WO2012054768A1 - Multiplexed shifted echo planar imaging - Google Patents

Abstract

MR imaging in which modulated gradient pulses applied in a special way in the Gs (slice select) axis and the Gp (phase) axis ensure that there is no, or little, dephasing effect from the accumulated Gs gradient pulses. Diphase-reducing gradient pulses are used, such that the phase of Gs is at or close to zero when the phase of Gp is at or close to zero (at ko in k-space). This is particularly important when the MB factor becomes larger than 3 slices, and the ko signal becomes higher amplitude. This new technique which shifts the center ol k-space and a time point when the periodic Gs phase modulation passes through or near zero phase to together have minimal net dephasing on the center of k-space, is referred to as multiplexed shifted EPI.

Description

MULTIPLEXED SHIFTED ECHO PLANAR IMAGING

Field

[001 ] This patent specification pertains to magnetic resonance imaging (MRI). Background and summary of the disclosure

[002] In magnetic resonance imaging, both diffusion imaging and functional imaging can rely on image formation via the echo planar imaging (EPI) technique, or a closely related variant such as spiral sampling. Efforts have been made to improve EPI and, particularly, to increase data acquisition efficiency in EPI. One approach has been to time-multiplex MRI signals from different slices of the patient's body and another, independent approach has been to frequency-multiplex MRI signals from different slices.

[003] Yet another approach is proposed in an abstract: Setsompop K, Gagoski BA, Polimeni J, Witzel T, Wedeen VJ, and Wald LL, Blipped CAIPARHINA for simultaneous multi-slice EPI with reduced g-factor penalty, Proc. Intl. Soc. Mag. Reson. Med 18 (2010), page 551 . The abstract proposes variations of time- multiplex and frequency-multiplex MR data acquisition aimed at increasing inter-slice image shift in the PE direction by introducing gradient blips in the slice-select (Gz) direction simultaneously with the conventional Gy blips. According to the abstract, a constant phase accumulation across ky (in k-space) is eliminated with back and forth jumps. The abstract shows Gz blips that are the same in strenght but alternate in sign for FOV/2 shift, and also a blip sequence for FOV/3 shift where two sequential blips that are the same strength and sign are followed by an opposite sign blip that is twice the strength of the preceding blip. However, the abstract only shows a low image multiplication factor of 2 and 3, and does not teach about a need for of effects of adjusting the relative phase pattern of Gz blips with respect to the concurrent Gp normal phase-encoding pulses in connection with image separation and image quality. [004] A novel technique described in detail below adds different considerations and is particularly advantageous for acquiring greater number of images in less time while suppressing undesirable consequences of other approaches. This new technique takes into account effects of the acquired MR signals on k-space, and particularly on the center ko of k-space, that are not apparent from prior work and perhaps are counterintuitive to known earlier proposals.

[005] The novel technique incorporates modulated gradient pulses in the Gs (slice select) axis and the Gp (phase) axis in a special way ensuring that the MR signal for ko in k-space would have maximum or near maximum amplitude when there is a node in the accumulated Gs at which there is no, or little, dephasing effect from the accumulated Gs gradient pulses. A goal of the new technique is to avoid dephasing of ko by the Gs accumulated gradient pulses by using gradient pulses such that the phase of Gs is at or close to zero when the phase of Gp is at or close to zero (at ko in k-space). This is particularly important when the MB factor becomes larger than 3 slices, and the ko signal becomes higher amplitude. If there were an effective Gs dephasing of ko at MB > 3 then there would be increasing intervoxel dephasing of the simultaneously acquired slices and large areas of attenuated signal would occur. This new technique which shifts the center of k-space and a time point when the periodic Gs phase modulation passes through or near zero phase to together have minimal net dephasing on the center of k-space, is referred to as multiplexed shifted EPI. The Setsompop et al. abstract cited above does not recognize or teach this, and does not show a concurrent zero phase of Gp and Gs. It is not apparent from the abstract that deals with lower MB factors of 3 or 2, or from other known prior material, how the controlled aliasing use of blipped Gs pulses for higher MB factors may be adversely affected when using the technique proposed in the abstract, according to which the signal peak would be lower and the dephasing interference between the simultaneously acquired signals at different slice positions on the Gs axis would give a much less severe signal loss. It may be counterintuitive to suggest, as taught in the new technique described further below, that signal loss in the images will become increasingly worse as the MB factor increases from 4 to 6, 8, 10, 12.... and higher, and how severe the interference would become absent reliance on the new technique disclosed in this patent specification. The new technique is discussed in greater detail in connection with Figs. 19-21 below, after a discussion of time-multiplexing and frequency-multiplexing MR imaging.

6] Time-multiplexing of MRI signals from different slices typically is called Simultaneous Echo Refocusing (SER) or Simultaneous Image Refocusing (SIR) EPI. Examples are described in (i) U.S. Patent No. 6,614,225, (ii) Feinberg, DS, Reese TG, Wedeen VJ, Simultaneous Echo Refocusing in EPI, Magn Res Med 48(1 ):1 -5 (2002), and (iii) Reese TG, Benner T, Wang R, Feinberg DA, and Wedeen VJ, Halving Imaging Time of Whole Brain Diffusion Spectrum Imaging and Diffusion Tractography Using Simultaneous Image Refocusing in EPI, J Magn Res Imaging 29:517-522 (2009). The patent and the two articles cited in this paragraph are hereby incorporated by reference in this patent specification, as is every other patent, abstact and article cited elsewhere in this patent specification. Fig. 5 in this specification illustrates an example of a classical EPI pulse sequence, and Fig. 6 illustrates an example of a 2-slice SIR pulse sequence. In Figs. 5 and 6, as in all other Figs, illustrating pulse sequences, the pulses and time spacings are not to scale. In the classical EPI example of Fig. 5, the 90° RF (radio frequency) pulse on the RF axis excites a slice in the body of a subject (e.g., a patient) in an imaging volume of an MRI scanner, according to the gradient pulses shown on the Gs axis. A train of MRI signals 1 , 2, 3, ...is emitted from the subject. This train of MRI signals is read out with the help of a sequence of rephrasing gradient pulses that alternate in polarity as illustrated on the Gr axis, and phase encoding gradient pulses as illustrated on the Gp axis. The MRI signals are detected with one or more RF receiver coils, and the coil outputs are computer-processed to produce MRI image data for a k-space matrix and thereafter to produce an MRI image of the slice as known in MRI technology. In the 2-slice SIR example of Fig. 6, two RF pulses RFa and RFb are applied to the subject in time succession and, with the help of the gradient pulses illustrated in the preparatory time period Tpre, cause the patient to emit a time succession of N sets of two MRI signals (a) and (b) each. The first set comprises, in time sequence, MRI signals b1 and a1 for respective slices Sb and Sa of the subject, obtained in a single read interval N1 ; the second set comprises MRI signals a2 and b2 for the same two slices but in reverse order, obtained in a second read interval N2; the third set comprises signals b3 and a3, obtained in a read interval N3, etc. A rephasing gradient shown on the gradient axis Gr alternates in polarity from one read interval N to the next, to thereby produce MRI signals for a total on N read intervals from the two RF pulses RFa and RFb, where N>2. The two RF excitation pulses typically are 90° pulses that are slightly offset in frequency from each other. As a result, MRI echo signals are acquired from two slices in the time that a single echo MRI signal would be acquired absent the use of SIR. Thus, time- multiplexing of images in the readout periods N of SIR EPI increases data acquisition efficiency to thereby reduce average scan time, especially in diffusion imaging.

7] SIR data acquisition is impacted by the ratio of the preparatory time Tpre to the total MRI echo signal time. In one example, in an MRI data acquisition known as HARDI acquisition, the preparatory period Tpre can be approximately 80 ms and the echo train approximately 20 ms. The sharing of Tpre with two or more slices creates a large gain in sequence efficiency, defined here as net time of analog-to-digital (ADC) signal encoding per total sequence time. Another gain in efficiency in SIR is by the sharing of the many gradient switchings Tsw. Therefore scanners with slower slew rate or gradient ramp-time in their gradient systems (longer Tsw) also become more efficient with SIR, and similarly for lower resolution imaging (shorter repetition time TR relative to Tsw) efficiency and time savings increase. Despite the overall advantages of scan time reduction of SIR, the sampling time and echo spacing for any one SIR slice are longer than for a classical EPI. The lengthening of the SIR echo train in the presence of local T2^* (time constant for loss of phase coherence among spins oriented at an angle to the static magnetic field due to a combination of magnetic field inhomogeneities and the spin-spin relaxation) and Bo (static magnetic field) inhomogeneity increase image distortions to varying degrees but without losses in SNR (signal-to-noise ratio) provided the TE (echo time, or the time between the application of the 90° pulse and the peak of the echo signal in EPI) is unchanged in SIR EPI from classical EPI. In conditions requiring a minimum obtainable TE, as in the case of optimized diffusion imaging, the minimum TE of SIR is affected by the additional time of applying multiple excitation pulses plus the lengthened ADC read periods, and up to 10% SNR reduction has been found in SIR EPI. Using SIR in fMRI (functional MRI), there is no penalty in SNR, as TE is typically lengthened from the minimally obtainable TE since BOLD (blood-oxygen level dependent) contrast is optimized when TE = T2^*. A disadvantage of SIR EPI is that the increased number of encoded slices results in a lengthening of the flat top of the read gradient waveform, and a lengthening of the spacing of signals used to form each image. Consequently, there can be increased distortions within the image. The net echo train lengthening increases with more slices readout in SIR, so that the reduction of scan time is less than the fraction of number of slices compared to EPI. The longer echo train causes a relative delay in the minimum achievable TE in diffusion imaging, where the preparation pulses in the beginning of the sequence prevent an earlier beginning of the echo train signal readout, thus the lengthened signal readouts can delay the encoding of the center of k-space causing a drop in signal by T2 decay rate.

8] Another technique that is independent of SIR involves frequency-multiplexing of images by combining excitation of slices at different off-resonance frequencies with subsequent de-multiplexing based on spatial sensitivity differences of RF receiver coils, a technique referred to as Multi-Band (MB) excitation. Examples of the MB approach are described in (i) Moller S, Auerbach E, van de Moortele PF, Adriany G, Ugurbil K, fMRI withl 6-Fold reduction using multibanded multislice sampling, Proc. Int. Soc. Magn. Reson. In Med., 2008. 16: p. 2366, (ii) Feinberg, D.A., Moeller, S., Smith, S.M., Auerbach, E., Ramanna, S., Glasser, M.F., Miller, K.L., Ugurbil, K., Yacoub, E., 2010. Multiplexed Echo Planar Imaging for Sub-Second Whole Brain FMRI and Fast Diffusion Imaging. PLoS One 5, e15710, and (iii) Larkman DJ, Hajnal JV, Herlihy AH, Coutts GA, Young IR, Ehnholm G. Use of multicoil arrays for separation of signal from multiple slices simultaneously excited. J Magn Reson Imaging 2001 ;13(2):313-317. In MB excitation, increased efficiency is achieved by exciting several slices simultaneously. The MRI signals from those slices are unfolded using spatial encoding information present in RF receiver systems. Each of the several receiver coils yields a combination of MRI signals from all excited slices weighted by the sensitivity of the respective coil. A matrix inversion can provide a solution to unfold these signals so as to reconstruct MR images of the respective slices.

9] The MB acquisition of multiple slices at one time accelerates the volume coverage by the number of bands used in an MB RF excitation pulse (and thus the number of simultaneously excited and read out slices, and also results in reduced gradient demands and consequent reduced levels of acoustic noise for an un- accelerated acquisition of the same number of slices in which each slice is acquired separately. An MB MRI data acquisition technique available in MRI scanners from Siemens under the name SENSE provides a solution to aliasing. The separation of the aliased slice signals requires a different reference acquisition for GRAPPA (another pulse sequence provided by Siemens), but not for SENSE which directly separates aliased voxels. The two reconstructions have been shown to perform equally for GE (FLASH) imaging, but with GRAPPA being more desirable for high- field EPI imaging. The data size is reduced by a factor equal to the number of bands as several slices are contained within one matrix. Compared to equivalent multi-slice acquisitions needed to achieve the same number of slices, the repetition time TR is reduced by this same factor, allowing a larger number of slice images (and thus a better characterization of the temporal dynamics) to be acquired over the same time. Lastly, since each slice is excited and sampled identically, there is no significant SNR loss due to reduced data collection as is encountered with conventional parallel imaging along the phase encode direction, where under-sampling is used to accelerate the acquisition. There can be, however, SNR losses associated with separation of aliased image slices. A disadvantage of the multiband RF excitation is that RF heating is increased by the square of the power in the pulse, therefore the quadratic increase in RF power and dependent SAR (specific absorption rate) limits the number of slices. The ability to separate the slices by de-aliasing the net signal is limited by the coil sensitivity variations on the slice axis. A challenge in MB MRI is that the slices need to be sufficiently spatially separated to ensure that the spatial sensitivities of the receiver coils can separate the concurrently occurring MRI signals coming from different slices. If the slices are too close or adjacent to each other, there may be insufficient information to separate their concurrently occurring signals. Using more closely spaced receiver coils improves the ability to separate the information from different slices along the axis of receiver coil placement but this may adversely influence other factors involved in coil design.

[0010] Yet another technique, called POMP, is discussed in Glover G, Phase Offset Multi-Planar (POMP) Volume Imaging: A New Technique, J Magn Reson Imaging, 1991 , 1 : 457-461 . In POMP, several slices are simultaneously excited by a composite RF pulse. The centers of the reconstructed images in a POMP section are offset from each other in the phase encoding direction by means of view- dependent phase modulation of the RF excitation pulses and are placed adjacent to each other in the reconstruction. Using a reconstruction matrix that is increased in size, the images can be made non-overlapping and stored separately. At constant imaging time, signal-to-noise ratio and resolution, POMP is said to produce a factor N_p more sections than a conventional sequence but with a reduced field of view. Alternatively, it is said that imaging time may be increased by a factor N_p to retain the same field of view with the expected signal-to-noise advantage. The average RF power deposited by the 90° RF pulse is said to be greater by the factor N_p. With the constraints of either longer imaging time or reduced field of view, and SAR (specific absorption rate) concerns inherent in POMP, the approach may have limited use in imaging humans unless other mitigating approaches are applied.

[001 1 ] Yet another technique involves exciting a first volume (rather than a slice) with a first linearly varying phase with respect to k-space and a second volume with a second linearly varying phase with respect to k-space. The first and second linearly varying phases have different slopes. After Fourier processing, each of the two volumes is formed in a reduced field of view. Fig. 2a in U.S. Patent No.7,710,1 15 to Hargreaves illustrates the POMP technique referred to above, and Fig. 2b illustrates the volume rather than slice imaging technique of the Hargreaves patent.

[0012] Each of the SIR EPI, MB MRI, and POMP techniques can have advantages in reduction of acquisition time but also limitations including how many slices can be effectively imaged in one pulse sequence. For example, it was not contemplated that more than four slices could be effectively imaged in a single pulse sequence using SIR EPI, due mainly to the lengthening of the readout time with the number of slices. And, it was not anticipated that the MB technique could effectively acquire much more than 4 images simultaneously due to limitations of coil sensitivity and due to increased SAR (Specific Absorption Rate) from the higher RF power in the banded RF pulses. When the readout period is increased by the nature of SIR, the effects of Bo inhomogeneity increase and time-dependent errors can lead to image distortions related to the total readout time and the echo spacing. In SIR diffusion encoding techniques, signal decay by T2 increases by the added time taken to use stronger or longer duration gradient pulses. The diffusion encoding part of the sequence is followed by the Gr readout period and the TE effective of the image affects the image SNR. Delayed TE reduces the SNR as T2 decay increases in later TE. The SIR readout due to its greater number of MRI signals has an additional delay in the ko signals position of all the slices and, therefore, there is typically a delay of, for example, 5 to 15 ms in the TE, reducing SNR. The Hargreaves technique is a volume imaging technique rather than a slice imaging technique so it involves additional concerns.

[0013] Another technique, which can be referred to as multiplicatively multiplexed MRI (MM MRI) provides advantages associated with the discovery that certain limitations of SIR MRI and of MB MRI are sufficiently independent of one another to make it possible to intertwine desirable characteristics of each in a single pulse sequence. MM MRI does not lead to significant penalties in signal readout time lengthening or increases in SAR. While some of the respective limiting factors of known SIR and MB may increase by a relatively small and linear amount in MM MRI, there is a highly desirable and non-linear multiplicative increase in the number of resulting slice images when certain features previously used only in SIR or only in MB are entwined as described in more detail below. Certain aspects of this technique also are described in Feinberg, D.A., Moeller, S., Smith, S.M., Auerbach, E., Ramanna, S., Glasser, M.F., Miller, K.L., Ugurbil, K., Yacoub, E., 2010. Multiplexed Echo Planar Imaging for Sub-Second Whole Brain FMRI and Fast Diffusion Imaging. PLoS One 5, e15710.

[0014] The concurrent use in MM MRI of selected features from two very different multiplexing techniques, SIR and MB, in a single pulse sequence gives not a summation of their imaging speed effects but instead gives a multiplication of their acceleration factors, so that the average time of acquiring data for one slice image becomes divided by the product rather than the sum of the two accelerations. This is unlike acceleration such as by the techniques known as Partial Fourier and Parallel Imaging, which only reduce the time of the signal readout period and so reduce only a portion of the pulse sequence's total time and do not increase the number of image slices to be read out in the echo train, such that their combined effect is not as large, not much greater than a factor of 2 in imaging speed, particularly given the need for the specific image contrasts for BOLD and diffusion imaging that determine the minimum required TE.

[0015] With so many slice images that can be acquired, recorded and separated from each other with a single pulse sequence, practical applications of MM MRI to neuroscience and cardiac imaging can be particularly advantageous in dynamic MRI imaging, where the temporal sampling frequency in the repeated scanning of the organ can be significantly increased, beyond the EPI that is currently the typical technique in clinical use. EPI is believed to be the fastest and most efficient imaging technique commonly now used for dynamic measurements of BOLD fMRI in clinical practice, and for encoding hundreds of scans of the brain with different diffusional b- value weightings, or to make cine time series images of heart movement, or to measure the dynamic changes in a bolus of contrast passage through an organ to calculate blood perfusion. The much faster data acquisition in MM MRI is applicable to increasing the temporal sampling in these important physiological imaging techniques.

16] In MM MRI, frequency and temporal multiplexing are intermingled in a single, super-multiplexed pulse sequence to give multiplicative increases in EPI imaging speed while providing good image quality. SIR alone can reduce the bandwidth of signal readout, causing increases in distortions, although this can be negated using parallel imaging to shorten the echo train. MB alone can increase SAR, which can become prohibitive at high magnetic field human imaging. MM MRI is able to obtain closely spaced, even adjacent slices by in effect inserting SIR-separated slices between MB-separated slices. MM MRI applies plural kinds of signal multiplexing in a single pulse sequence such that SIR and MB features are used to multiplex images and Parallel Imaging is used to multiplex signals from within each image. The SIR-type multiplexing applies additional RF excitation pulses and records time- sequential MRI signals from different images. The RF excitation pulses in the super- multiplexed MM MRI pulse sequence are different from those used in SIR multiplexing in that they are modified to have specific spatial frequencies to create multiple bands across physical space where NMR excitation occurs. Spatial sensitivity differences in RF receiver coils are used to separate the MRI signals from different excitation bands. M bands are excited in each banded excitation RF pulse and S of these M-banded RF excitation pulses are applied in time sequence within the same EPI pulse sequence, to thereby create MRI signals from M x S slice locations, where each of M and S is a positive integer and preferably M>2 and S>2. Additional de-phasing gradient pulses are applied between the M-banded pulses to temporally encode each group of M bands differently. In one example, the M bands are widely spaced to coincide with the spatial sensitivity of phased array receiver coils. The S slices within each M band can be adjacent to each other so they fill up the spatial locations between the more widely separated M bands. The resulting S groups of signal are first separated by their timing in each refocused readout period of the EPI echo train. The M banded signals within each previously separated S demodulated signals are decoded using sensitivity differences of RF receiver coils to produce M images from each of the S groups of signals to give complete separation of M x S k-space data sets. Fourier reconstruction of 2D images can be used as known in MRI technology.

[0017] In a specific and non-limiting example, the MM MRI process comprises (a) applying, to a subject in an MRI scanner, a set of S radio frequency (RF) excitation pulses in a time sequence, each pulse S being a multi-band pulse that simultaneously excites M slices, one in each of M volumes or bands in the subject, thereby causing the subject to emit, in a time sequence, S MRI signals, where S and M are positive integers and S>2 and M>2, (b) refocusing to thereby generate a train of N sets of S MRI signals each, where N is a positive integer and N>2, (c) acquiring the MRI signals with at least one RF receiver coil system having spatial sensitivity characteristics, (d) computer-processing the MRI signals acquired with the at least one RF receiver coil system to produce magnetic resonance image data for SxM slices of the subject; and (e) further computer-processing at least some of the image data to produce and display magnetic resonance images of at least some of the SxM slices of the subject. When S>3 and M>4, magnetic resonance image data for 12 or more slices of the subject are produced. The refocusing can use refocusing magnetic gradients alternating in polarity such that the MRI signals in each succeeding set of the N sets of S MRI signals each are time-ordered in a reverse of their order in the preceding one of said N sets. It is recognized here that instead of exciting M slices, the slice thickness can be made greater by using a weaker slice selective gradient to create thicker slabs and these slabs can be phase encoded with gradient pulses applied to two gradient axes, for which 3D FT produces a set of T of images (where T is a positive integer and T>2) instead of an image of one slice, in which case MRI data for a total of MxSxT images are recorded in each pulse sequence. It is also recognized that the MRI signals can be oriented differently in k- space, using radial k-space trajectories or sinusoidal or spiral k-space trajectories.

[0018] A technique that overcomes many of the limitations and challenges described above can be called Multiplexed Shifted EPI. As described in more detail below, it involves exciting several slices such that the MRI signals of all the excited slices occurs substantially at the same time in each of the readout periods. This can be improved by the additional use of modulating controlled aliasing Gs pulses to improve the separation of the slices. The ko signal is shifted to a position in which the modulating phase gradient is passing through a zero point by means of rewind gradient pulse. By insuring that there is minimal or reduced intravoxel through slice dephasing with the shifted alignment of ko and the zero phase of the modulation gradient, the image signal is not cancelled and becomes considerably higher particularly at higher M accelerations. This approach to reading out signals in combination with parallel imaging helps reduce or eliminate the disadvantages of image signal loss E in MB EPI and MM EPI and other forms of imaging in which multiple slices are encoded and readout simultaneously or essentially simultaneously, and might additionally help in other imaging sequences that might benefit from reduced scan times. While parallel imaging (GRAPPA and SENSE) have been used extensively to shorten scan time by reducing the number of acquired signals, the new approach of multiplexed shifted EPI applies certain aspects of parallel imaging in a different way. Additional slices are encoded and produce signals that are readout simultaneously with the coil arrays on the slice axis. The parallel imaging can be used to increase the net FOV on the phase axis, thus to acquire an asymmetric FOV to separate the images. The multiplexed shifted EPI acquires fewer EPI echo trains so that the overall scan time is reduced. In one example of multiplexed shifted EPI the RF pulses in which there are different linear phase shifts and amplitude modulations applied during the readout of signals in the echo train. A second example, called shifted MB EPI or s-MB EPI in this specification, applies a different modulated phase shift on the Gs (slice select) axis to each of several consecutively applied slice selective pulses followed by a single EPI readout train. In a third example, called shifted SIR MB EPI or s-SIR MB EPI in this specification, a combination of composite RF pulses with frequency offset for MB excitation can be used in place of the single phase shifted single slice excitation, above, to give (MB) x (shifted SIR) number of slices. By the Fourier shift theorem, a phase shift due to phase modulation in the time domain signal (also known as k- space) is transformed into a displacement in the image domain. By acquiring an image with n-slice times the FOV required to cover the head or other body region on the image phase axis, the simultaneously excited images are separated along the image axis. This process requires n times as many phase encoded signals to enlarge the FOV or the different signals will simply be aliased with respect to each other in the FOV of one image.. The simultaneous image signals using MB by applying phase modulation so that their corresponding images that are simultaneously readout in the EPI echo train are separated on the phase axis of the image matrix. While signals from different excited image slices are simultaneously refocused in the echo train, the signals from different images are not separated in time on the multiple Gr refocused read periods and instead the slice signals occur superimposed and simultaneous with no time separation. The sequence excites slices with multiple selective RF pulses unlike existing SIR techniques, the small Gr pulses between the RF pulses are no longer used as they are not required to separate the signals later in time on the Gr readout periods.

19] This sequence differs from conventional MB in several ways. First, it does not require slices to have wide spacing between them on the slice axis. Second, it does not rely on coil sensitivity on the Gs slice axis to separate images. It takes advantage of multiple modulations in accumulated Gs (Gz) phase and parallel imaging on the in-plane phase axis of the image. The parallel imaging is used to measure additional signals to increase the FOV on the image phase axis. The signals are separated by the Fourier shift theorem as each excited slice has a different phase, and therefore is displaced into a different region of the 2D FT image FOV. The enlarged FOV by means of parallel imaging on the kp axis, not on the slice axis, allows complete separation of the slices which are forced to not overlap by means of matching the phase shift and FOV. The composite pulse is used in EPI, with the parallel imaging used to increase the FOV to avoid image overlap of EPI images while the total imaging time does not increase. Unlike POMP, the method does not require additional acquisition time but is acquired in a single echo train without acquiring additional signal. This can be combined with additional composite excitation pulses using phase offsets so that the overall RF heating is reduced from that which would result in exciting the same number of slices using a single POMP pulse incorporating the larger number of slices. Several possible combinations of the new S-SIR sequence, SIR and MB sequence can be made for advantages in reducing SAR heating, or increasing total acquisition speed.

20] The s-SIR EPI technique is changed to allow the signals from different slice images to be recorded simultaneously rather than at different times on each read period so there is no longer an increase in the net time of signal readout. The net lengthening of the signal readout causes image distortions and signal decay. Each of the RF excitation pulses has a different phase, with R(t) representing the RF envelope for a given linear-phase excitation pulse. An RF excitation R(t) with a linear phase shift is R'(t, y), where R'(t,k) = R(t)e' , with k_yy will with FT produce the image offset by y on the phase axis of the 2D image. Thus, an image whose center is offset from the origin by y may be produced by replacing the RF pulse envelope with one whose phase varies with as in the above equation. By the Fourier shift theorem, a phase shift after FT results in a displacement of the image. The encoded images using this s-SIR technique will be separated on the 2D image phase axis rather than in time on the read axis of k-space. The field of view (FOV) of PS-SIR must be increased by the number of SIR encoded images in order for the images not to overlap. The FOV is increased by using parallel imaging (GRAPPA or SENSE) by achieving smaller increments of sampling of the kp data axis. Smaller dK_p after FT gives larger FOV where FOV = 1/dK_p. Therefore the SNR reduction in the image is not due to accelerated imaging in which fewer signals are acquired as is used in SENSE and GRAPPA but rather it is only due to coil coupling losses, "g-factor." Each of the s-SIR EPI and the MB MRI data acquisition techniques has advantages in reduction of acquisition time but also limitations on how many slices can be effectively imaged in one pulse sequence. For example, it was contemplated that more than 4 slices could be effectively imaged in a single pulse sequence using s- SIR EPI, due to the increased g-factor coil sensitivity on the image phase axis. And, it is anticipated that the MB technique could effectively acquire 4 images simultaneously given limitations of coil sensitivity on the slice axis and due to increased SAR (Specific Absorption Rate) from the higher RF power in the banded RF pulses.

Brief Description of the Drawing

[0021 ] Fig. 1 illustrates and MRI super-multiplexed pulse sequence employing a new arrangement of certain features of Simultaneous Image Refocusing (SIR) MRI and Multi-Band (MB) MRI technologies.

[0022] Figs. 2a, 2b and 2c illustrate various aspects of the pulse sequence of Fig. 1 and its use.

[0023] Fig. 3 includes certain portions of Figs, 1 and 2a-2c and adds an illustration of using the super-multiplexed MRI signals to multiplicatively increase MRI signals acquisition speed and the number of slices compared to SIR and MB MRI.

[0024] Fig. 4 is a block diagram illustrating an MRI system using the pulse sequences discussed in this patent specification.

[0025] Fig. 5 illustrates a classical prior art spin echo pulse sequence with refocusing.

[0026] Fig. 6 illustrates a prior art SIR pulse sequence.

[0027] Fig. 7 illustrates a shifted SER EPI pulse sequence.

[0028] Fig. 8 illustrates a flowchart of main steps in using the pulse sequence of Fig. 7.

[0029] Fig. 9 illustrates a variant of the pulse sequence of Fig. 7

[0030] Fig. 10 illustrates a shifted MB EPI pulse sequence, Fig. 1 1 illustrates a related flowchart, and Fig. 12 illustrates a variant of Fig. 10.

[0031 ] Fig. 13 illustrates a shifted SIR MB EPI pulse sequence and Fig. 14 illustrates a related flowchart.

[0032] Fig. 15 illustrates a multiplexed EPI pulse sequence using a time sequence of composite RF excitation pulses and MRI signals that occur in time sequential groups, and Fig. 16 illustrates a related flowchart.

[0033] Fig. 17 illustrates main steps in a process using a pulse sequence as in Fig. 13.

[0034] Fig. 18 illustrates main steps in a process using a pulse sequence as in Fig. 7. [0035] Fig. 19 illustrates a multiband (MB) pulse sequence using controlled aliasing by

FOV/2 with no ko dephasing.

[0036] Fig. 20 illustrates a multiband (MB) pulse sequence using controlled aliasing by

FOV/3 with no ko dephasing.

[0037] Fig. 21 illustrates a multiband (MB) pulse sequence using controlled aliasing by

FOV/4 with no ko dephasing.

[0038] Fig. 21 illustrates sifted ko controlled aliasing EPI pulse sequence combined with additional phase variation in the RF pulses that leads to additional phase shifts in the

MR signal.

Detailed Description of Preferred Embodiments

[0039] In describing examples and preferred embodiments illustrated in the drawings, specific terminology is employed for the sake of clarity. However, the disclosure of this patent specification is not intended to be limited to the specific terminology so selected and it is to be understood that each specific element includes all technical equivalents that operate in a similar manner and features of one embodiment can be used in another.

[0040] Fig. 1 illustrates an example of a super-multiplexed MRI pulse sequence using the principles described above, in which MRI image data can be acquired for a total of 12 slices of a subject in an MRI scanner. A first excitation pulse 100a, labeled to "MB 90°," is a multi-band pulse that includes four frequency offsets, so that it contains excitation information for four different frequencies, slightly offset from each other as known in MB MRI, to excite four different volumes or bands in the subject. In combination with the concurrent gradient pulse illustrated on the Gs axis, RF pulse 100a excites four slices in the subject. This is illustrated in more detail in Figs. 2a, 2b and 2c, where Fig. 2a illustrates RF excitation pulse 100a concurrent with the gradient pulse 102a on the Gs axis. Fig. 2b illustrates the four slices 100a1 , 100a2, 100a3 and 100a4 in the head of a patient excited by the combination of pulses 100a and 102a. Fig. 2c illustrates the four frequency bands that are included in RF pulse 102a, in alignment with the slices in the patient's head that they excite. These frequencies are fo and three offsets from fo, one offset by an increment (-f), another offset by an increment (+f) and the third offset by an increment (+2f). Fig. 2c also illustrates the Gs gradient pulse 102 rising in the z-direction (along the length of the patient) from a magnetic field amplitude (-B) to (+B), so that the concurrence in time of the components of RF pulse 100a and slice select gradient 102a excites the four slices seen in Fig. 2b.

[0041 ] Returning to Fig. 1 , another RF excitation pulse 100b is applied to the subject in the MRI scanner after RF pulse 100a, concurrently with another magnetic gradient pulse 102b. Pulse 100b is otherwise similar to pulse 100a but differs in two important respects: (i) it is spaced in time from pulse 100a, and (ii) its base frequency content fo is slightly offset from the frequency content fo illustrated in Fig. 2c for pulse 100a. Because of this offset between the two base frequency contents fo of pulses 100a and 100b, the four slices in the subject's head that pulses 100b and 102b excite are offset in space from those excited by pulses 100a and 102a, so that now a total of 8 slices in the patient's head are excited. Later in time, RF pulse 100c and magnetic gradient pulse 102c are applied in a similar manner, except that the base frequency content of pulse 100c again is offset from those in pulses 100a and 100b such that an additional four slices of the patent's head are excited, for a total of 12 slices. Dephasing magnetic gradient pulses illustrated on the Gs axis in time period Tpre are used as is known in SIR technology.

[0042] After Tpre, a read gradient waveform is applied, alternating in polarity, so that three MRI signals 106c1 , 106b1 and 106a1 are included in the first readout period Tr1 , another three MRI signals are included in the second readout period Tr2 but in the reverse time order, another three MRI signals are included in the next readout period Tr3 but in a time order reversed from that in the preceding period Tr, and so on for a total of N readout periods Tr, with appropriate phase encoding as illustrated in the Gp axis. Thus, in this example, the pulse sequence illustrated in Fig. 1 applies to the subject a time sequence of S=3 RF excitation pulses 100a, 100b and 100c, each of said pulses being a multi-band pulse that simultaneously excites M=4 slices or bands in the subject, thereby causing the subject to emit, in a time sequence, S=3 MRI signals 106a, 106b and 106c, which MRI signals are refocused repeatedly to generate a train of N sets of three MRI signals each, containing MRI information from which MRI data can be derived for a total of SxN=12 slices in the subject, as described below.

[0043] Fig. 3 repeats some of the information of Figs. 1 and 2a-2c but adds an illustration of how the MRI signals are used in this example to produce MRI data for 12 slices in the patient's head. In upper left, Fig. 3 shows the pulses seen in the Tpre time period of Fig. 1 , and in upper right shows the MRI signals and the read magnetic gradient waveform 104 seen in the right hand portion of Fig. 1 . The MRI signals are detected with an RF receiving coil system having spatial sensitivity characteristics and are computer-processed into k-space data populating a total of 12 k-space matrices, although only three RF excitation pulses were used in this example. The contents of the 12 k-space matrices are further computer-processed, for example using the two-dimensional Fourier Transform (2D FT) process known in MRI technology, to produce and, if desirable, display on a monitor, images of some or all of the 12 slices of the patient's head illustrated in lower left in Fig. 3. The 12 slices are in M=4 groups of S=3 images, where the location of each image is indicated by a respective vertical line over the image of the patient's head.

[0044] In populating the k-space matrices seen in Fig. 3, the horizontal selection of a matrix to populate within each row of k-space matrices is based on the time sequence of the MRI signals resulting from the three RF pulses 100a, 100b and 100c, while the selection of a row of k-space matrices is based on frequency demultiplexing of the information content of the MRI signals, using for example a matrix inversion process applied to MRI signals obtained with RF receiver coil systems of appropriate sensitivity differences as described in the articles cited in paragraph 005 above, which are incorporated by reference. Each of the rectangles in lower right represents a k-space matrix for a respective one of the SxM=12 slices. For example, the uppermost left rectangle represents the k-space for the slice corresponding to S=1 , M=1 , the next rectangle in the same row represents the k- space for the slice corresponding to S=2, M=1 , and the rightmost rectangle in the same top row represents the k-space for the slice corresponding to S=3, M=1 . The rectangles in the second row represent the k-spaces corresponding to the three S slices in the group where M=2, the rectangles in the third row represent the k-spaces corresponding to the three S slices in the group where M=3, and the rectangles in the fourth row represent the k-spaces corresponding to the three S slices in the group where M=4. The arrows in each k-space represent lines in k-space and the direction in which they are populated. As in known SIR MRI and in EPI, the successive lines in k-space are populated in opposite directions. Only a few lines (4 lines in this illustration) are in each k-space matrix, to show the principle, although typically there would be a much greater number of lines, such as 256 or 512 or some other number much greater than the illustrated 4 lines. Parallel Imaging would populate several of these lines in k-space. Preferably each of the 12 populated k- space matrices is converted to an MR image of a slice of the subject, although this is not necessary - some of the k-space matrices may be not be converted so no MR images are generated for the corresponding slices. Similarly, preferably all k-spaces (12 in this example) are populated with respective MRI signals, but this is not necessary - some may be left empty.

5] It is possible with this technique to obtain with current commercially available MRI hardware up to 12 fold reduction in the acquisition time of multi-slice 2D EPI covering the brain in approximately 200 milliseconds instead of 2.4 seconds. Twelve images can be recorded in approximately 50 milliseconds instead of a single image in about 40 milliseconds using a classical single echo train pulse sequence, also called single-shot EPI. The pulse sequence would then be repeated using different frequency offsets of the excitation RF pulses to record images at different spatial locations. Using four excitations of the pulse sequence, 12x4=48 images would be produced in 200 milliseconds, enough to cover all regions of the human brain, or other body regions. Different number of M bands and S simultaneously refocused images can be used to obtain a multiplicative number of slices in each echo train. It is possible to increase the M factor by designing phased array receiver coils with many rows of coils to encode many bands of excitation. It is also possible to increase the S factor by reducing the time between each of the S excitation pulses and by reducing the time between the S signals in each readout period to maintain the signal amplitude in the presence of T2 and T2^* exponential decay parameters. Therefore it should be possible with further refinements that M=10 and S=10 could be used to give up to 10x10=100 images within a single echo train of EPI signals. This can be further improved by utilizing a parallel imaging reduction R-factor, to use multiple RF receiver coils within each banded region to reduce the number of required phase encoded signals to form each 2D image and to shorten the echo train so that T2 and T2^* decay occurs over less time for maintaining higher signal amplitude. The use of higher S (greater number of time-sequenced excitation pulses) that will increase the time of the read period and concurrently increase the echo train time, can be compensated by increasing the parallel imaging reduction factor R. Therefore larger S parameter with larger R parameter in combination with M parameter will produce super-linear accelerations of the number of slices within an essentially unchanged echo train time. The higher signal bandwidth and signal- to-noise reductions known to occur with higher R parameter caused by incomplete decoupling of receiver coils and by the reduced number of acquired signals can lead to noise reductions upon averaging in Fourier Transform image reconstruction can lead to overall SNR decrease in Multiplexed EPI. The SNR can be increased and the frequency sensitivity improved in each coil by using higher Bo magnetic field for MRI, currently up to 7 Tesla compared to more widely available MRI scanners of 1 .5 Tesla and 3.0 Tesla field strength.

[0046] The Multiplexed EPI principles described here can be extended to obtaining 3D FT images instead of 2D FT images and also can be used with multi-shot segmented data acquisition methods instead of single-shot data acquisition pulse sequences. The new method can be used with SE EPI utilizing M bands within the refocusing RF pulse in addition to the described technology.

[0047] One of the biggest technical challenges facing diffusion and functional MR imaging concerns the safety limitations imposed on the MR hardware so that the research subject or patient is not harmed. Specifically, the two major areas of concern are tissue heating arising from RF energy absorption (as quantified by the specific absorption rate, SAR) and nerve stimulation arising from the slope of switched gradient fields, dB/dt. Some modern scanners are already operated at the SAR and dB/dt limits that can be tolerated by humans. Therefore, in order to derive better spatial or temporal resolution from NRI scans, or to reduce examination times, it is important to design pulse sequences that optimize image acquisitions using the hardware available. The new method described here can provide significant advances in terms of imaging speed for both diffusion and functional neuroimaging, utilizing twin approaches to multiplexing that can be combined into a single, ultra-fast method.

8] Functional MRI (fMRI) based on blood oxygenation changes and measured using a gradient echo (GE) EPI sequence is now a commonly applied method to functional neuroimaging. Using the most modern generally available hardware at 3T, such as with a 32-channel receive-only phased array head coil, coupled with the most advanced pulse sequences presently generally available, with parallel imaging, still limits the performance of GE-EPI to around 2 mm spatial resolution with whole brain coverage in about two seconds. For conventional event-related fMRI analyses these performance specifications may be adequate because the hemodynamic response to neuronal activation is "low-pass filtered," peaking perhaps some five to six seconds after the stimulus event. Thus, with a TR = 2 sec the Nyquist condition should be satisfied. However, there are confounding physiological fluctuations, most notably brain pulsation coupled to the cardiac cycle, that occur at frequencies faster than 0.5 Hz. Any method that can maintain spatial resolution while improving temporal resolution has the potential to allow increased statistical power in an event- related fMRI scan, thus reducing the number of stimulus trials needed and reducing the scan duration. Furthermore, new 'data-driven' analysis methods are now replacing static mapping of activation locations with functional causality maps, the idea being to establish the temporal order in which brain regions are engaged in a particular task. Methods aimed at mapping causality, or any form of temporal information, benefit greatly from increased digitization, i.e. the largest degree of sampling per unit time. This can reduce artifacts, e.g. reduce ringing after Fourier transformation of a voxelwise time series data set, as well as improve the discriminating power between two events that are temporally close; while fMRI signals are inherently low-pass filtered by several seconds, there is a preservation of timing information as low as hundreds of milliseconds.

[0049] The Multiplexed-EPI variants of GE-EPI can permit significant gains in the temporal resolution of fMRI sampling rate in the data time series without any significant penalties in terms of spatial resolution. Using the lowest S and M factors of two for each and combining them will permit whole brain coverage with high spatial resolution of up to 2 mm in a greatly reduced TR of about 500 ms at 3T. This four-fold acceleration means that cardiac fluctuations would no longer alias, and there can be improved precision for advanced time series analyses, such as Granger causality. The temporal sampling rate arguments pertain to high field fMRI also but at 7T there are also large increases in contrast-to-noise ratio CNR and in BOLD contrast for a super-linear increase in sensitivity and CNR of functional signal of the brain.

[0050] The method described here therefore would reduce the overall acquisition time of multi-slice 2D EPI or other 2D multi-slice MRI techniques including those known as RARE and GRASE by means of intermixing features of two different multiplexing schemes such that super-linear increases obtain in imaging speed, as illustrated in Fig. 3 for the example of S=3 and M=4. Of course, a selection of S>3 and/or M>4 can be made within the principles of the new method.

[0051 ] Fig. 4 illustrates in block diagram form an MRI scanner system operating in accordance with the pulse sequences described in this patent specification. A patient 400 is on a patient support 40 of an MRI scanner 404 such that the anatomy of interest, e.g., the patient's head, is in an imaging volume of the scanner established by a source 406 of a steady magnetic field Bo, which can be a superconducting or other magnet. Schematically illustrated gradient coils 408 are pulsed with the magnetic gradient pulses discussed above that are illustrated on the Gs, Gr and Gp axis of Fig 1 , and an RF coil system 410 applies the RF excitation pulses and receives the MRI signals illustrated on the RF axis of Fig. 1 . The RF coil system can comprise sending coils and one or more receiving coils having sensitivity characteristics suitable to MB MRI imaging. The gradient coils 408 are powered and controlled through gradient pulse control unit 412 to generate the required gradient field pulses and the RF coil system 410 is controlled by an RF system pulse control unit 414 to apply the required RF excitation pulses and receive and digitize the MRI signals. Each of units 412 and 414 is in turn in communication with a computer control and signal processing unit 416 that operates in accordance with an MRI pulse sequence described in this patent specification to cause the generation of the required RF excitation pulses and magnetic gradient pulses and to receive and computer-process the MRI signals and produce MR images. A console 418 communicates with computer control 418 to initiate or otherwise control MRI sequences, including a sequence according to the new method described above, and display MRI images.

52] Thus, an example of the magnetic resonance imaging (MRI) method described above comprises the steps of: (i) applying, to a subject in an MRI scanner, a set of S radiofrequency (RF) excitation pulses in a time sequence, each pulse S being a multi-band pulse that simultaneously excites M respective slices in each of M volumes of the subject, thereby causing the subject to emit, in a time sequence, S MRI signals, where S>2 and M>2, but preferably S>2 and M>2; (ii) refocusing the signals to thereby generate a train of N sets of S MRI signals each, where N>2, but preferably N>2; (iii) acquiring the MRI signals with at least one RF receiver coils having spatial sensitivity characteristics; (iv) computer-processing the MRI signals acquired with said RF receiver coils to produce magnetic resonance image data for SxM slices of the subject; and (v) further computer-processing at least some of said image data to produce and display magnetic resonance images of at least some of said SxM slices of the subject. In the method, when S>3 and M>4, and the computer-processing can produce magnetic resonance image data for 12 slices of the subject, and preferably no less that (S+M+1 ) slices. The refocusing preferably is carried out using refocusing gradients alternating in polarity such that the MRI signals in each succeeding set of said N sets of S MRI signals each are time- ordered in a reverse of the order in the preceding one of said N sets.

[0053] In another example, this patent specification describes an MRI scanner system comprising (i) an MRI scanner having a source of a steady magnetic field Bo, sources of gradient magnetic fields acting on a subject in an imaging volume of the MRI scanner, and an RF system selectively applying RF excitation pulses to the subject and receiving MRI signals from the subject in response thereto; (ii) a control and signal processing computer system coupled with said RF system and said sources of gradient fields, said computer system being configured to: (a) apply to said subject, through said RF system, a set of S radiofrequency (RF) excitation pulses in a time sequence, each pulse S being a multi-band pulse that simultaneously excites M respective slices in each of M volumes of the subject, thereby causing the subject to emit, in a time sequence, S MRI signals, where S>2 and M>2, but preferably S>2 and M>2; (b) refocus the signals, through said sources of gradient pulses, to thereby generate a train of N sets of S MRI signals each, where N>2, but preferably N>2; (c) acquire the MRI signals through said RF system with spatial sensitivities encoded is the acquired MRI signals; and (d) computer- process the MRI signals acquired through said RF system to produce magnetic resonance image data for more than S+M but no more than SxM slices of the subject. Either the computer system or a separate console couples therewith can receive therefrom and further process said magnetic resonance image data and display MRI images based thereon on a monitor.

[0054] Yet another example can be embodied in a computer program product stored on a computer-readable medium in non-transitory form which, when loaded on and executed with an MRI scanner system comprising an MRI scanner having a source of a steady magnetic field Bo, sources of gradient magnetic fields acting on a subject in an imaging volume of the MRI scanner, and an RF system selectively applying RF excitation pulses to the subject and receiving MRI signals from the subject in response thereto, causes the system to carry out the steps of: (i) applying to said subject, through said RF system, a set of S radiofrequency (RF) excitation pulses in a time sequence, each pulse S being a multi-band pulse that simultaneously excites M respective slices in each of M volumes of the subject, thereby causing the subject to emit, in a time sequence, S MRI signals, where S>2 and M>2, but preferably S>2 and M>2; (ii) refocusing the signals, through said sources of gradient pulses, to thereby generate a train of N sets of S MRI signals each, where N>2, but preferably N>2; (iii) acquiring the MRI signals through said RF system with spatial sensitivities encoded is the acquired MRI signals; (iv) computer-process the MRI signals acquired through said RF system to produce magnetic resonance image data for more than S+M but no more than SxM slices of the subject; and (v) further computer-process said magnetic resonance image data and display MRI images based thereon on a monitor.

5] A different new pulse sequence, called shifted SIR EPI or shifted SER EPI (or shift SIR/SER EPI, or s-SIR/SER EPI) in this specification is illustrated in Fig. 7, and a flowchart illustrating main steps in its use in MRI is illustrated in Fig. 8. Three slices are excited in the example of Fig. 7, although a different number S>2 may be selected instead. Three 90° RF excitation pulses excite three slices in sequence. Each of these RF pulses has a different respective frequency offset f1 , f2 and f3, and a different respective phase Φ1 , Φ2 and Φ3. Each of the slice select pulses on the Gs axis is followed by a negative refocusing pulse that is half the time duration of the positive slice select. There is no gradient to separate the slices. The MRI signals from each of the three slices occur at the same time in each readout period on the Gr axis (each of the flat positive and negative portions of the read gradient alternating waveform). As illustrated in Fig. 8, in the more generalized case of exciting n slices, where n is a positive integer n>2, the first slice is excited with an RF pulse with phase shift Φ1 and frequency offset f1 , the second slice with Φ2,ί2, and the n-th slice with Φη,ίη (where n=N, and N is the total number of slices). The MRI signals on the RF axis are read out with a phased array RF receiver coil arrangement to separate the signal information by slice. A k-space matrix is used that is larger in size than for a single slice to correspond to a larger field of view (FOV) on the phase axis (NxFOV in size). 2D FT is performed for that larger k- space matrix using parallel image reconstruction. N images are shifted to non- overlapping locations according to the shift theorem and are displayed as MR images.

[0056] Fig. 9 illustrates a shifted SIR pulse sequence that is otherwise the same as in Fig. 7 but omits the refocusing negative pulses between the slice select pulses on the Gs axis, and alternates the polarities of the slice select pulses. As a result the slices that are imaged can be closer to each other and TE differences between the slices can be reduced. The flowchart of Fig. 8 applies to the pulse sequence of Fig. 9 as well to that of Fig. 7.

[0057] Fig. 10 illustrates a pulse sequence called shifted MB EPI of s-MB EPI in this specification. It excites multiple slices (3 in the illustrated example) with a composite RF pulse comprising the concurrent combination of 3 RF pulses, each with a respective phase and frequency offset. Again, the number of slices is n>2, selected with the slice select pulse illustrated on the Gs axis The MRI signals from all three slices occur at the same time in each readout period on the Gr axis, and are read out and separated using a phase array coil system. The flowchart in Fig. 1 1 illustrates main steps in using the sequence of Fig. 10 in the more general case. A composite RF pulse is created to excite a total of n>3 slices, each with a respective phase a frequency offset. The simultaneously occurring MRI signals in each readout period are read out with a phased array coil system, and are processed using FT (Fourier Transform or Fourier Series) processing using an image matrix that is larger in the phase axis direction (n x FOV) to accommodate the information for the n slices, using parallel image reconstruction. The n slice images are shifted to non- overlapping locations and displayed. Fig. 12 illustrates another variant of the shifted MB EPI pulse sequence that is particularly suitable for diffusion imaging. It differs from the Fig. 10 sequence only by including a 180° RF pulse between the slice excitation composite RF pulse and he MRI signal train and a concurrent slice select pulse on the Gs axis that is wide enough to encompass all n simultaneously excited slices (n=3 in this example). By making the 180° RF pulse broad to cover the three slices, the slices can be adjacent or at least close to each other and RF power for this pulse need not be high. The flowchart of Fig. 1 1 applied to the pulse sequence of Fig. 12 as well.

[0058] Fig. 13 illustrates a pulse sequence called shifted SIR MB EPI (or s-SIR MB EPI) in this specification, in which the MRI signals from multiple slices still occur at the same time in each readout period but the number of slices is multiplicatively increased, and Fig. 14 illustrates main steps in using this pulse sequence. A time sequence of n>2 excitation pulses in used in this example and each is a composite RF pulse exciting a respective plurality of n>2 slices. The number of excitation pulses need not be the same and the number of slices that each pulse excites, although the same designator n is used for each in Figs. 13 and 14. The first RF pulse excites n slices (e.g., n=3) as it is a composite RF pulse exciting the slices with different frequency offsets F1 , F2, Fn and different linear phase offsets Φ1 , Φ2, ... , Φη, such that the first composite RF pulse comprises sinc1 (F1 , Φ1 ), sinc2(F2, Φ2), ... , sincn(Fn, Φη). The second composite RF pulse uses the same linear phase offsets as the first RF pulse but different respective frequency offsets, i.e., uses a composite of F(n+1 ), Φ1 ); F(n+2), Φ2), ... , F(n+n, Φη), where F1≠F(n+1 ), etc. The third RF excitation pulse also is a composite pulse and has the same linear phase offsets as the first and second ones but has different respective frequency offsets, and so on if more than three time-spaced RF excitation pulses are used. In this manner, in the example of Fig. 13 3x3=9 slices are excited (and in the general case nxn slices are excited) and the MRI signal from all occurs at the same time in each readout period on the Gr axis. The slice select pulses on the Gs axis alternate in polarity and are followed by a half-width negative refocusing pulse. As illustrated in Fig. 14, the MRI signals are readout with a phased array coil system and separation of signals is performed on the basis of phase and amplitude differences in the RF receiver coil matrix. The resulting MR images are formed and displayed.

[0059] Fig. 15 illustrates a pulse sequence in which each of several time-sequential RF excitation pulses is a composite pulse exciting a respective set of several slices but the MRI signals occur in a time sequence for each set of slices. For example, if each composite RF excitation pulse excited 4 slices and there is a time sequence of 3 such pulses, the MRI signals for the 4 slices excited by the first composite pulse occur at the same time in each readout period, followed within the same readout period by the signals for the 4 slices in the second composite excitation pulse, followed again in the same readout period by the signals for the 4 slices excited by the third composite RF pulse. Because the slice select pulses alternate in polarity, the order of the MRI signal sets alternates from one readout period to the next. Fig. 16 illustrates main steps in using the pulse sequence of Fig. 15. The MRI signals are read out with a phased array coil system with coils on two axes in k-space, the phase axis kp and the slice axis. Slice separation of the MRI signals is performed based on phase and amplitude difference in the coil matrix, and the images are formed and displayed. The number of images in this case is the product of the number of slices excited by one of the composite RF pulses and the number of time- sequenced composite RF excitation pulses.

[0060] Fig. 17 illustrates several steps in a process using a pulse sequence such as in Fig. 13, where each of the time-sequenced RF excitation pulses is a composite pulse exciting a respective set of 4 slices. In the excitation step #1 , a phase shift Φ on each of the 3 composite RF excitation pulses moves the images to a different pixel shift (ΜΦ/2π) on the image phase axis. In the signal readout step #2, the MRI signals that are read out have phase shift but occur at the same time. Parallel imaging increases matrix size and field of view on the image phase axis. Thus, in the first readout period, three sets of MRI signals are readout, each set illustrated by a respective arrow and comprising signals for four slices. Thus, each readout period provides information for 4x3=9 slices. With 2D FT, the slices are separated in step #3 in a larger FOV, where each of the three SIT images contains 4 aliased MB images. With MB image processing reconstruction, a total of MBxs-SIR images (4x3=9) is obtained from each EPI sequence and is available for display.

[0061 ] Fig 18 is similar in other respects to Fig. 17 but the RF excitation pulses are not composite pulses, i.e., the pulse sequence is as in Fig. 7. In step #2, the MRI signals that occur in a single readout period are for the three slices corresponding to the three RF excitation pulses in this example. In Fig. 18, each of the arrows in step #2 represents a single slice. The order is the same because the slice select pulses on the Gs axis are the same polarity. In step #3 in this case each of the three sSIR images is a single image. These slices represented by these images can be adjacent or very closely spaced, unlike the case with conventional MB images that need to be widely separated because of the nature of conventional MB EPI. If the slices are spaced in the z-direction, the frequency offset determines the spacing in this case.

62] Fig. 19 illustrates an MRI pulse sequence according to one example of the new technique of controlled aliasing with no ko dephasing. The normal blipped phase encoding gradient pulses applied on the Gp axis occur time when the read gradient Gr is being switched. An initial larger area gradient pulse of opposite polarity is applied on the Gr axis before the read gradients so that the net accumulated phase encoding of signals will linearly traverse k-space and will pass through ko in k-space when the net Gp phase encode gradients sum is at or close to zero. With no dephasing, the ko signal has maximum or near maximum amplitude and contributes a large amount of the images' total intensity. The Setsompop et al. abstract cited above illustrates how a modulated pattern rather than a linearly increasing pattern of blipped gradient pulses on the slice gradient axis, Gs, applied between the signal readout periods at the same time of the Gp blipped phase encoded gradients, causes controlled aliasing (displacement on the image phase axis) of the different simultaneously encoded images in multiband imaging. This used periodic refocusing pulses (also called rewinder pulses) to undo the dephasing effects of prior pulsed Gs gradient pulses. Several different modulation patterns of Gs can be applied, to cause different field of view (FOV) shifts ½, 1/3, according to the pattern of Gz blipped pulses. FOV controlled aliasing with shift = ½ is created by alternating pattern of polarity of Gz, i.e. +1 , -1 , +1 , -1 ... where 1 is an arbitrary unit of phase accumulation adjusted to give the desired FOV shift, and similarly +1 , +1 , -2, +1 , +1 , -2, give FOV 1/3 controlled aliasing shifts, respectively. However in Setsompop et al. abstract only a low MB factor of 3 was shown, and there is no recognition or a teaching of effects of the relative phase pattern of Gz (Gs) pulses with respect to the concurrent Gp normal phase encoding pulses.

[0063] In contrast, according to this patent specification benefits accrue by incorporation of controlled aliasing Gs blipped modulated pulses into both MB EPI and SIR x MB EPI sequences in a specific constrained combination to always have the ko signal at or near maximal amplitude signal coincide with a node in the Gs accumulated modulation gradient when there is no or little dephasing effect of the accumulated blipped Gs pulses. By doing this, there is no dephasing of ko by the Gs modulated gradient pulses in the controlled aliasing scheme. The Gs phase is at or near zero when the Gp phase is at or near zero (at ko in k-space). This is particularly important when MB factor becomes larger than 3 slices, and the ko signal becomes higher amplitude. If there were an effective Gs dephasing of ko at MB > 3 then there would be increasing intervoxel dephasing of the simultaneously acquired slices and large areas of attenuated signal would occur. The Setsompop et al. abstract did not teach this and does not show the concurrent zero phase of Gp and Gs. It is not apparent from the abstract or other known prior material related to MB sequences how the controlled aliasing use of blipped Gs pulses would cause a severe artifact if lower MB factors of 3 or 2 were being used because the signal peak would be lower and the dephasing interference between the simultaneously acquired signals at different slice positions on the Gs axis would give a much less severe signal loss. It is not apparent from such earlier proposal, and perhaps is counterintuitive that the signal loss in the image will become increasingly worse as the MB factor increases from 4 to 6, 8, 10,12.... and higher.

[0064] Figs. 19-21 illustrate different representative controlled aliasing Gs modulation schemes in which the ko phase encoding signal and the controlled aliasing periodic defocusing and refocusing are concurrent. In EPI usually more than 30 to a 100 echoes are acquired and the ko is towards the middle of the echo train. Partial Fourier in EPI can shift the ko to an earlier TE with a shortened echo train. The Gs blipped gradients are applied to periodically refocus to a zero phase point throughout the echo train. Many different modulations can be applied. When the net summed moment of the different polarity gradients sum to zero or close to zero then the phase effect will also be refocused completely or essentially completely. For example the FOV/3 controlled aliasing of images is accomplished by the periodic polarity and relative amplitudes (-1 ,-1 ,+2,-1 ,-1 ,+2...) where the sum -1 ,-1 ,+2 pulses give total refocusing on the subsequent echo.

[0065] Figs. 19-21 shows relative net amplitude x duration of pulsed Gp gradients.

There is net zero accumulated phase at the third echo in this representative sequence. The controlled aliasing pulses applied on the Gs gradient axis shows relative amplitude x time of these pulses. Note the amplutude x time of the Gs and Gp pulses can be different, hence it is only one representative example shown here where they have the same +1 area, while in practice the Gs and Gp blipped pulses can and typically will have different areas (also known as gradient moment). The critically important feature newly introduced in this patent specification is to have the ko signal occur at one of the many different periodic points following a phase rewind gradient pulse causing the net Gs phase to become zero or close to zero, hence the signal has complete or essentially complete refocusing.

[0066] The EPI sequence can also have additional short echo trains of 2-3 echoes without any phase encoding gradient pulses, placed before the actual image echo train. These so called, navigator echoes are used to measure the timing errors in echo position and serve to align the actual image data in k-space. It is important that any accumulated gradient moment causing phase shifts in the ko has matched phase shifts in the navigator echoes. In the ideal case where ko has no gradient moment dephasing, the navigator echoes will similarly have no net gradient moment which would cause dephasing.

[0067] It is believed that in the work of the lead author of the Setsompop et al. abstract, an additional so called 'pre-winding' pulse on the Gs axis prior to the initial signal is replaced with a smaller balancing blip pulse of area Aprewind = -Ablip/2, where Ablip is the applied area of the subsequent blipped gradient pulses. This balancing blip is used to match the phase of the two edges of the image slice to reduce a potential small residual ghost artifact in images. The balancing blip pulse is not used to impose concurrent timing of Ko signal by shifting the position of Gs refocusing of any of the phase refocused points in the signal echo train. In fact, as described for N/2 FOV shift, the balancing pulse has exactly half the effect on the phase of signals compared to the subsequent Gs blipped pulses hence it is not intended to shift the position in the echo train where the phase is completely refocused and cannot in fact be used to impose zero net gradient moment on Ko, the strongest signal, to avoid it being dephased and reduced in amplitude with resulting loss of signal in images.

[0068] As the number of simultaneous slices increases with larger MB, the spatial separation of slices reduces and the slices are closer together, which can make it more difficult to identify differences in their respective coil sensitivity profiles, causing poorer separation and reduce the effect of controlled aliasing separation. In general, this new shifted Ko to zero net gradient moment dephasing encoding scheme in the EPI sequence can be further modified with refocusing 180 degree RF pulses and diffusion gradient pulses prior to the echo train, as well as many other labeling schemes, including ASL and inversion recovery pulses.

[0069] One potential improvement illustrated in Figure 4, of the shifted phase is to acquire either a longer echo train, or to use parallel imaging techniques to synthesize additional echoes in k-space and to create a much larger field of view along the phase encoded image axis. This is similar to POMP in which the simultaneously recorded slices would be moved by means of larger phase modulation frequencies, to a completely separated non-overlapping position in the larger FOV on the phase axis. A higher frequency modulation on the Gs axis can be accomplished by creating interference patterns in summed phase of signal from simultaneously recorded signals from the different slice. Imposing a linear or nonlinear phase shift in the RF pulses combined in the MB pulse will give an additional modulated phase variation as the modulated Gs blipped pulses impose the controlled aliasing of these simultaneously recorded slices. By either acquiring an N times longer echo train, with a 1/N reduction in amplitude of the Gp phase encoding blipped gradient, then the FOV of the image will be increased by N whereas the spatial resolution will remain unchanged. In Fig. 4, a sifted ko controlled aliasing EPI pulse sequence combined with additional phase variation in the RF pulses that leads to additional phase shifts in the MR signal. The arrows in each cluster of 3 arrows pointing in different direction represent respective slice signals - thus, MR signals for 3 slices are acquired in each readout period represented by a positive gradient pulse on the RF axis.

70] The above specific examples and embodiments are illustrative, and many variations can be introduced on these examples and embodiments without departing from the spirit and scope of the disclosure. For example, elements and/or features of different illustrative embodiments may be combined with each other and/or substituted for each other within the scope of this disclosure.

Claims

CLAIMS:

1 . A magnetic resonance imaging (MRI) method comprising:

applying, to a subject in an MRI scanner, radiofrequency (RF) excitation pulses that simultaneously excite M respective slices of the subject, where M>3, and produce time-sequential sets of M concurrent MR signals related to the M slices;

applying magnetic gradient pulses to the patient in a sequence that reduces the net gradient effects in the peak MR signals to reduce dephasing that would otherwise reduce an amplitude of the MR signals at a selected point in k-space and/or increase interference between concurrent MR signals for different slices;

acquiring the MRI signals;

computer-processing the acquired MRI signals to produce magnetic resonance image data for at least some of said M slices of the subject; and

further computer-processing at least some of said magnetic resonance image data to produce magnetic resonance images of at least some of said M slices of the subject.

2. The method of claim 1 in which said applying of gradient pulses comprises applying refocusing gradient pulses on one axis and applying dephase-reducing gradient pulses on another axis at times and amplitudes causing controlled aliasing by FOV/M that reduces ko dephasing, where FOV is a field of view.

3. The method of claim 2 in which the dephase-reducing pulses are applied to a Gs axis and comprise alternating sets of (M-1 ) pulses of one polarity followed by a pulse that has an opposite polarity and a time-amplitude area greater that a preceding one of the (M-1 ) pulses.

4. The method of claim 3 comprising applying the refocusing pulses on a Gp axis, at times related to transitions in polarity of readout pulses, wherein said refocusing pulses are combined with other pulses on the Gp axis to cause a null of cumulative rephrasing at the time of the MR signals for a ko portion of k-space.

5. The method of claim 4, in which the refocusing pulses on the Gp axis and the dephase-reducing pulses on the Gs axis coincide in time.

6. The method of claim 1 including parallel imaging to multiplex MRI signals.

7. The method of claim 1 including displaying magnetic resonance images of at least some of said M slices of the subject.

8. The method of claim 1 including utilizing said magnetic resonance images in functional MRI (fMRI) studies.

9. A magnetic resonance (MRI) scanner system comprising:

an MRI scanner having a source of a steady magnetic field Bo, sources of gradient magnetic fields acting on an imaging volume of the MRI scanner for containing a subject, and an RF system selectively applying RF excitation pulses to the imaging space and receiving MRI signals from the imaging space and subject therein in response thereto;

a control and signal processing computer system coupled with said RF system and said sources of gradient fields, said system being configured to cause said MRI scanner to:

apply radiofrequency (RF) excitation pulses that simultaneously excite M respective slices of a subject in the imaging space, where M>3, and produce time-sequential sets of M concurrent MR signals related to the M slices;

apply magnetic gradient pulses to a patient in said imaging space in a sequence that reduces the net gradient effects in the peak MR signals to reduce dephasing that would otherwise reduce an amplitude of the MR signals at a selected point in k-space and/or increase interference between concurrent MR signals for different slices;

acquire the MRI signals;

computer-process the acquired MRI signals to produce magnetic resonance image data for at least some of said M slices; and further computer-process at least some of said magnetic resonance image data to produce magnetic resonance images of at least some of said M slices.

10. The system of claim 9 in which said system is further configured to cause the scanned to apply refocusing gradient pulses on one axis and applying dephase- reducing gradient pulses on another axis at times and amplitudes causing controlled aliasing by FOV/M that reduces ko dephasing, where FOV is a field of viewg.

1 1 . The system of claim 10 in which said system is further configured to cause the scanner to apply the dephase-reducing pulses to a Gs axis in a sequence of alternating sets of (M-1 ) pulses of one polarity followed by a pulse that has an opposite polarity and a time-amplitude area greater that a preceding one of the (M-1 ) pulses.

12. The system of claim 1 1 in which said system is further configured to apply the refocusing pulses on a Gp axis, at times related to transitions in polarity of readout pulses, wherein said refocusing pulses are combined with other pulses on the Gp axis to cause a null of cumulative rephrasing at the time of the MR signals for a ko portion of k-space.

13. The system of claim 12 in which said system is further configured to cause the refocusing pulses on the Gp axis and the dephase-reducing pulses on the Gs axis to coincide in time

14. The system of claim 13 in which said system is further configured to to display at least some of said magnetic resonance images.

15. A computer program product comprising computer-readable programs stored on a computer-readable medium in a non-transitory form which, when loaded on and executed with an MRI scanner system comprising an MRI scanner having a source of a steady magnetic field Bo, sources of gradient magnetic fields acting on a subject in an imaging volume of the MRI scanner, and an RF system selectively applying RF excitation pulses to the subject and receiving MRI signals from the subject in response thereto, causes the MRI scanner system to carry out the steps of:

applying radiofrequency (RF) excitation pulses that simultaneously excite M respective slices of a subject in the imaging space, where M>3, and produce time-sequential sets of M concurrent MR signals related to the M slices;

applying magnetic gradient pulses to a patient in said imaging space in a sequence that reduces the net gradient effects in the peak MR signals to reduce dephasing that would otherwise reduce an amplitude of the MR signals at a selected point in k-space and/or increase interference between concurrent MR signals for different slices;

acquiring the MRI signals; computer-processing the acquired MRI signals to produce magnetic resonance image data for at least some of said M slices; and further computer-procesing at least some of said magnetic resonance image data to produce magnetic resonance images of at least some of said M slices.

16. The computer program product of claim 15 in which said product causes the scanned to apply refocusing gradient pulses on one axis and applying dephase- reducing gradient pulses on another axis at times and amplitudes causing controlled aliasing by FOV/M that reduces ko dephasing, where FOV is a field of viewg.

17. The computer program product of claim 16 in which said product causes the scanner to apply the dephase-reducing pulses to a Gs axis in a sequence of alternating sets of (M-1 ) pulses of one polarity followed by a pulse that has an opposite polarity and a time-amplitude area greater that a preceding one of the (M-1 ) pulses.

18. The computer program product of claim 17 in which said product causes the scanner to apply the refocusing pulses on a Gp axis, at times related to transitions in polarity of readout pulses, wherein said refocusing pulses are combined with other pulses on the Gp axis to cause a null of cumulative rephrasing at the time of the MR signals for a ko portion of k-space.

19. The computer program product of claim 18 in which said product causes the refocusing pulses on the Gp axis and the dephase-reducing pulses on the Gs axis to coincide in time

20. The computer program product of claim 19 in which said product causes the scanner to display at least some of said magnetic resonance images.