalier Universitaire Bretonneau de
`Tours, France. He thanks Prof Leandre Pourcelot for encour-
`aging this work. The authors also acknowledge Peter Jakab,
`MSE. David Feinberg, MD, PhD, Nobuya Higuchi. MD, and
`Sam Wong, PhD, for their interest and helpful discussions
`during the course of this work.
`
`2.
`
`3.
`
`4.
`
`5.
`
`6.
`
`7.
`
`References
`1.
`Hennig J. Naureth A. Friedburg H. RARE imaging: a
`fast imaging method for clinical MR. Magn Reson Med
`1986; 3:823-833.
`Hennig J, Friedbug H. Clinical applications and meth-
`odological developments of the RARE technique. Magn
`Reson Imaging 1988: 6:391-395.
`Twieg DB. The k-space trajectory formulation of the
`NMR imaging process with applications in analysis and
`imaging methods. Med Phys 1983: 10:610-621,
`Mulkern RV. Wong STS. Winalski C. Jolesz FA. Con-
`trast manipulation and artifact assessment of 2D and
`3D RARE sequences. Magn Reson Imaging 1990; 8:557-
`566.
`Santyr GE. Henkelman RM, Bronskill MJ. Variation in
`measured transverse relaxation in tissue resulting from
`spin locking with the CPMG sequence. J Magn Reson
`1988; 79:28-44.
`Allerhand A. Analysis of Carr-Purcell spin-echo NMR
`experiments on multiple-spin systems. 1. The effect of
`homonuclear coupling. J Chem Phys 1966: 44: 1-9.
`Carver JP. Richards RE. A general two-site solution for
`the chemical exchange produced dependence of T2 upon
`the Carr-Purcell pulse separation. J Magn Reson 1972:
`6:89-105.
`8.
`Gutowsky HS, Vold RL, Wells EJ. Theory of chemical
`exchange effects in magnetic resonance. J Chem Phys
`1965: 43:4107-4 125.
`Bendel P. Spin-echo attenuation by diffusion in non-
`uniform field gradients. J Magn Reson 1990; 86:509-
`515.
`Rozenman Y, Zou X, Kantor HL. Signal loss induced by
`superparamagnetic iron oxide particles in NMR spin-
`echo images: the role of diffusion. Magn Reson Med
`1990; 14:31-39.
`Mulkern RV, Spencer RGS. Diffusion imaging with
`paired CPMG sequences. Magn Reson Imaging 1988:
`6:623-631.
`Drayer BP. Burger P, Darwin R, et al. Magnetic reso-
`nance imaging of brain iron. AJNR 1986: 7:373-380.
`Saunders RD, Smith H. Safety aspects of NMR clinical
`imaging. Br Med J 1984; 40: 148- 154.
`Federal Register 1988; 53:48981.
`Hennig J. Multiecho imaging sequences with low refo-
`cusing flip angles. J Magn Reson 1988: 78:397-407.
`Oshio K, Singh M. A computer simulation of T2 decay
`effects in echo planar imaging. Magn Reson Med 1989;
`1 1 :389-397.
`
`9.
`
`10.
`
`11.
`
`12.
`
`13.
`
`14.
`15.
`
`16.
`
`General Electric Co. 1008 - Page 8