Echo Planar Imaging in MRI - Essay Example

EPI in MRI Optimizing EPI Image Echo Planar Imaging (EPI) is among the most efficient magnetic resonance imaging (MRI) techniques. However, this technique is predisposed to severe artefacts, and as a result, it has proved to be a technically demanding experiment. More often than not, there are a number of parameters which have to be changed in the event that and EPI image has to be optimized. These parameters include magnetic susceptibility, k-space line, spin echo sequence, duration of the EPI trajectory, phase direction and gradient (Ye et al, 1996, p.219). Magnetic susceptibility refers to the effect of a tissue on the magnetic field. Notably, the boundaries existing between fat and compact bone are affected by the rapid de-phasing of the transverse magnetic field as well as distorts the signal. Resultantly, this leads to the wrong positioning of the frequency in the phase direction. K-space line is another parameter that could be optimized so as to change the EPI. A shift in the k-space line from its normal line leads to geometric distortion (Amin & Afzal, 2009, p.232). Such a shift may be resulting from field in-homogeneities. However, this artefact is reduced by spin echo sequences, which do rephrase the spins each time a 180-degree pulse is applied. On the other hand, gradient echo structures of the refocusing of the artefact. Undeniably, EPI makes use of a gradient echo sequence, thus leading to the accumulation of the effect with no radiofrequency (RF) pulse to correct the same. It is worth noting that any accumulation of errors arising from phase encoding will be wrongly registered, thus exposing EPI to this artefact. In the event that the duration of the EPI trajectory if considerably long, the EPI itself gets to be affected even by small field in-homogeneities (Ye et al, 1996, p.219). Change in the parameter can be so achieved through increasing the bandwidth. Such an action will lead to a reduction in the echo time (TE). So as to realize more ramp sampling, the dwell time can be increased. In optimizing EPI, the frequency of the phase encoding can be increased using a zero filling. The end result of this is an increase in resolution. On the other hand, TE can be increased so as to increase the transverse magnetization, thus giving a greater T2 and T2*. Moreover, a gradient with a relatively high performance can be used in optimising EPI (Ye et al, 1996, p.220). Consequences of Increasing the Receiver bandwidth Bandwidth represents the frequency range resulting from the already read gradient across the field-of-view (FOV). This measure is quite imperative in the determination the decay filter effects of signal-to-noise (SNT) ratio, T2 and distortion. It is also equally important in determining how well a sequence gets to be flow compensated (Ben, 2012). Based on the type of pulse in use and the strength of the scanner gradient, bandwidths may necessitate modification (Cohen, 1998, p.5). The acquisition or receiver bandwidth, denoted as rBW, refers to the range of frequencies that the receiver can accept for purposes of sampling the magnetic resonance signal. This rBW is changeable, and as a result, it has a direct relationship with the SNR. For instance, if the receiver bandwidth is increased, the SNR is reduced; thus, there is more noise from the outside of the spectrum. Nevertheless, a larger receiver bandwidth permits considerably faster imaging (McCarthy et al, 1993, p.4952). Different resonant frequencies cause a chemical shift artefact in the frequency encoding direction in voxels enclosing fat and water. In the same manner, incorrect larmor frequencies due to increased receive bandwidth also causes geometric distortion. These larmor frequencies are usually formed around metal embeds (Ye et al, 1996, p.220). On increasing the receiver bandwidth, the chemical shift artefact is measurably reduced. This, in some scanners, is so realized in the event that the water-fat shift parameter gets to be abridged. This leads to the range of resonant frequencies (over which the distortion is spread) to occupy a relatively smaller pixel range. In addition, the in-plane geometric distortion now occupies a smaller area within the field-of-view. However, it is advisable that if the scanner has an option of using a higher gradient performance, the same ought to be selected. The argument behind this is that in the event that all other factors are held constant, higher receiver bandwidth is achievable through the employment of a considerably higher frequency encoding gradient heft. Resultantly, the freedom to cut down the geometric distortion will have lengthened. Moreover, it is worth noting that an increased rBW allows a shorter echo-spacing in the turbo spin echo (TSE) echo train as well as a shorter minimum echo delay time (TE) (Ye et al, 1996, p.220). Benefits of circular spiral and/or square spiral EPI methods There are a number of alterations to the Echo Planar Imaging method. Some of these include constant phase encoding, circular square sampling and circular spiral sampling (Ben, 2012). As is the case with EPI, spiral imaging (be it circular or square) is a considerably faster magnetic resonance imaging technique that is capable of efficiently sampling the all of the k-space of interest in a single segment or even in countable k-space segments. Rather than using a zigzag route so as to map out a Cartesian grid in k-space, spiral scanning ordinarily makes use of an Archimedean spiral path in the event of the complex vector k (Ye et al 1996, p.220). This aspect ensures that along the axes of the Cartesian, the criterion so used in sampling the correct FOV representation is adequately satisfied. On the aspect of spatial resolution, the number of the spiral resolutions has to be half the number of equidistant samples along the axes of the Cartesian. This number of the spiral resolutions is the one responsible for determining the range of the monotonic scalar function. Additionally, since that spiral EPI does cross the Cartesian axes the same number of times, it is possible for an individual to work with a square (an isotropic) FOV with the same in-pane resolution in whatever dimension. Spiral EPI methods also prolong the limits of the k-space (Ye et al 1996, p.220). Notably, spiral EPI methods permit considerably shorter signal acquisition windows and echo times as compared to EPI. This is because the spiral methods have relatively low sensitivity to movements of the brain and have efficient gradient utilization. These two aspects result in shorter acquisition windows. As a result, these spiral images display a higher quality whose resolution of structural borders and details are better. Moreover, spiral imaging ensures that there is less distortion of the image (Amin & Afzal, 2009, p.233). With the spiral readouts starting at the echo time (TE), it is certain that all components of higher spatial frequency will be gathered for longer evolution times. In summary, therefore, there are principally three advantages that both circular and square spiral EPI methods present. Firstly, it is the circularly symmetric T2 weighting which is as a result of the circularly point spread function in the domain of the image. Secondly, spiral EPI methods do eliminate perceived discontinuities in gradient waveforms. The advantage of this is that it leads to a reduction in the initial transient and steady-state distortions. Spiral EPI methods also see to it that an effective rapid spiral-scan to a high frequency from a dc is continuous. This ensures the realization of multiple pulsing with interlacing for further improvements in resolution without decay image distortion (Ye et al 1996, p.220). Advantages of using segmentation in EPI As discussed earlier, EPI is ordinarily characterized by geometric distortion, and as a result, there is a need to ensure that this distortion is reduced or done away with. One of the most commonly used techniques to reduce the above mentioned geometric distortion is through the segmentation of the acquisition. Segmentation, also referred to as multi-shot EPI or parallel imaging, entails the division of the k-space into various segments which are filled in an interleaved manner (Cohen, 1998, p. 7). Besides, the echo time in multi-shot imaging denotes the time from the very end of the magnetic isocentrer to the centre of the k-space view or segment. However, it is worth noting that there is a need for numerous repetitions to be carried out so as to ensure that the entire k-space is covered. In this case also, the degree of segmentation can be the number of shots used for purposes of acquiring the full image or even the length of the echo train per excitation (Ye et al, 1996, p.220). For instance, 32 shots having 256 phase encoding steps usually have an echo train length of 8. For many years, the single-shot EPI has been the dominant class of pulse sequence in the majority of the imaging applications. Nevertheless, multi-shot acquisition, which we refer to as segmentation, has represented a viable alternative. Segmentation in EPI has brought about a number of advantages as will be discussed below (Ben, 2012). One of the key advantages of segmentation is that there is no more limitation of the realizable spatial resolution resulting from the anticipated echo time. The argument behind this advantage is that the readout of a single segment can now be made subjectively as brief as possible. It is for this reason, therefore, that segmented EPI gets to be more often than not preferred for functional MRI, whose resolution is very high (in sub-millimetre) (Ye et al, 1996, p.220). In EPI, segmentation is also used for purposes of increasing resolution. This has facilitated the application of high-resolution functional magnetic resonance imaging in very high field human MRI. Through this increased resolution, there are high possibilities that improvements will be realized in SNR as well as in the BOLD contrast relative to f-magnetic resonance imaging. Moreover, the achieved high resolution broadens the application of very high field magnetic resonance imaging (McCarthy et al, 1993, p.4954). At such high resolutions, any uncorrected geometric distortions are in the offing of entirely shifting the functional activity out of the cortical reconstruction. This will, in the end, lead to such an activity becoming effectively undetectable to analyses which are surface-based. In such a scenario, it may be approvingly worthwhile minimizing distortions noticed at the outset, thus preserving the rationality of a consequent surface-based analysis (Ye et al, 1996, p.221). Another advantage of segmentation is the fact that it places lesser stress on the gradients as compared to the single-shot echo planar imaging approach. In slice direction, gradients are predisposed to causing dropout. These dropouts are more likely to cause geometric distortion. As earlier mentioned, segmentation is characterized by high resolution, and this increases the strength of the gradient to be so applied (Amin & Afzal, 2009, p.234). Resultantly, the increased gradient strength culminates into faster traversal of the k-space, and as a result, lowers distortion. In the multi-shot approach, this aspect is very imperative and more especially in the event that the available signal-to-noise ratio and hardware make obtaining of the necessary k-space data somewhat difficult prior the elimination of the magnetic resonance through transverse relaxation. Moreover, this makes segmented EPI to be able to run on conventional systems without complications, where it counterpart (the single-shot EPI) may not (Ye et al 1996, p.221). In segmented EPI, phase errors usually have less build up time as compared to the case with single-shot EPI. The benefit of this reduced phase errors is reduction in the magnetic susceptibility artefacts. What the segmentation does is effectively resetting background field gradients by dividing the entire k-space traversal into numerous segments, thus increasing the effective readout bandwidth proportionately to the number of shots. With an increase in the degree of segmentation, the number of echoes so acquired per radiofrequency excitation gets to be proportionately reduced (Ye et al, 1996, p.221). As a result, phase errors equally reduce, leading to a reduction in the geometric distortion. In multi-shot EPI, field inhomogeneity is straightforwardly visualized and assessed in the domain of spatial frequency data (the k-space). The echo time shift approach incrementally alters the positioning of the echo train, thus making improvement of the phase error function via the distribution of phase discontinuities far away from the k-space centre. In addition, segmentation allows the introduction of T1 weighting (Ben, 2012). Disadvantages of using segmentation in EPI On the other hand, segmentation in EPI presents various disadvantages. To begin with, parallel imaging is very sensitive to shot-to-shot instabilities resulting from respiration, head motion or even instabilities borne of instrumentation. As a result, there are high likelihoods that multi-shot imaging will display reduced temporal signal-to-noise ratio when it is compared to the single-shot approach. Moreover, multi-shot imaging takes a longer period for it to perform (Ye et al, 1996, p.221). The scan time in segmented EPI equals repetition time (TR) multiplied by number of segments (Ns) multiplied by the number of acquisitions (NEX). As a result of this longer time frame, it is likely that segmented EPI will be more predisposed to motion artefacts as compared to single-shot EPI (Cohen, 1998, p.12). In a multi-shot acquisition, there are different shots and any motion – be it respiration or head motion – during the diffusion encodings cause phase errors. In addition, phase differences in k-space data among the different shots lead to the cancellation of phase artefacts necessary in the reconstruction of the image. The subsequent correction of the phase error via the use of low-resolution phase subtraction may prove to be incomplete due to the under-sampling of the regions of the shot. As earlier noted, segmented EPI has a complex coverage of the k-space. This predisposes the approach to various specific artefacts (Ye et al 1996, p.222). Differences between SMASH and SENSE in Parallel Imaging From the elements in an EPI array, each and every image has a reduced FOV. Besides, the area of interest is normally under-sampled. This, therefore, calls for aliasing or image wrapping. Image wrapping in EPI refers to the introduction of an error or an artefact in the sampling of an intermittent signal in the event that the sampling frequency is relatively low to properly capture the signal. This, therefore, generates a false frequency alongside the correct one during the process of frequency sampling (Ben, 2012). Thus, there is a need to correctly bring together the information so obtained from the partial phase encoding from each and every coil element. In so doing, two methods are used, and these include sensitivity encoding (SENSE) and simultaneous acquisition of spatial harmonics (SMASH) (Ye et al, 1996, p.222). Both SENSE and SMASH are parallel acceleration techniques, but each of them is unique. SENSE and SMASH are magnetic resonance imaging techniques which are designed for purposes of decreasing the scan time. This reduction is so achieved via under-sampling of the k-space and the simultaneous recording of images from numerous imaging coils. The role that the under-sampling does is to reduce the acquisition time. On the other hand, the use of the numerous radiofrequency (RF) coils does eliminate the wraparound resulting from the under-sampling (Ye et al, 1996, p.223). In Sensitivity Encoding, image processing is attained on the image domain data. SENSE is usually founded on the fact that receiver sensitivity in normal circumstances has an encoding effect which compliments the Fourier preparation by field gradients which are linear in nature. As a result, this reduces the use of the multiple receiver coils in parallel scan time (McCarthy et al, 1993, p.4955). SENSE also enables a considerable reduction in the scan time in MRI. This is so realized via the utilization of the spatial information related to the receiver coils to reduce the conventional Fourier encoding. Principally, SENSE is applicable to any imaging sequence as well as k-space trajectories. Nonetheless, this approach is predominantly practicable for Cartesian sampling schemes (Ye et al, 1996, p.223). In simple terms, SENSE can be characteristically defined as an image domain unfolding procedure. In the k-space which has been Cartesian sampled, both the location of and the distance between the periodic reappearances in an image domain are well known. Owing to the undeniable fact that SENSE parallel imaging entails the reconstruction of mages from sensitivity encoded data, this approach addresses majority of the general cases of the integration of sensitivity encoding and gradient. This is to mean that there are hardly any restrictions made to the configuration of the coil and/or to the k-space sampling pattern. SENSE presents dual reconstruction strategies; the first being the strong reconstruction, which solely focuses on optimal voxel shape for convenience purposes. The second strategy is the weak reconstruction. This strategy lays a weaker preference on the voxel shape criterion to the signal-to-noise ratio. However, the common ground for these two strategies is the reconstruction of the algorithm, which is normally numerically demanding. So as to efficiently perform a SENSE reconstruction on a Cartesian, it is a prerequisite to first create an aliased image for each and every array element using the discrete Fourier transform (DFT) (Cohen, 1998, p.15). This is closely followed by the creation of a full-FOV image from the set on transitional images. Sensitivity encoding is dependent on the fact that each individual pixel on an image from a sole element whose array is phased is a superposition of a number of pixels which represent the tissues spin density. This denotes that each and every pixel in the aliased image is the total sum of the various signals in the space weighted by the coil sensitivity profile and the realistic spin density. It is also worth noting that SENSE demands a reference scan to be taken for purposes of ascertaining the coil sensitivity for individual elements in the phase array. The reason as to why the coil sensitivity information is a prerequisite is that the same is utilized in separating signals coming from varied spatial locations for each pixel. It can also be used for unfolding the aliasing. The notable advantage of SENSE is that there are no limitations to any special element arrangement within the phased array (Ye et al, 1996, p.224). Simultaneous acquisition of spatial harmonics (SMASH), on the other hand, is a parallel imaging technique whereby image processing is so achieved on k-space data, that is raw data. Being a new and fast-imaging technique, SMASH increases the speed of magnetic resonance image acquisition by an integer factor over the already in existence fast-imaging methods without necessarily making quantifiable sacrifices in either SNR or spatial resolution. SMASH does the reduction of the image acquisition time by taking advantage of spatial information characteristic in the geometry of a surface coil array as a substitute for some of the phase encoding, resulting from the magnetic field gradients (Amin & Afzal, 2009, p.235). This ordinarily permits for partially parallel image acquisitions via the use of the already existing fast-imaging arrangements. Unlike in SENSE, reconstructions in SMASH entail a relatively small set of magnetic resonance signal combinations prior to FT. This aspect can prove to be advantageous for both practical implementation and artefact handling (Ye et al, 1996, p.224). Although SMASH requires prior approximation of an individual coil sensitivity of the receiver array, as is the case with SENSE, SMASH is basically grounded on a linear combination of these approximated coil sensitivities which are in the offing of generating missing phase-encoding steps. Bibliography Amin, N & Afzal, M 2009, ‘The Impact of Variation in the Pulse Sequence Parameters on Image Uniformity in Magnetic Resonance Imaging, Journal of Pak Medical Association, vol. 59, no.4, pp. 231-235 Ben 2012, Magnetic Resonance Imaging, module 4 to 6 Cohen, MS 1998, Echo-Planar Imaging (EPI) and Functional MRI, PA Bandettini and C Moonen, pp.1-17 McCarthy, G, Blamire, A, Rothman, DL, Gruetter, R & Shulman, RG 1993, ‘Echo-Plana Magnetic Resonance Imaging Studies of Frontal Cortex Activation During Word Generation in Humans, Proc. Natl. Acad. Sci, vol. 90, pp.4952-4956, USA. Ye, FQ, Pekar, J, Jezzard, P, Duyn, JH, Frank, JA & McLaughlin, AC 1996, ‘Perfusion Imaging of the Human Brain at 1.5 T using a Single-Shot EPI Spin Tagging Approach, Magn. Reson. Med. Vol.36, pp.219-224. Read More