cuarres 13) | EZ Magnetic Resonance Imaging: Advanced Image Acquisition Methods, Artifacts, Spectroscopy, Quality Control, Siting, Bioeffects, and Safety ‘The essence of magnetic resonance imaging (MRI) in medicine is the acquisition, manipulation, display, and archive of datasets that have clinical relevance in the context of making a diagnosis or performing research for new applications and opportunities. There are many advantages and limitations of MRI and MR spectros- copy (MRS) as a solution to a clinical problem. Certainly, as described previously (note that this chapter assumes a working knowledge of Chapter 12 content), the great advantages of MR are the ability to generate images with outstanding tissue contrast and good resolution, without resorting to ionizing radiation. Capabilities of MR extend far beyond those basics, into fast acquisition sequences, perfusion and diffusion imaging, MR angiography (MRA), tractography, spectroscopy, and a host of other useful or potentially useful clinical applications. Major limitations of MR are also noteworthy, including extended acquisition times, MR artifacts, patient claustro- phobia, issue heating, and acoustic noise to name a few. MR safety. often ignored, 1s also of huge concern to the safety of the patient In this second of two MR chapters, advanced pulse sequences and fast image acquisition methods, dedicated radiofrequency (RF) coils, methods for perfusion diffusion, and angiography imaging, image quality metrics, common artifacts, spec- troscopy, MR equipment and siting, as well as MR safety issues are described and discussed with respect to the underlying phystes The concepts of image acquisition and timing issues for standard and advanced pulse sequences into k-space is discussed first, with several methods that can be used to reduce acquisition times and many of the trade-offs that must be considered 33] Image Acquisition Time A defining character of MRI is the tremendous range of acquisition time needed to image a patient volume. Times ranging from as low as 50 ms to tens of minutes are commonly required depending on the study, pulse sequence, number of images in the dataset, and desired image quality. When MR was initially considered to be a potential diagnostic imaging modality in the late 1970s, the prevailing conventional wisdom gave no chance for widespread applicability because of the extremely long limes required to generate a single slice from a sequentially acquired dataset, which 4ag450 Section Il ¢ Diagnostic Radiology required several minutes or more per slice. Breakthroughs in technology, equip. ment design, RF coils, the unique attributes of the k-space matrix, and methods of acquiring data drastically shortened acquisition times (or effective acquisition times) quickly, and propelled the rapid adoption of MRI in the mid-1980s, By the early 1900s, MRI established its clinical value that continues to expand today. Acquisition Time, Two-Dimensional Fourier Transform Spin Echo Imaging ‘The time to acquire an image is determined by the data needed to fill the fraction of kespace that allows the image to be reconstructed by Fourier transform methods, For a standard spin echo sequence, the relevant parameters are the TR, number of phase encoding steps, and number of excitations (NEX) used for averaging identical repeat cycles, as Acquisition time = TR X # PEG Steps x NEX Even though there may be multiple echoes as illustrated in Figure 13-1, there ts also the same number of k-space repositories to capture the data in a specific, single row of k-space defined by the strength of the PEG, as shown for the first echo with proton density weighting and second echo with T2 weighting for this double echo acquisition. Thus, effective imaging time can be reduced by producing wo (or more) images of the same slice within the TR interval. In addition, the matrix size that defines k-space is often not square (e.g., 256 X 256, 128 X 128), but 0 A —r ~ a _~ ~ cS i S ke Vv ke Vv echo, proton density weighted 2echo, T2 weighted FIGURE 13-1 Standard spin echo pulse sequence is shown with two echoes per TR interval to encode proton density contrast (short TE, first echo}, and T2 contrast (long TE, second echo). In this acquisition, two separate images are acquited independently by storing in a designated k-space matnx accorcing to echo time. ‘Assingle PEG strength is momentarily applied to induce phase variations to encode the row to be filled in each of the matrices (see the red PEG encoding for the last row in k-space, for instance). The full k-space matrix requires the sequence to be repeated with incremental variations in the PEG strength until each k-space row is, fully populated. If averaging is desired, then an identical sequence (without incrementing the PEG) is repeated and averaged in the same row.Chapter 13 * Magnetic Resonance Imaging 451 rectangular (e.g, 256 X 192, 256 X 128) where the small matrix dimension is most frequently along the phase encode direction to minimize the number of incremental PEG strength applications during the acquisition. A 256 X 192 image matrix and two averages (NEX) per phase encode step with a TR = 600 ms (for T1 weighting) requires imaging time of 0.6 s X 192 X 2 = 230.4 s = 3.84 min for a single slice! For a proton density and T2-weighted double echo sequence with TR = 2,500 ms (Fig, 13-1), this increases to 16 min, although two images are created in that time Of course, a simple first-order method would be to eliminate the number of aver- ages (NEX), which reduces the time by a factor of 2; however, the downsides are an increase in the statistical variability of the data, which decreases the image signal-to- noise ratio (SNR) and makes the image appear “noisy.” Methods to reduce acquisi- tion time and/or time per slice are crucial to making MR exam times reasonable, as described by various methods below: Multislice Data Acquisition The average acquisition time per reconstructed image slice in a single-slice spin echo sequence is clinically unacceptable. However, the average time per slice is significantly reduced using mulkislice acquisition methods, where several slices within the tissue volume are selectively excited in a sequential timing scheme during the TR interval to fully utilize the dead time waiting for longitudinal recovery in an adjacent slic as shown in Figure 13-2. This requires cycling all of the gradients and tuning the RF excitation pulse many times during the TR interval. The total number of slices that can be acquited simultaneously isa function of TR, TE, and machine limitations Total Number of Slices = TRATE + C), where C is a constant dependent on the MR equipment capabilities (computer speed, gradient capabilities, sequence options, additional pulses, e g.. spoiling pulses in standard SE; use of spatial saturation; and chemical shift, among others). Each slice and each echo, if multiecho, requires its own k-space repository to store data as its acquired. Long TR acquisitions such as proton density and T2-weighted sequences +“_______ e— hy gp #lices = TR /(TE+C) FIGURE 13-2 Multisice two-dimensional image acquisition is accomplished by discretely exciting different slabs of tissue during the TR period; appropriate changes of the RF excitation bandwidth, SSG, PEG, anc FEG parameters are necessary. Because cf diffuse excitation profiles, RF irradiation of acjacent slices leads to partial saturation and loss of contrast. The number of slices (volume) that can be obtained is a function of the TR. TE, and C, the latter representing the capabilities of the MR system and type of pulse sequence.452 Section Ii + Diagnostic Radiology can produce a greater number of slices over a given volume than T!-weighted sequences with a short TR. The chief trade-off is a loss of tissue contrast due to cross- excitation of adjacent slices due to nonsquare excitation profiles, causing undesired proton saturation as explained in Section 13.5 on artifacts. Data Synthesis Data “synthesis” takes advantage of the symmetry and redundant characteristics of the frequency domain signals in k-space. The acquisition of as little as one-half the data plus one row of k-space allows the mirroring of "complex conjugate” data to fill the remainder of the matrix (Fig. 13-3). In the phase encode direction, “half Fourier,” “Ya NEX,” or “phase conjugate symmetry” (vendor-specific names) techniques effec- tively reduce the number of required TR intervals by one-half plus one line, and thus can reduce the acquisition time by nearly one-half. In the frequency encode direc- tion, “fractional echo" or “read conjugate symmetry” refers to reading a fraction of the echo. While there is no scan time reduction when all the phase encode steps are acquired, there is a significant echo time reduction, which can reduce motion-related artifacts, such as dephasing of blood. However, the penalty for either half Fourier or fractional echo techniques is a reduction in the SNR (caused by a reduced NEX or Fractional NEX: Fractional Echo: Acquired data = % matrix + 1 line minimum TE reduced ky zz WAM ‘Synthesized mirror image data from Mirror’ k, Data ‘opposite quadrants image extracted FIGURE 13-3 Fractional NEX and Fractional Echo. Left. Data synthesis uses the redundant characteristics of the frequency domain. This is an example of phase conjugate symmetry. in which 2 of the PEG views +1 tna are acquired, and te complex conagateof he data eflectedn the symmetric quadrants. Acoust time is thus reduced by approumately Y2 (~ 40%), although image nosse 1s increased by approximately V2 Right: Fractional echo acquisition is performed when only part of the echo is read during the application of the FEG. Usually, the peak of the echo is centered in the middle of the readout gradient, and the echo signals prior to the peak are identical mirror images after the peak. With fractional echo, the echo is no longer centered, and the sampling window is shifted such that only the peak echo and the dephasing part of the echo are sampled. As the peak of the echo is closer to the RF excitation pulse, TE can be reduced, which can improve TI and proton density weighting contrast. A larger number of slices can also be obtained with a shorter TE in a multislice acquisition (see Fig. 13-2)Chapter 13 * Magnetic Resonance Imaging 453, data sampling in the volume) and the potential for artifacts if the approximations in the complex conjugation of the signals are not accurate. Other inaccuracies result from inhomogeneities of the magnetic field, imperfect linear gradient fields, and the presence of magnetic susceptibility agents in the volume being imaged. Fast Pulse Sequences Fast Spin Echo (FSE) techniques use multiple PEG steps in conjunction with mul- Liple 180-degree refocusing RF pulses to produce an echo train length (ETL) with corresponding digital data acquisitions per TR interval, as illustrated in Figure 13-4 Multiple k-space rows are filled during each TR equal to the ETL, which is also the reduction factor for acquisition time. “Effective echo time” is determined when the central views in k-space are acquited, which are usually the first echoes, and subsequent echoes are usually spaced apart via increased PEG strength with the same echo spacing time. “Phase re-ordering” optimizes SNR by acquiring the low- frequency information with the early echoes (lowest amount of T2 decay), and the high-frequency, peripheral information with late echoes, where the impact on over- all image SNR is lower. The FSE technique has the advantage of spin echo image acquisition, namely immunity from external magnetic field inhomogeneities, with 4X, 8X, 10 16% faster acquisition time. However, each echo experiences difler- ent amounts of intrinsic T2 decay, which results in image contrast differences when compared with conventional spin echo images of similar TR and TE. Lower sig- nal levels in the later echoes produce less SNR, and fewer images can be acquired in the image volume during the same acquisition. A T2-weighted spin echo image CTR = 2,000 ms, 256 phase encode steps, one average) requiresapproximately 8.5 min, while a corresponding FSE with an ETL of 4 (Fig. 13-4) requires about 2.1 min, PEG __ FES te Echo a a Bteaive Echo train length (ETL) = 4 Effective TE = 16 ms FIGURE 13-4 Conventional FSE uses multiple 180-degree refocusing RF pulses per TR interval with incre- mental changes in the PEG to fil several views in k-space (the ETL). This example illustrates an ETL of four. ‘with an “effective” TE equal to 16 ms. Total time of the acquisition 1s reduced by the ET factor. The reversed polarity PEG steps reestablish coherent phase before the next gradient application. Slightly different PEG strengths are applied to fill the center of k-space first, and then the periphery with later echoes, continuing Until all views are recorded. As shown, data can be mirrored using conjugate symmetry to reduce the overall time by another factor of two,asa Section II » Diagnostic Radiology Longer TR values allow for a greater ETL, which will offset the longer TR in terms of overall acquisition time, and will also allow more proton density weighting Specific FSE sequences for T2 weighting and multiecho FSE are employed with variations in phase reordering and data acquisition. FSE is also known as “turbo spin echo” or “RARE” (rapid acquisition with refocused echoes) A Gradient Echo (GE) Acquisition pulse sequence is similar to a standard spin echo sequence with a readout gradient reversal substituting for the 180-degree pulse (Fig. 13-5). Repetition of the acquisition sequence occurs for each PEG step and with each average. With small f_ip angles and gradient reversals, a consider- able reduction in TR and TE is possible for fast image acquisition: however, the ability to acquire multiple slices is compromised. A PEG rewinder pulse of oppo- site polarity is applied to maintain phase relationships from pulse to pulse in the coherent image acquisition. Spoiler gradients are used to eliminate persistent trans- verse magnetization from stimulated echoes for incoherent GE (see Chapter 12, Section 12.5) Acquisition times are calculated in the same way as spin echo; a GE sequence for a 256 X 192 image matrix, two averages, and a TR = 30 ms, results in an imaging, time equal to 192 X 2 X 0.03s = 15.5 A conventional spin echo requires 3.84 min for a TR = 600 ms. Trade-offs for fast acquisition speed include SNR losses, magnetic susceptibility anifacts, and less immunity from magnetic field inhomo- geneities. There are several acronyms for GE sequences, including GRASS, FISP, Spoiled GRASS, FLASH, SSFP. etc., depending on the manufacturer of the equip- ment. Table 13-1 describes a partial list of the different GE sequences and their method of data acquisition. Echo Planar Image (EPI) Acquisition is a technique that provides extremely fast imaging time. Spin Echo (SE-EP!) and Gradient Echo (GE-EPI) are two methods used for acquiring data, and a third is a hybrid of the two, GRASE (Gradient and Spin Echo). Single-shot (all of the image information 1s acquired within 1 TR imerval) or multishot EPI has been implemented with these methods. For single-shot SE-EPI, en JES IB FIGURE 13-5 Coherent GE pulse sequence uses a smal fip angle (30 to 40 degrees) RF pulse simultaneous to the SSG. Phase and frequency encode gradients are applied shortly afterward (with a TE of less than 3 ms in cert sequences). A PEG “rewinder” (reverse polanty) reestablishes the phase conditions prior to the next pulse, smmultaneous with the extended FEG durationChapter 13 * Magnetic Resonance imaging 455, TABLE 13-1 COMPARISON OF MANUFACTURER-NAMED ACRONYMS FOR GE SEQUENCES SEQUENCE GENERAL ELECTRIC PHILIPS SIEMENS TOSHIBA Coherent GE GRASS, FGRFMPGR FFE FISP. Field echo Incoherent GE SPGR, FSPGR THFFE Field echo (RF spoiled) Incoherent GE ‘MPGR FLASH Field echo (Gradient spoiled) Steady-state free SSFP, DE FGR T2FFE PSI precession ‘SSFP: balanced FIESTA Balanced FFE True FISP. True SSFP sequence / true FiSP Note: Not ail manufacturers ae listed in this table, ner are all GE sequences. (Blank areas indicate particular ‘sequence is not performed (at time of publication) Image acquisition typically begins with a standard 90-degree flip, then a PEG/FEG gradient application to initiate the acquisition of data in the periphery of the k-space, followed by a 180-degree echo-producing RF pulse. Immediately after, an oscillating, readout gradient and phase encode gradient “blips” are continuously applied to stimu- late echo formation and rapidly fill k-space in a stepped “zig-zag” pattern (Fig, 13-6) The “effective” echo time occurs at a time TE, when the maximum amplitude of the induced GEs occurs. Acquisition of the data must proceed in a period less than 12* (around 50 ms), placing high demands on the sampling rate, the gradient coils (shielded coils are required, with low induced “eddy currents"), the RF transmitter/ receiver, and RF energy deposition limitations, For GE-EPI, a similar acquisition strat- egy is implemented but without a 180 degrees refocusing RF pulse, allowing for faster acquisition time. SE-EPI is generally longer, but better image quality is achieved; on the other hand, larger RF energy deposition to the patient occurs. EPI acquisition can FIGURE 13-6 Single shot Echo Planar Spin Echo image (SE-EP) ‘axcquisition sequence. Data is deposited in k=pace, initially posi tioned by a simultaneous PEG and FEG application to locate the initial row and column position (in this ‘example, the upper left), followed by phase encode gradient “blips” simultaneous to FEG oscillations, to fill k-space line by line by intro- ‘ducing 1-row phase changes in ‘a ng-2ag pattern. Image matrix sizes of 64 X 64 and 128 X 64 are common.Section Il * Diagnostic Radiology be preceded with any type of RF pulse, for instance FLAIR (EPI-FLAIR), which will prodiice images much faster than the corresponding conventional FLAIR sequence The GRASE (Gradient and Spin Echo) sequence combines the initial spin echo with a series of GEs, followed by an RF rephasing (180 degrees) pulse, and the pat- tem is repeated until k-space is filled. This hybrid sequence achieves the benefits of both types of rephasing: the speed of the gradient and the ability of the RF pulse to compensate for T2* effects, providing significant improvements in image quality compared to the standard EP! methods. A trade-off isa longer acquisition ume (¢.g , greater than 100 ms) and much greater energy deposition from the multiple 180 degrees RF pulses. EPI acquisitions typically have poor SNR, low resolution (matrices of 6+ X 64 or 128 X 64 are typical), and many artifacts, particularly of chemical shift and magnetic susceptibility origin, Nevertheless, EPI offers real-time “snapshot” image capability with 50 ms total acquisition time. EPI is emerging as a clinical tool for studying time- dependent physiologic processes and [unctional imaging, Concems of safety with EPI, chiefly related to the rapid switching of gradients and possible nerve stimulation of the patient, the associated acoustic noise, image artifacts, distortion, and chemical shift are components that will limit use for many imaging procedures Other K-Space Filling Methods Methods to fill k-space in a nonsequential way can enhance signal, contrast, and achieve rapid scan times as shown in Figure 13-7. Centric k-space filling has been discussed with FSE imaging (above), where the lower strength phase encode gradi- Outer rows tilled last A Centric filing Central rows filed first Outer rows filed last Outer rows filed first B Keyhole hiting Contra rows ies nth Outer rows filed first Equal AT between points k-space re-binning of spiral data is required before image reconstruction © Spiral fing FIGURE 13-7 Alternate methods of filing k-space. A. Centnc filing applies the lower strength PEG's frst to ‘maximize signal and contast from the eartest echoes ofa FSE or GE sequence. B. Keyhole filing applies PEG's ‘of higher strength first to fil the outer portions of k-space, and the central lines ae filed only during a certain part of the sequence, such as with arrnal of contrast signal. C Spiral data acquisition occurs with sinusovdal ostilation of the X and ¥ gradients 90 degrees out of phase with each other. with samples beginning in the center of kspace and spiraling out to the periphery. interpolating the data into the k,,k, matrix is required in order to apply 2DFT image reconstructionChapter 13 ¢ Magnetic Resonance Imaging 457 ents are applied first filling the center of k-space when the echoes have their highest amplitude. This type of filling is also important for fast GE techniques, where the image contrast and the SNR fall quickly with time from the initial excitation pulse. Keyhole filling methods fil k-space similarly to centric filling, except the central lines are filled when important events occur during the sequence, in situations such. as contrast-enhanced angiography. Outer areas of k-space are filled first, and when gadolinium appears in the imaging volume, the center areas are filled. At the end of the scan, the outer and central k-space regions are meshed to produce an image with both good contrast and resolution Spiral filling is an alternate method of filling k-space radially, which involves the simultaneous oscillation of equivalent encoding gradients to sample data points during echo formation in a spiral, starting at the ongin (the center of the k-space) and spiraling outward to the periphery in the prescribed acquisition plane. The same contrast mechanisms are available in spiral sequences (e.g., TI, T2, proton density weighting). and spin or GEs can be obtained. After acquisition of the signals, an addi- onal post-processing step, re-gridding, is necessary to convert the spiral data into the rectilinear matrix for two-dimensional Fourier transform (2DFT), Spiral scanning is an efficient method for acquiring data and sampling information in the center of k-space, where the bulk of image information is contained A variant of radial sampling with enhanced filling of the center of k-space is known generically as “blade” imaging, and commonly as propeller: Periodically Rotated Overlapping Parallel Lines with Enhanced Reconstruction, where a rectangular block of data is acquired and then rotated about the center of k-space. Redundant informa- tion concentrated in the center of k-space is used for improvement of SNR or for the identification of times during the scan in which the patient may have moved, so that those blocks of data can be processed with a phase-shifting algorithm to eliminate the movernent effect on the data during the reconstruction process and to mitigate motion artifacts toa great extent. Filling of k-space for this method is shown in Figure 13-8, Parallel Imaging Parallel imaging is a technique that fills k-space by using the response of multiple receive RF coils that are coupled together with independent channels, so that data can be acquired simultaneously Specific hardware and software are necessary for the electronic orchestration of this capability. Typically, 2, 4, 5, 7, 8, 16, 18 (or more) colls ate arranged around the area to be imaged; if a 4-coil configuration is used, then during each TR period, each coil acquires a view of the data as the acquisition sequence proceeds. Lines in k-space are defined only after the processing of linear combinations of the signals that are received by all of the coils. Since 4 views of the data are acquired per TR interval, scan time can be decreased by a factor of 4 (known as the reduction factor). However, the acquisition of the signals have gaps, and the FOV in the phase direction is reduced to one-quarter of its original size This results in a known aliasing of the information (a wrapped image—see section on Artifacts) that i rectified by using the measured sensitivity profile of each coil to calculate from where the signal is coming, This sensitivity profile determines the position of the sig- nal based on ts amplitude, where the signal near the coil has a higher amplitude than that farthest away: As a result of the process, commonly known as SENSE (SENSiti ity Hncoding—thete are several acronyms coined by the manufacturers), the image can be unwrapped and combined with the unwrapped images from each of the othe, coils. A simple two-coil example is shown in Figure 13-9 for a breast image applica- tion of SENSE, in which improved resolution is desired over reduced sean tine458 Section II + Diagnostic Radiology Rectangular filing Propeller filling Motion during the acquisition FIGURE 13-8 The propeller data acquistion compared to a rectangular filing of k-space is shown above Instead af acquiring single lines of information to fill kspace consecutively as shown in the upper left and middle left, a rectangular data acquisition at a specific angle (e.g., 0 degree) acquired encompassing several lines of k-space, which represents a “blade” of mformation. The partial acquisiton is rotated about the center of k-space at angular increments, which provides 3 dense sampling of data at the center of k-space and less in the periphery as shown by the schematic (upper right illustration). I the patient moves during a portion of the ‘examination (lower left image). the blades in which the motion occurred can be identified, reprocessed, and the image reconstructed without the motion artifact lower right mage) Single coil FIGURE 13-9 Parallel imaging with two RE ois. Top. A single coil acquisition of a breast (MR exam over the full FOV. Middle. indwvidual cols with every-other row of k-space being filed represent Ys FOV, with image overlap caused by aliasing. Bottom. After SENSE pro- cessing, images are combined to deliver twice the spatial resolution in the left/nght (Phase) direction, with the same imaging time Right coilChapter 13 * Magn ic Resonance Imaging 459 Parallel imaging can be used to either reduce scan times or improve resolution. It also can be used with most pulse sequences. There are obvious benefits in terms of scan times and/or resolution, but there isa slight loss of SNR due to the manipulation of the signals, and chemical shift artifacts (explained in the Anifacts section) may increase. Patient motion can also cause misalignment between the undersampled data and the reference scans of the coils. Three-Dimensional Fourier Transform Image Acquisition Three-dimensional image acquisition (volume imaging) requires the use of a broad- band, nonselective, or “slab-selective” RF pulse to excite a large volume of pro- tons simultaneously. Two phase gradients are discretely applied in the slice encode and phase encode directions, prior to the frequency encode (readout) gradient (Fig, 13-10). The image acquisition time is equal to ‘TR X # Phase Encode Steps (z-axis) X # Phase Encode Steps (y-axis) X # Signal Averages A three-dimensional Founer transform (three one-dimensional Fourier transforms) is applied for each column, row, and depth axis in the image matrix “cube.” Volumes obtained can be either isotropic, the same size in all three ditections, oF anisotropic, where at least one dimension ts different in size. The advantage of the former is equal resolution in all directions; reformations of images from the volume do not suller from degradations of larger sample size from other directions. After the spatial domain data are obtained, individual two-dimensional slices in any arbitrary plane are extracted by interpolation of the cube data When using a standard TR of 600 ms with one average for a T1-weighted exam, a 128 X 128 X 128 cube requires 163 min or about 2.7 h! Obviously, this is unaccept- able for standard clinical imaging. GE pulse sequences with TR of 50 ms acquire the same image volume in about 15 min. Another shortcut is with anisotropic voxels, where the phase encode steps in one dimension are reduced, albeit with a loss of resolution A major benefit to isotropic three-dimensional acquisition is the uniform resolution Isotropic Slice Encode (phase #1) e ‘pres #2) Frequency Encode PM Enas® FIGURE 13-10 Three-dimensional image acquisition requires the application of a broadband RF pulse to excite all of the protons in the volume simultaneousy, followed by a phase encode gradient along the sice encode direction, a phase encode gradient along the phase encode direction, and a frequency encode gra: dient in the readout direction. Spatial location ss decoded sequentially by the Fourier transtorm along each encode path, storing intermechate results inthe three-dimensional k-space matrix.460 Section Il * Diagnostic Radiology in all directions when extractin, : ig any two-dimensional image from the ma tition, high SNRs achieved compared to asm twccimenstnal mage allowing feconsirutin of very thin slices with good detail les paral volume averaging) and ig A downside is the increased probability of motion artifacts and computer hardware requirements for data andl, and storage nl 132| MR Image Characteristics Spatial Resolution and Contrast Sensitivity Spatial resolution, contrast sensitivity, and SNR parameters form the basis for evalu- aling the MR image characteristics. The spatial resolution is dependent on the FOV, which determines pixel size, the gradient field strength, which determines the FOV. the receiver coil characteristics (head coil, body coil, and various surface coil designs), the sampling bandwidth, and the image matrix. Common image matrix sizes are 128 128, 256 X 128, 256 X 192, and 256 x 256, with 512 x 256, 512 x 512, and 1,024 x 512 becoming prevalent. In general, MR provides spatial resolution approx- imately equivalent to that of CT, with pixel dimensions on the order of 0.5 to 1.0 mm for a high-contrast object and a reasonably large FOV (greater than 250 mm) ‘4.250 mm FOV and a 256 X 256 matrix will have a pixel size on the order of 1 mm. In small FOV acquisitions with high gradient strengths and with surface coil receivers, the effective pixel size can be smaller than 0.1 to 0.2 mm (of course, with 4 limited FOV of 25 to 50 mm. Slice thickness in MRI is usually 5 to 10 mm and 11 that produces the most partial volume averaging, higher field strength magnets due toa larger SNR. which allows thinner slice acquisition, and/or higher sampling rates (smaller pix- cls) for a given acquisition, However, with higher B,. increased RF absorption, artifact production, and a lengthening of TI relaxation occur. The later decreases TI contrast Sensitivity because of increased saturation of the longitudinal magnetization ‘Contrast sensitivity is the major attribute of MR. The spectacular contrast sensitiv” ity of MR enables the exquisite discrimination of soft issues and contrast due to blood few This sensitivity is achieved through dillerences inthe T!, T2, proton density and flow velocity characteristics. Contrast, which is dependent upon these parameter, sueved through the proper application of pulse sequences, as discussed previously: SiR contrast materials, usually suscepubihty agents that disrupt the local magnetic field to enhance T2 decay or provide a relaxation mechanism [or shorter TI recovery time (eg. bound water in hydration layers), are becoming import enhancement agents fonnlifferentiation of normal and diseased tissues. The absolute contrat sensitivity of the MR image is ultimately limited by the SNR and presence of image artifacts represents the dimensiot ‘Spatial resolution can be improved with Signal-to-Noise Ratio, SNR ultimate SNR achievable by the MR system numerous dependencies on the Te bibede signal tmensity based on TH, T2, and proton density parameters has been discussed; to summarize, the TR, TE, and flip angle will have an impact on the magnitude of the signal generated in the image While there are many mitigaing fears, a long TR increases the longitudinal magnelzst 0 recovery aa ten the SNR. a Jong TE increases the transverse magne a0 decay and reduces th SNR: a smaller flip angle (reduced from 90 degrees) reduces the SN . ‘ ith large flip angle, long TR, short TE, coarse mans spin echo pulse sequences WChapter 13 * Magnetic Resonance Imaging 461 large FOV, thick slices, and many averages will generate the best SNR: however, the resultant image may not be clinically relevant or desirable. While SNR is important, its not everything For a given pulse sequence (TR, TE, flip angle), the SNR of the MR image is dependent on a number of variables, as shown in the equation below for a (wor dimensional image acquisition: ANEX SNR 1>¢voxel,,. XSF X (QE) X (B)X face gap) x freconstruction) where Is the intrinsic signal intensity based on pulse sequence, voxel... 1s the voxel volume, determined by FOV, image matrix, and slice thickness, NEX is the number of excitations, determined by the number (or fractional number) of repeated signal acquisitions into the same voxels, BW is the frequency bandwidth of the RE receiver, {QP is the function of the coil quality factor parameter (tuning the coil), (B) is the function of magnetic field strength, B, {(slice gap) is the function of interslice gap effects, and {(reconstruction) is the function of the reconstruction algorithm Other factors in the above equation are explained briefly below: Voxel Volume The voxel volume is equal to FOV, FOV, — FOV, _,__FOM, _. stice thickness, No. of pixels, x \ No.of pixels, y © t Uuckness. 2 Volume SNRs linearly proportional to the voxel volume. Thus, by reducing the image matrix size from 256 X 256 to 256 X 128 over the same FOV, the effective voxel size increases by a factor of wo, and therefore increases the SNR by a factor of two for the same image acquisition time (e.g., 256 phase encodes with one average versus 128 phase encodes with two averages). Signal Averages Signal averaging (also known as number of excitations, NEX) is achieved by averag- ing sets of data acquired using an identical pulse sequence (same PEG strength). The SNR is proportional to the square root of the number of signal averages. A 2-NEX acquisition requires a doubling (100% increase) of the acquisition time for a 40% imerease in the SNR (V2 =1.4) Doubling the SNR requires 4 NEX. In some cases, less than | average (e.g,, Y4 or %4 NEX) can be selected. Here, the number of phase encode steps is reduced by % or %, and the missing data are synthesized in the k-space matrix. maging time is therefore reduced by a similar amount: however, a loss of SNR accompanies the shorter imaging times by the same square root factor. RF Bandwidth The receiver bandwidth defines the range of frequencies to which the detector is tuned during the application of the readout gradient. A narrow bandwidth (a nar- row spread of frequencies around the center frequency) provides a higher SNR, 1 proportional to YBW . A twofold reduction in RF bandwidth—from 8 to 4 kHz,462 Section Il * Diagnostic Radiology for instance—increases the SNR by 1.4 X (40% increase). This is mainly related to the fact that the white noise, which is relatively constant across the bandwidth, does not change, while the signal distribution changes with bandwidth. In the spa- tial domain, bandwidth is inversely proportional to the sample dwell ume, AT to sample the signal: BW = 1/AT. Therefore, a narrow bandwidth has a longer dwell time, which incteases the signal height (Fig. 13-11), compared to the shorter dwell lime for the broad bandwidth signal, thus spreading the signal over a larger range of frequencies. The SNR is reduced by the square root of the dwell time. However, any decrease in RF bandwidth must be coupled with a decrease in gradient strength to maintain the sampling across the FOV, which might be unacceptable if chemt- cal shift artifacis are of concern (see Anifacts, below). Narrower bandwidths also require a longer time for sampling, and therefore affect the minimum TE time that is possible for an imaging sequence. Clinical situations that can use narrow band- widths are with T2-weighted images and long TEs that allow the echo to evolve over an extended period, particularly in situations where fat saturation pulses are used to reduce the effects of chemical shift in the acquired images. Use of broad bandwidth settings is necessary when very short TEs are required, such as in fast GE imaging 10 reduce the sampling time RF Coil Quality Factor The coil quality factor is an indication of RF coil sensitivity to induced currents in response to signals emanating from the patient. Coil losses that lead to lower SNR are caused by patient “loading” effects and eddy currents, among other factor. Patient load- ing refers o the electric impedance characteristics of the body, which to a certain extent acts like an antenna, This effect causes a variation in the magnetic field that is different ‘Signal broad (16 kHz) oe “I \\ “aT \ (KHz) ——~Il\v - sal | narrow (4 kHz) RF bandwidth = 1/ dwell time = 1/aT ‘Sample dwell time MEFIGURE 13-11 RF Receiver Bandwidth is determined by the FEG strength, the FOV, and sampling rate. This figure illustrates the spatial domam ew of SNR and corresponding sample dwell time. Evolution of the echo in the broad bandwidth situation occurs rapidly with minimal dwell time, which might be needed in situations where very short TE is required, even though the SNR is reduced. On the other hand, in T2 weighted images requiring a long TE, narrow bandwidth can improve SNRChapter 13 ¢ Magnetic Resonance Imaging 463 for each patient, and must be measured and corrected for. Consequently, tuning the receiver coil to the resonance frequency is mandatory before image acquisition. Eddy currents ate signals that are opposite of the induced current produced by transverse ‘magnetization in the RF coil, and reduce the overall signal. Quadrature coils increase the SNR as two coils are used in the reception of the signal; phased array coils increase the SNR even more when the data from several coils are added together (see Paralle! Imag- ing, Section 13.1). The proximity of the receiver coil to the volume of interest affects the coll quality factor, but there are trade-offs with image uniformity, Positioning of the coil with respect to the direction of the main magnetic field is also an issue that occurs with air core (horizontal B.) to solid core (vertical B,) magnets. Body receiver coils positioned in the bore of the magnet have a moderate quality factor, wheteas surface coils have a high quality factor. With the body coil, the signal is relatively uniform across the FOV; however, with surface coils, the signal falls off abruptly near the edges of the field, lmit- ing the useful imaging depth and resulting in nonuniform brightness across the image Magnetic Field Strength Magnetic field strength influences the SNR of the image by a factor of BI" to B!* Thus, one would expect a three- to fivefold improvement in SNR with a 1.5 T magnet over 0.5 T magnet, Although the gains in the SNR are real, other considerations mitigate the SNR improvement in the clinical environment, including longer TL relaxation times and greater RF absorption, as discussed previously: Cross-Excitation Cross-excitation occurs from the nonrectangular RF excitation profiles in the spatial domain and the resultant overlap of adjacent slices in muhtislice image acquisition sequences. This saturates the protons and reduces contrast and the contrast-to-noise ratio. To avoid cross-excitation, interslice gaps or interleaving procedures are neces- sary (see Artifacts section, below) Image Acquisition and Reconstruction Algorithms Image acquisition and reconstruction algorithms have a profound effect on SNR. The various acquisition/reconstruction methods that have been used in the past and those used today are, in order of increasing SNR, point acquisition methods, line acquisition methods, two-dimensional Fourier transform acquisition methods, and three-dimen- sional Fourier transform volume acquisition methods. In each of these techniques, the volume of tissue that is excited is the major contributing factor to improving the SNR and image quality. Reconstruction filters and image processing algorithms will also affect the SNR. High-pass filtration methods that increase edge definition will ‘generally decrease the SNR, while low-pass filtration methods that smooth the image data will generally increase the SNR at the cost of reduced resolution Summary, Image Quality The best possible image quality is always desirable, but not always achievable because of the trade-off between SNR, scan speed, and spatial resolution. To increase one of these three components of image quality involves the consideration of reducing one or both of the other two. Itis thus a balancing act that is chosen by the operator, the protocol, and the patient in order to acquire images with the best diagnostic yield464 Section II + Diagnostic Radiology MR parameters that may be changed include TR, TE, TI, ETL, Matrix Size, Slice Thickness, Field of view, and NEX. Working with these parameters in the optimiza- tion of acquisition protocols to achieve high image quality is essential B Signal from Flow The appearance of moving fluid (vascular and cerebrospinal fluid [CSF]) in MR images is complicated by many factors, including flow velocity, vessel orientation, laminar versus turbulent flow patterns, pulse sequences, and image acquisition modes. Flow- related mechanisms combine with image acquisition parameters to alter contrast. Signal due to flow covers the entire gray scale of MR signal intensities, from “black blood’ to “bright blood’ levels, and flow can be a source of artifacts, The signal from flow can also be exploited to produce MR angiographic images Low signal intensities (low voids) are often a result of high-velocity signal loss (HSL), in which protons in the flowing blood move out of the slice during echo ref- ormation, causing a lower signal. Flow turbulence can also cause flow voids, by causing a dephasing of protons in the blood with a resulting loss of the tissue magetization in the area of turbulence. With HVSL, the amount of signal loss depends on the velocity of the moving fluid. Pulse sequences to produce “black blood’ in images can be very useful in cardiac and vascular imaging. A typical black blood pulse sequence uses a “double inversion recovery” method, whereby a nonselective 180-degree RF pulse 1s initially applied, inverting all protons in the body, and is followed by a selective 180-degree RF pulse that restores the magnetization in the selected slice. During the inversion time, blood outside of the excited slice with inverted protons flows into the slice, producing no signal; therefore, the blood appears dark Flow-Related Enhancement Flow-related enhancement is a process that causes increased signal intensity due to lowing protons; it occurs during imaging of a volume of tissues. Even-echo rephas- ing is a phenomenon that causes flow to exhibit increased signal on even echoes in a multiple-echo image acquisition. Flowing protons that experience two subsequent 180-degree pulses (even echoes) generate higher signal intensity due to a construc- live rephasing of protons during echo formation. This effect is prominent in slow Jaminar flow (e.g., veins show up as bright structures on even-echo images), Flow enhancement in GE images is pronounced for both venous and arterial structures, as well as CSF The high intensity is caused by the wash-in (between subsequent RF excitations) of fully unsaturated protons into a volume of partially saturated protons due to the short TR used with gradient imaging. During the next excitation, the signal amplitude resulting from the moving unsaturated protons is, about 10 times greater than that of the nonmoving saturated protons. With GE tech- niques, the degree of enhancement depends on the velocity of the blood, the slice or volume thickness, and the TR. As blood velocity increases, unsaturated blood exhibits the greatest signal, Similarly, a thinner slice or decreased repetition time results in higher flow enhancement. In arterial imaging of high-velocity flow, it is possible to have bright blood throughout the imaging volume of a three-dimensional acquisition if unsaturated blood can penetrate into the volume prior to experiencing an RF pulse, Signal from blood is dependent on the relative saturation of the surrounding tissues and the incoming blood flow in the vasculature. In a multislice volume,Chapter 13 + Magnetic Resonance Imaging 465 FIGURE 13-12 The repeated RF excitation within an imaging volume produces partial saturation of the tissue magnetization (top figure, gray area) Unsaturated protons flowing into the volume gener- ate a large signal cifference that 's bright relate to the surrounding tissues. Bright blood effects can be reduced by applying pre-saturation RF pulses adjacent to the imaging volume, so that protons in infiowing blood will have a similar partial saturation (bottom figure; note ne blood signal) Flow-Related Enhancement Pre-saturated spins: equal signal Flow presaturation repeated excitation of the tissues and blood causes a pantial saturation of the protons, dependent on the T1 characteristics and the TR of the pulse sequence. Blood out- side of the imaged volume does not interact with the RF excitations, and therefore these unsaturated protons may enter the imaged volume and produce a large signal compared to the blood within the volume, This is known as llow-related enhance- ment. As the pulse sequence continues, the unsaturated blood becomes partially saturated and the protons of the blood produce a similar signal to the tissues in the inner slices of the volume (Fig. 13-12). In some situations, flow-related enhance ment is undesirable and is eliminated with the use of “presaturation” pulses applied to volumes just above and below the imaging volume. These same saturation pulses are also helpful in reducing motion artifacts caused by adjacent tissues outside the imaging volume MR Angiography Exploitation of blood flow enhancement isthe basis for MRA, Two techniques to create images of vascular anatomy include time-of-flight and phase contrast angiography Time-of-Flight Angiography The time-of-flight technique relies on the tagging of blood in one region of the body and detecting it in another. This differentiates moving blood from the surround station- ary tissues. Tagging is accomplished by proton saturation, inversion, of relaxation to change the longitudinal magnetization of moving blood. The penetration of the tagged blood into a volume depends on the T1, velocity, and direction af the blood. Since the detectable range is limited by the eventual saturation of the tagged blood, long vessels are difficult to visualize simultaneously in a three-dimensional volume. For these rea- sons, a two-dimensional stack of slices is typically acquired, where even slowly mov- ing blood can penetrate the region of RF excitation in thin slices (Fig. 13-13). Each slice is acquired separately, and blood moving in one direction (north or south, €.. aneries versus veins) can be selected by delivering a presaturation pulse on an adja- cent slab superior or inferior to the slab of data acquisition. Thin slices are also help- ful in preserving resolution of the flow pattern. Often used for the two-dimensional466 Section Il » Diagnostic Radiology FIGURE 13-13 The time of fight MRA acquisition collects each slice separately with a sequence to enhance biood flow. Explcitation of blood flow is achieved by detecting unsaturated protons moving into the volume, producing a bright signal. A coherent GE image acquisition pulse sequence 1s shown, TR = 24 ms, TE = 3.1 ims, Flip Angle = 20 degrees. Every 10th image in the stack 1s displayed above, from left to right and top to bottom, image acquisition is a “GRASS” or “FISP” GE technique that produces relatively poor anatomic contrast, yet provides a high-contrast “bright blood” signal. Magnetization transfer contrast sequences (see below) are also employed to increase the contrast of the signals due to blood by reducing the background anatomic contrast Two-dimensional TOF MRA images are obtained by projecting the content of the stack of slices at a specific angle through the volume. A maximum intensity projec- tion (MIP) algorithm detects the largest signal along a given ray through the volume and places this value in the image (Fig. 13-14). The superimposition of residual MLFIGURE 13-14 simple illustration shows Projections are cast through the image stack (volume) how the MIP algorithm extracts the highest ‘The maximum signal along each line is projected (maximum) signals nthe two-dimensional stack of images along a specific direction in the volume, and produces projection images with maximum intensity variations as a func MIP images tion of angleChapter 13 * Magnetic Resonance Imaging 467 stationary anatomy often requires further data manipulation to suppress undesirable signals. This is achieved in a vanety of ways, the simplest of which is setting a win- dow threshold. Another method is to acquire a dataset without contrast, and subtract the noncontrast MIP from the contrast MIP to reduce background signals. Clini- cal MRA images show the three-dimensional characteristics of the vascular anatomy from several angles around the volume stack (Fig. 13-15) with some residual signals from the stationary anatomy Time-ol-flight angiography often produces variation in vessel intensity dependent on orientation with respect to the image plane, a situation that is less than optimal Phase Contrast Angiography Phase contrast imaging relies on the phase change that occurs in moving protons such as bload. One method of inducing a phase change is dependent on the applica tion of a bipolar gradient (one gradient with positive polarity followed by a second gradient with negative polarity, separated by a delay time AT). In a second acquisi- tion of the same view of the data (same PEG), the polarity of the bipolar gradients is reversed, and moving protons are encoded with negative phase, while the stationary protons exhibit no phase change (Fig. 13-16). Subtracting the second excitation from the first cancels the magnetization due to stationary protons but enhances magne- tization due to moving protons. Alternating the bipolar gradient polarity for each subsequent excitation during the acquisition provides phase contrast image informa- tion. The degree of phase shift is directly related to the velocity encoding (VENC) ime, AT, between the positive and negative lobes of the bipolar gradients and the velocity of the protons within the excited volume. Proper selection of the VENC time 2D Projection ‘Angiograms from MIP FIGURE 13-15 A volume stack of bright blood images (left) is used with MIP processing to create a series of projection angiograms at regular intervals, the three-dimensional perspective is appreciated in a stack wew, with virtual rotation of the vasculature468 Section Il + Diagnostic Radiology bipolar gradient encoding T T Excitation <1 Excitation | I-—1 | a #2 ‘ Excitation #1 - Excitation #2 = netresidual phase - ( Vd secre net phase shift = 0 Moving spins, low velocity, forward direction: | le. 7 __talphaee siti amet ~ = Backward direction, high 7 | eccrmeemne | ene: axpoene IEFIGURE 13-16 Phase Contrast Angiography uses consecutive excitations that have a bipolar gradient ‘encoding withthe polarity reversed between the fist and second excitation, as shown in the top row. Mag netization vectors (lower two rows) illustrate the effect of the bppolar gracients an stationary and moving spins forthe frst and second excitations. Subtracting the two will cancel stationary tissue magnetization and enhance phase differences caused by the velocity of moving blood. is necessary to avoid phase wrap error (exceeding 180-degree phase change) and to ensure an optimal phase shift range for the velocities encountered. Intensity varia- tions are dependent on the amount of phase shift, where the brightest pixel values represent the largest forward (or backward) velocity, a mid-scale value represents 0 velocity, and the darkest pixel values represent the largest backward (or forward) velocity Figure 13-17 shows a representative magnitude and phase contrast image of the cardiac vasculature. Unlike the time-of-flight methods, the phase contrast image Magnitude Image Phase Image EFIGURE 12-17 Magnitude (left) and phase (ight) images provide contrast of flowing blood. Magnitude images are sensitive to flow, but not to direction, phase images provide cirection and veloaty information ‘The blood flow from the heart shows forward flow in the ascending aorta (dark atea) and forward flow in the descending aorta at this point in the heart cycle forthe phase image. Some bright ow pattems inthe ascend ing aorta represent backward flow to the coronary arteries. Grayscale amplitude is propartional to velocity, where intermediate grayscale s 0 velocity.Chapter 13 * Magnetic Resonance Imaging 469. is inherently quantitative, and when calibrated carefully, provides an estimate of the mean blood flow velocity and direction, Two- and three-dimensional phase contrast image acquisitions for MRA are possible Gradient Moment Nulling In spin echo and gradient recalled echo imaging, the slice select and readout gradients are balanced, so that the uniform dephasing with the initial gradient application is rephased by an opposite polarity gradient of equal area, However, when moving protons are sub- jected to the gradients, the amount of phase dispersion is not compensated (Fig, 13-18). This phase dispersal can cause ghosting (laint, displaced copies of the anatomy) in images. Its possible, however, to rephase the protons by a gradient moment nulling technique. With constant velocity flow (first-order motion), all protons can be rephased using the application of a gradient tnplet. In this technique, an initial positive gradient of unit area is followed by a negative gradient of twice the area, which creates phase changes that are compensated by a third positive gradient of unit area. The velocity compensated gradi- cent (right graph in Fig. 13-18) depicts the evolution of the proton phase back to zero for both stationary and moving protons. Note that the overall applied gradient has a net area of zero—equal to the sum of the positive and negative areas. Higher order cortec- tions such as second- of third-order moment nulling to correct for acceleration and other ‘motions are possible, but these techniques require more complicated gradient switching, Gradient moment nulling can be applied to both the slice select and readout gradients to correct for problems such as motion ghosting as elicited by CSF pulsatile flow: -— Perfusion and Diffusion Contrast Imaging Perfusion of tissues via the capillary bed permits the delivery of oxygen and nutrients to the cells and removal of waste (¢.g,, CO,) from the cells. Conventional perfusion measurements are based on the uptake and wash-out of radioactive tracers or other ‘exogenous tracers that can be quantified from image sequences using well-characterized imaging equipment and calibration procedures. For MR perfusion images, exogenous RF pulse N\ ff} Stationary Poa ¢ ! Stationary “a andmaving spins spins FIGURE 13-18 Left. Phase dispersion of stationary and moving spins under the influence of an applied gradient (no flow compensation) as the gradient is inverted is shown. The stationary spins return to the onginal phase state, whereas the moving spins do not. Right. Gradient moment nulling of first order linear velocity (flow compensation) requires a doubling of the negative gradient amplitude followed by a positive gradient such that the total summed area is equal to zero. This will retum both the stabonary spins and the moving spins back to their original phase state.and endogenous tracer methods are used. Freely diffusible tracers using nuclei such as ‘H (deuterium), ‘He, "O, and “F are employed in experimental procedures to produce differential contrast in the tissues, More clinically relevant are intravascular blood pool agents such as gadolinium-diethylenetriaminepentaacetic acid, which ‘modify the relaxation of protons in the blood in addition to producing a shorter T2* This produces signal changes that can be visualized in pre and post contrast images (Fig. 13-19). Endogenous tracer methods do not use externally added agents, but instead depend on the ability to generate contrast from specific excitation or diffusion mechanisms. For example, labeling of inflowing protons (black blood” perfusion) uses protons in the blood as the contrast agent. Tagged (labeled) protons outside of the imaging volume perfuse into the tissues, resulting in a drop in signal intensity, a time course of events that can be monitored by quantitative measurements Functional MRI (fMRI) is based on the increase in blood flow to the local vascu- lature that accompanies neural activity in the specific areas of the brain, resulting in a local reduction of deoxyhemoglobin because the increase in blood flow occurs without an increase in oxygen extraction, As deoxyhemoglobin isa paramagnetic agent, it alters the T2*-weighted MRI image signal. Thus, this endogenous contrast enhancing agent serves as the signal for [MRI Area voxels (represented by x-y coordinates and 2 slice thickness) of high metabolic activity resulting from a task-induced stimulus produce correlated signal for Blood Oxygen Level Dependent (BOLD) acquisition techniques. A. BOLD sequence produces multiple T2*-weighted images of the head before the appli- cation of the stimulus. The patient is repeatedly subjected to the stimulus and multiple BOLD images are acquired. Because the BOLD sequence produces images that are highly dependent on blood oxygen levels, areas of high metabolic activity will dem- onstrate a change in signal when the prestimulus image data set is subtracted, voxel by voxel, from post-stimulus image data set. Voxel locations defined by significant rE Pre {top row) and post (bottom row) gadolinium contrast T!-weighted MR axial images of the brain ilustrate the signal change that occurs with the appearance of gadolinium by shortening the T1 time of the perfused tssuesChapter 13 Magnetic Resonance Imaging 471 signal changes indicate regions of the brain activated by a specific task. Stimuli in (MRI experiments can be physical (finger movement), sensory (light flashes of sounds), or cognitive (repetition of “good” or “bad” word sequences, complex problem solving) among others. To improve the SNR in the {MRI images, a stimulus is typically applied ina repetitive, periodic sequence, and BOLD images are acquired continuously, tagged with the timing of the stimulus. Regions in the brain that demonstrate time-dependent activity and correlate with the time-dependent application of the stimulus are sta- Uistically analyzed, and coded using a color scale, while voxels that do not show a significant intensity change are not colored. The resultant color map is overlaid onto a grayscale image of the brain for anatomic reference, as shown in Figure 13-20 High-speed imaging and 12* weighting necessary for {MRI is typically achieved with EPL acquisition techniques that can be acquired in as little as 50 ms for a 64 64 acquisition matrix, Gradient Recalled Echo acquisitions using standard sequences are also employed with multiple contiguous slices (e.g., 16 slices, slice thickness 5 mm, TR = 3s, TE = 60 ms, 90-degree flip angle) at 1.5 T, with 25 to 30 complete head acquisitions. The latter acquisition techniques provide better spatial resolution but rely on very cooperative subjects and a much longer exam time Diffusion Weighted Ima ig Diffusion relates to the random motion of water molecules in tissues. Interaction of the local cellular structure with the movement of water molecules produces aniso- tropic, directionally dependent diffusion (e.g,, in the white matter of brain tissues) Bilateral finger tapping paradigm 1-0.728 11 ovo} FIGURE 13-20 Functional MR image bilateral finger tapping paradigm shows the areas of the brain acti- vated by this repeated activity. The paradigm was a right finger tap alternated by a left finger tap (time sequence on the right side of the figure) and the correlated BOLD signals (black traces) derived from the echo planar image sequence. A voxel-by-voxel correlation of the periodic stimulus and MR signal is performed, and When exceeding a correlation threshold, a color overlay is added to the grayscale image. In this example, red inccates the right finger tap thatexctes the left motar cortex, which appears onthe nght seo the image and blue the left finger tap. (Figure compliments of MH Buonacore, MD, PhD University of California Davis472 Section Il + Diagnostic Radiology FIGURE 13-21 The basic elements of @ DWI pulse sequence are shown. The diffusion weight- re—h h ing gradients are of amplitude G, duration of the gradients & and time between gradients is A $86 7 *—_———_ pw Ls ! gradients i = a PEG eG —— $$ r S.qnai —> Diffusion sequences use strong MR gradients applied symmetrically about the relo- cusing pulse to produce signal differences based on the mobility and directionality of water diffusion, as shown in Figure 13-21. Tissues with more water mobility (normal) have a greater signal loss than those of lesser water mobility (injury) under the influ- ence of the diffusion weighted imaging (DWI) gradients. The in vivo structural integrity of certain tissues (healthy, diseased, or injured) can be measured by the use of DWI, in which water dilfusion characteristics are deter- mined through the generation of apparent diffusion coefficient (ADC) maps. This, equires two or more acquisitions with different DWI parameters. A low ADC cor responds to high signal intensity on a calculated image, which represents restricted diffusion (Fig, 13-22). ADC maps of the brain and the spinal cord have shown prom- 1 FIGURE 13-22 Left Diffusion weighted image. Right Calculated ADC image. shownng an area of increased brightness related to restricted mobulty of water molecules.Chapter 13 * Magnetic Resonance Imaging 473 ise in predicting and evaluating pathophysiology before it is visible on conventional T1- or T2-weighted images. DWI is also a sensitive indicator for early detection of ischemic injury Areas of acute stroke show a drastic reduction in the diffusion coefficient compared with nonischemic tissues. Diagnosis of multiple sclerosis is a potential application, based on the measurements of the diffusion coefficients in three- dimensional space Various acquisition techniques are used to generate diffuston-weighted contrast Standard spin echo and EPI pulse sequences with applied diffusion gradients of high strength are used. Challenges for DWI are the extreme sensitivity to motion of the head and brain, which is chiefly caused by the large pulsed gradients required for the diffusion preparation. Eddy currents are also an issue, which reduce the effectiveness of the gradient fields, so compensated gradient cails are necessary, Several strategies hhave been devised to overcome the motion sensitivity problem, including common electrocardiographic gating and motion compensation methods 134] Magnetization Transfer Contrast Magnetization transfer contrast is the result of selective observation of the interaction, between protons in free water molecules and protons contained in the macromolecules of a protein, Magnetization exchange occurs between the two proton groups because of coupling or chemical exchange, Because the protons exist in slightly different mag- netic environments, the selective saturation of the protons in the macromolecule can be excited separately from the bulk water by using narrow-band RF pulses (because the Larmor frequencies are different). A transfer of the magnetization from the protons in the macromolecule partially saturates the protons in bulk water, even though these pro- tons have not experienced an RF exeitation pulse (Fig, 13-23). Reduced signal from the Hydration Layer ny Hy e fon ng Mn RaPay Se Macromolecule He an Bulk Water Sone Sen? ye Tags" “8 NS ne Oe Meta “ote Putonchoma | ) Rr Excitaon: _ 4 Protons of (off-resonance pulse) Ve bulk water ~1.5 kHz shift Frequency Spectrum IEFIGURE 13-23 Magnetization transfer contrast is implemented with an off-resonance RF pulse of about 1,500 He from the Larmor frequency. Excitation of hydrogen atoms on macromolecules is transferred via the hydration layer to agjacent "free-water” hydrogen atoms. A partial saturation of the tissues reduces the sig- nals that would otherwise compete with signals from blood flow, making this useful for time-of-flight MRA,47a Section | * Diagnostic Radiology adjacent free water protons by the saturation “label” affects only those protons having a ‘chemical exchange with the macromolecules and improves local image contrastin many situations by decreasing the otherwise large signal generated by the protons in the bulk water. This technique is used for anatomic MRI of the heart, the eye, multiple sclerosis, knee cartilage, and general MRA. Tissue characterization is also possible, because the image contrast in partis caused by the surface chemistry of the macromolecule and the tissue-specific factors that affect the magnetization transfer characteristics. Magnetization transfer contrast pulse sequences are often used in conjunction with MRA time-of-flight methods. Hydrogen atoms constitute a large fraction of mac- romolecules in proteins, ate tightly bound to these macromolecules, and have a very short T2 decay with a broad range of resonance frequencies compared to protons in fice water. Selective excitation of these protons is achieved with an off-resonance RF pulse of approximately 1,500 Hz from the Larmor frequency, causing their satura- tion. The protons in the hydration layer bound to these molecules are affected by the magnetization and become partially saturated themselves. MR signals from these tissues are suppressed, with an impact of reducing the contrast variation of the anat- omy As a result, the differential contrast of the flow-enhanced signals is increased. with overall better image quality angiographic sequence MR Artifacts Artifacts manifest as positive or negative signal intensities that do not accurately rep- resent the imaged anatomy. Alhough some artifacts are relatively insignificant and are easily identified, others can limit the diagnostic potential of the exam by obscur- ing or mimicking pathologic processes or anatomy: One must realize the impact of MR acquisition protocols and understand the etiology of artifact production 10 exploit the information they convey. ‘To minimize the impact of MR artifacts, a working knowledge of MR physics as well as image acquisition techniques is requited. On the one hand. there are many variables and options available that complicate the decision-making process for MR image acquisition, On the other, the wealth of choices enhances the goal of achieving diagnostically accurate images. MR artifacts are classified into three broad areas— those based on the machine, on the patient, and on signal processing, Machine-Dependent Artifacts Magnetic field inhomogeneities are either global or focal field perturbations that lead to the mismapping of tissues within the image, and cause more rapid T2 relaxation Distortion or misplacement of anatomy occurs when the magnetic field 1s not com- pletely homogeneous, Proper site planning, sell-shielded magnets, automatic shim- ming, and preventive maintenance procedures help to reduce inhomogeneities Focal field inhomogeneities arise from many causes. Ferromagnetic objects in oF oon the patient (e.g, makeup, metallic implants, prostheses, surgical clips. dentures) produce field distortions and cause protons to precess at frequencies different from the Larmor frequency in the local area. Incorrect proton mapping, displacement, and appearance as a signal void with a peripherally enhanced rim of increased signal ate Common findings Geometric distortion of surrounding tissue ts also usually evi- dent, Even nonferromagnetic conducting materials (e.g.. aluminum) produce field distortions that distur the local magnetic environment, Partial compensation by the spin echo (180-degree RF) pulse sequence reduces these artifacts; on the otherChapter 13 * Magnetic Resonance Imaging 475, hand, the gradient-refocused echo sequence accentuates distortions, since the protons always experience the same direction of the focal magnetic inhomogencities within the patient Susceptibility Artifacts Magnetic suscepubility is the ratio of the induced internal magnetization in a tissue to the external magnetic field. As long as the magnetic suscepuilty of the tissues being imaged is relatively unchanged across the ficld of view, then the magnetic field will remain uniform. Any drastic changes in the magnetic susceptibility will distort the magnetic field, The most common susceptibility changes occur at tissue-air anter= faces (e.g, lungs and sinuses), which cause a signal loss due to more rapid dephasing (12+) at the tissue-air interface (Fig, 13-24). Any metal (ferrous or not) may have a significant elfect on the adjacent local ussues due to changes in suscepubility and the resultant magnetic field distortions. Paramagnetic agents exhibit a weak magne- tization and increase the local magnetic field causing an artifactual reduction in the surrounding T2* relaxation Magnetic susceptibility can be quite helpful in some diagnoses. Most notable is the ability to diagnose the age of a hemorrhage based on the signal characteristics of the blood degradation products, which are different in the acute, subacute, and chronic phases. Some of the iron-containing compounds (deoxyhemoglobin, methemoglobin, hemosiderin, and ferritin) can dramatically shorten T1 and T2 relaxation of nearby pro- tons. The amount of associated free water, the type and structure of the iron-containing molecules, the distribution (intracellular versus extracellular), and the magnetic field strength all influence the degree of relaxation effect that may be seen. For example, in the acute stage, T2 shortening occurs due to the paramagnetic susceptibility of the organized deoxyhemoglobin in the local area, without any large effect on the T1 relax- ation time. When red blood cells lyse during the subacute stage, the hemoglobin altered into methemoglobin, and spin-lattice relaxation is enhanced with the formation of a hydration layer, which shortens TI relaxation, leading to a much stronger signal BE FIGURE 13-24 Suscepnbaity artifacts due to dental filings are shown in the same axial image sce. Left. Axial T2-weighted fast spin echo image illustrates significant suppression of susceptibility artifacts with 180-degree refocusing pulse. Right. Axial T2*-weighted gracient echo image ilstrates significant image vord exacerbated by the gradient echo, where external inhomogeneities are not canceled inthe reformed echo476 Section Il + Diagnostic Radiology on Tl-weighted images. Increased signal intensity on T1-weighted images not found in the acute stage of hemorrhage identifies the subacute stage. In the chronic stage, hemosiderin, found in the phagocytic cells in sites of previous hemorrhage, disrupts the local magnetic homogeneity, causes loss of signal intensity, and leads to signal void, producing a characteristic dark rim around the hemorthage site Gadolinium-based contrast agents (paramagnetic characteristics shorten T2 and hydration layer interactions shorten TL) are widely used in MRI. Tissues that uptake gadolinium contrast agents exhibit shortened T1 relaxation and demonstrate increased signal on T1-weighted images. Although focal inhomogeneities are generally consid- eted problematic, there are certain physiologic and anatomic manifestations that can be idemtfied and diagnostic information obtained Gradient Field Artifacts Magnetic field gradients spatially encode the location of the signals emanating from excited protons within the volume being imaged. Proper reconstruction requires lin- car, matched, and properly sequenced gradients. The slice select gradient defines the volume (slice), Phase and frequency encoding gradients provide the spatial informa- tion in the other wo dimensions ince the reconstruction algorithm assumes ideal, linear gradients, any deviation or temporal instability will be represented as a distortion. Gradient strength has a tendency to fall off at the periphery of the FOV. Consequently, anatomic compres- sion occurs, especially pronounced on coronal and sagittal images having a large FOV, typically greater than 35 cm (Fig, 13-25), Minimizing the spatial distortion entails either reducing the FOV by lowering the gradient field strength or by hold- ing the gradient field strength and number of samples constant while decreasing the frequency bandwidth. Of course, gradient calibration is part of a continuous quality control (QC) checklist, and geometric accuracy must be periodically verified. Anatomie proportions may simulate abnormalities, so venfication of pixel dimen- sions in the PEG and FEG directions are necessary. I the strength of the FEG and the strength ofthe largest PEG are different, the height or width of the pixels can become dis- torted and produce inaccurate measurements, Ideally, the phase and frequency encod ing gradients should be assigned to the smaller and larger dimensions of the object respectively, to preserve spatial resolution while limiting the number of phase encoding steps. In practice, this is not always possible, because motion artifacts or high-intensity FIGURE 13-25 Gradient nonlinearity causes image distortions by mis-mapping anatomy. In the above ‘examples, the strength of the gradient at the periphery is iess than the iceal (orange line versus black line). This results in a compression of the imaged anatomy, with inaccurate geometry (images with orange border). For ‘comparison, images acquired with linear corrections are shown above.Chapter 13 * Magnetic Resonance Imaging 477 signals that need to be displaced away from imporuant ateas of interest after an initial scan might require swapping the frequency and phase encode gradient directions RF Coil Artifacts RF surface coils produce variations in uniformity across the image caused by RF excitation variability, attenuation, mismatching, and sensitivity falloff with distance Proximal to the surface coil, receive signals are intense, and with distance, signal Intensity 1s attenuated, resulting in grayscale shading and loss of brightness in the image, Nonuniform image intensities are the all-too-frequent result. Also, compensa- tion for the disturbance of the magnetic field by the patient is typically compensated by an automanie shimming calibration. When this is not performed, or performed snadequately, a significant negative impact on image quality occurs. Examples of vari- able response are shown in Figure 13-26. Other common artifacts from RF coils occur with RF quadrature coils (coils that simultaneously measure the signal from orthogonal views) that have two separate ampli- fer and gain controls. If the amplifiers are imbalanced, a bright spot in the center of the image. known as a center point artifact, arises as a “O frequency” direct current offset Vanations in gain between the quadrature coils can cause ghosting of objects diagonally in the image. The bottom line for all RF coils is the need for continuous measurement and consistent calibration of their response, so that artifacts are minimized. RF Artifacts RF pulses and precessional frequencies of MRI instruments occupy the same fre- quencies of common RF sources, such as TV and radio broadcasts, electric motors, fluorescent lights, and computers, Stray RF signals that propagate to the MRI antenna can produce various artifacts in the image. Narrow-band noise creates noise pat- tems perpendicular to the frequency encoding direction. The exact position and spatial extent depends on the resonant frequency of the imager, applied gradient field strength, and bandwidth of the noise. A narrow band pattern of black/white ~ Coil close to skin Inadequate shimming for fat saturation recewve cols are 100 close to the skin, as exern- FIGURE 13-26 Signal intensity vanations occur when surface RF recewe coi i plified by the MR breast imace on the left. With inadequate shimming calibration, saturation pulses for adipose tissue in the breast 1s uneven, causing a significant variation in the unvformity of the reconstructed image. From Hendrick RE, Breast MRI. fundementals and technical aspects. New York, NY: Springer, 2007. By permission478 Section Il * Diagnostic Radiology alternating noise produces a “zipper” artifact, Broadband RF noise disrupts the image over a much larger area of the reconstructed image with diffuse, contrast-reducing “herringbone” artifacts. Appropriate site planning and the use of properly installed RF shielding materials (e.g., a Faraday cage) reduce stray RF interfetence to an accept- ably low level. An example RF zipper artifact is shown in Figure 13-44 RF energy received by adjacent slices during a multislice acquisition excite and saturate protons in adjacent slices, chiefly due to RF pulses without sharp off/on/ off transitions. This is known as cross-excitation. On T2-weighted images, the slice- to-slice interference degrades the SNR, on T1-weighted images, the extra partial saturation reduces image contrast by reducing longitudinal recovery during the TR interval, A typical truncated “sinc” RF profile and overlap areas in adjacent slices are shown in Figure 13-27. Interslice gaps reduce the overlap of the profile tails, and pseudo-rectangular RF pulse profiles reduce the spatial extent of the tails. Important anatomic findings could exist within the gaps, so slice interleaving is a technique to mitigate cross-excitation by reordering slices into two groups with gaps. During the first half of the TR, the first slices are acquited (slices 1 to 5), followed by the second group of slices that are positioned in the gap of the first group (slices 6 to 10). This method reduces cross-excitation by separating the adjacent slice excitations in time The most effective method is to acquire two independent sets of gapped multislice images, but the image time is doubled. The most appropriate solution is to devise RF pulses that approximate a rectangular profile; however, the additional time necessary for producing such an RF pulse can be prohibitive K-Space Errors Errors in k-space encoding affect all areas of the reconstructed image, and cause the artifactual superimposition of wave patterns across the FOV. Each individual pixel value in k-space contributes to all pixel values in image space as a frequency harmonic with a signal amplitude. One bad pixel introduces a significant artifact, rendering the image suboptimal, as shown in Figure 13-28. FIGURE 13-27 Top. Poor pulse pro- files are caused by truncated RF pulses, and resulting profile overlap causes ‘unwanted partial saturation in adjacent slices. with a loss of SNR and CNR. Opti- mized pulses are produced by consider- ing the tradeotf of pulse duration versus excitation profile. Bottom. Reduction of cross-excitation is achieved with inter- slice gaps, but anatomy at the cap loca- tion might be missed. An interleaving technique acquires the first half of the images with an intersice gap, and the second half of the images are positioned in the gaps of the fist images. The sep- aration in time reduces the amount of contrast reducing saturation of the adja- cent slices “sinc profile overlapped Interslice gap Interleaving “computer-optimized” “rectangular” profile profile a Nig Agno, is J ieChapter 13 * Magnetic Resonance Imaging 479 Bad pixel in k-space Resultant image IB FIGURE 13-28 A single bad pixel in k-space causes a significant artifact in the reconstructed image. The bad pixel is located at k, = 2, k, = 3, which produces a superimposed sinusoidal wave on the spatial domain image as shown above Motion Artifacts The most ubiquitous and noticeable artifacts in MRI arise with patient motion, including voluntary and involuntary movement, and flow (blood, CSF). Although motion artifacts are not unique to MRI, the long acquisition time of certain MRI sequences increases the probability of motion blurring and contrast resolution losses. Motion anifacts occur mostly along the phase encode direction, as adjacent phase encoding measurements in k-space are separated by a TR interval that can last 3,000 ms or longer. Even very slight motion can cause a change in the recorded phase variation across the FOV throughout the MR acquisition sequence. Examples of motion artifacts are shown in Figure 13-29 The frequency encode direction is less aflected, especially by peiodic motion, since the evolution of the echo signal, frequency encoding, and sampling occur simultaneously over sevetal milliseconds, Ghost images, which are faint copics of the image displaced along the phase encode direction, are the visual result of patient motion Several techniques can compensate for motion-related artifacts, The simplest technique transposes the PEG and FEG to relocate the motion artifacts out of the ins, are most always displayed in the phase encode le ghost images of the anatomy, since the variation im FIGURE 13-29 Motion arifacts, particularly of flow patt ‘gradient direction. Slight changes in phase produce multi in phase caused by motion can be substantial between excitations480 Section Il * Diagnostic Radiology region of diagnostic interest with the same puilse sequences. This does not reduce the magnitude of the artifacts, however, and often there is a mismatch when placing the PEG along the long axis of a rectangular FOV (c.g. an exam of the thoracic spine) in terms of longer examination times of a significant loss of spatial resolution or of SNR. There are other motion compensation methods: 1, Cardiac and respiratory gating—signal acquisition at a particular cyclic location synchronizes the phase changes applied across the anatomy (Fig. 13-30) 2. Respiratory ordering of the phase encoding projections based on location within the respiratory cycle, Mechanical or video devices provide signals to monitor the cycle 3, Signal averaging to reduce artifacts of random motion by making displaced sig- nals less conspicuous relative to stationary anatomy. 4. Short TE spin echo sequences (limited to proton density, Tl weighted scans, frac- tional echo acquisition, Fig. 13-3). Note: Long TE scans (T2 weighting) are more susceptible to motion. 5. Gradient moment nulling (additional gradient pulses for flow compensation) to help rephase protons that are dephased due to motion, Most often, these techniques require a longer TE and are more useful for T2-weighted scans (Fig, 13-18). 6. Presaturation pulses applied outside the imaging region to reduce flow artifacts {rom inflowing protons, as well as other patient motions that occur in the periph- ery (Fig. 13-12) 7. Muluple redundant sampling in the center of k-space (¢-g., propeller) to identify and remove those sequences contnbuting to motion, without deletenously alfect- ing the image (Fig 13-8) Chemical Shift Artifacts of the First Kind There are two types of chemical shift artifacts that affect the display of anatomy due to the precessional frequency differences of protons in fat versus protons in water. Chemical shift refers to the resonance frequency variations resulting {rom intrinsic Non Gated ECG Signal | Velocity i 13 i i [eis Nias ANETTISZ ECG Gated cesta | | Velocity i 4 on | Like Le, LA EFIGURE 13-30 Motion artifacts occur when data is acquired without consideration of physiologic periodic- ity. Top. The electrocardiogram measures the R-wave at each heartbeat, but data acquisition proceeds in a linear fashion without regard to reproducibility. The result is a set of images degraded with motion artifact, with diagnostic usefulness marginal, at best. Bottom. Acquis'ton of images proceeds with the detection of the R-wave signal and synchronization of the collection of image data in a stepwise fashion over the period between Rewaves. A reduced number of mages or extended accuisition time ts required to collet the data,Chapter 13 * Magnetic Resonance Imaging 481 magnetic shielding of anatomic structures. Molecular structure and electron orbital characteristics produce fields that shield the main magnetic field and give rise to dis- tinct peaks in the MR spectrum. In the case of proton spectra, peaks correspond to water and fat, and in the case of breast imaging, silicone material is another material to consider. Lower frequencies of about 3.5 parts per million for protons in fat and 5.0 parts per million for protons in silicone occur, compared to the resonance frequency of protons in water (Fig, 13-31). Since resonance frequency increases linearly with field strength, the absolute difference between the fat and water resonance also increases, making high field strength magnets more susceptible to chemical shift artifact. Data acquisition methods cannot directly discriminate a frequency shift due to the application of a frequency encode gradient or a chemical shift artifact, Water and {at differences therefore cannot be distinguished by the frequency difference induced by the gradient. The protons in fat resonate at a slightly lower Irequency than the corresponding protons in water, and cause a shift in the anatomy (misregtstration of water and fat moieties) along the frequency encode gradient direction, ‘A sample calctilation in the example below demonstrates frequency variations in fat and water for two different magnetic field and gradient field strengths. ‘Chemical shift artifact numerical calculation for field strength, with a 3.5-ppm (3.5 X 10-®) variation in resonance frequency between fat and water results in the following frequency differences: L5 T: 63.8 X 10°Hz X 3.5 x 10° 30 T: 127.7 X 10°Hz x 3.5 X 10° = 447 He 223 Hz ‘Thus, the chemical shift is more severe for higher field strength magnets Chemical shift artifact numerical calculation for gradient strength results in the following numerical calculations for a 25-cm (0.25 m) FOV, 256 X 256 matrix: Low gradient strength: 2.5 mT/m % 0.25 m = 0.000625 T variation, gives frequency range of 0.000625 T X 42.58 MH2/T = 26.6 kHz actoss FOV and 26.6 kH2/256 pixels = 104 Hz/pixel High gradient strength: 10 m ange of 0.0025 T X 42.58 M pixels = 416 Hz/pixel \T/m X 0.25 m = 0.0025 T variation, gives frequency H2/T = 106.5 kHz actoss FOV and 106.5 kH2/256 ‘Thus, a chemical shift occurrence is more severe for lower gradient strengths, since displacement will occur over a large number of pixels. With a higher gradient Strength, water and fat are more closely contained within the broader pixel boundary bandwidths. Normal and low bandwidth images are illustrated in Figure 13-32 MEFIGURE 13-31 Chemical shift refers to ssa . ss ; water Chemical sen the slightly different precessional frequen- - ifferent materials oF tis- tat Ges of protons in i NRT 24 pm renee ess po ae eters 0 -_ " water for fat and silicone. Fat chernical shift ‘ artifacts are represented by a shitt of water: ‘and fat in the images of anatomical struc + * ‘ture, mainly in the frequency encode gract ent direction Swapping the PEG and the FEG wall cause a shift ofthe fat and water = a ‘components ofthe tissues in the image. —_ — Fi PEG «6Section I « Diagnostic Radiology High Bandwidth FIGURE 13-32 MR images of the breast, containing glandular and adipose tissue, are acquired under a high bandwidth (32 kHz) and a low bandwidth (4 kH2),dlustrating the more severe chemical shift with low readout gradient steength and bandwidth (Reprinted by permission, Hendrick RE. Breast MRI: fundamentals and tech- ical aspects. New York, NY: Springer, 2007.) RF bandwidth and gradient strength considerations can mitigate chemical shift ari- facts. While higher gradient strength can confine the chemical shift of fat within the pixel bandwidth boundaries, a significant SNR penalty occurs with the broad RF bandwidth required to achieve a given slice thickness. A more widely used method is to use lower gradient strengths and narrow bandwidths in combination with off-resonance “chemi- cal presaturation” RF pulses to minimize the fat (or the silicone) signal in the image (Fig 13-33). Another alternative is to use STIR techniques to eliminate the signals due to fat at the “bounce point.” Swapping the phase and frequency gradient directions or chan; ing the polarity of the frequency encode gradient can displace chemical shift artifacts from a specific image region, even though the chemical shift displacement still exists Identification of fat in a specific anatomic region is easily discerned from the chemical shift artlact displacement caused by changes in FEG/PEG gradient directions Chemical Shift Artifacts of the Second Kind (Chemical shift anifacts ofthe second kind occur with GE images. resulting from the rephas ing and dephasing of the echo in the same direction relative to the main magnetic field Signal appearance is dependent on the selection of TE. This happens because of construc tive (in phase) or destructive (out of phase) transverse magnetization events that occur periodically due to the difference in precessional frequencies. At 1.5 T, the chemical shift IE FIGURE 13-33 The left image of the lumbar spine is Spin Echo T1 weighted, TR = 450 ms, TE = 14 ms. The night image is T1 weighted with chemical fat saturation pulses, TR = 667 ims, TE = 8 ms, In both images, the FEG ts verti= cally orientedChapter 13 * Magnetic Resonance Imaging 483 Water & Fat Out-of-Phase at 22, 6.6, 11.0 ms, FIGURE 13-34 For GE image sequences, signal intensity of the transverse magnetization vector due to the 220 Hz lower precessional frequency of fat protons, where in-phase magnetization occurs every 4.4 ms (172205), and out-of-phase magnetization occurs every 4.4 ms shifted by Ys cycle (2.2 ms). Signal intensity is dependent on the selection of TE, as shown above. is 220 Hz, and the periodicity of each peak (in phase) between water and fat occurs at 0. 4.5, 9.0, 13.5, ms, and each valley (out of phase) at 2.25, 6.75, 11.0, .... ms, as shown in Figure 13-34 Thus, selection of TE at 9 ms will lead to a constructive addition of water and fat, and TE at 7 ms will lead to a destructive addition of water and fat. The in-phase timing will lead to a conventional chemical shuft image ofthe first kind, while the out-of- pphase timing will lead to chemical shift image of the second kind, manifesting a dark rim around heterogeneous water and fat anatomical structures, shown in Figure 13-35. Ringing Artifacts Ringing anifact (also known as Gibbs phenomenon) occurs near sharp boundaries, and high-contrast transitions in the image, and appears as multiple, regularly spaced parallel bands of alternating bright and dark signal that slowly fades with distance. TE=9ms TE =7ms Water & Fat in-phase Water & Fat out-of-phase FIGURE 13-35 Breast MAI images show the effect of selecting a specific TE for a GE acquisition. On the left, chemical shift ofthe “Tirst kind” is shown wath TE = 9 ms and water and fat in phase for tansverse magnet zation, shifted only due to the intrinsic chemical shift differences of fat and water. On the right, chemical shift Of the second kind is additionally manifested with TE = 7 ms, due to fat and water being out of phase, creating ‘a lower signal at all fat-water interfaces, and resuiting in reduced intensity (Reprinted by permission, Hendnck RE. Breast MRt: fundamentals and technical aspects. New York, NY: Springer, 2007.)Section Il + Diagnostic Radiology The cause is the insufficient sampling of high frequencies inherent at sharp disconti- rites in the signal, Images of objects can be reconstructed from a summation of sinu- soidal waveforms of specific amplitudes and frequencies, as shown in Figure 13-36 for a simple rectangular object. In the figure, the summation of frequency harmonics, each with a particular amplitude and phase, approximates the distribution of the object, but initially does very poorly, particularly at the sharp edges. As the number of higher fre- quency harmonics increase, a better estimate is achieved, although an infinite number of frequencies are theoretically necessary to reconstruct the sharp edge perfectly. In the MR acquisition, the number of frequency samples is determined by the number of pixels (frequency, k,, or phase, k,, increments) across the k-space matrix. For 256 pixels, 128 discrete frequencies are depicted, and for 128 pixels, 64 discrete frequencies are specified (the k-space matrix is symmetric in quadrants and dupli- cated about its center). A lack of high-frequency signals causes the “ringing” at sharp transitions described as a diminishing hyper- and hypointense signal oscillation from the transition. Ringing artifacts are thus more likely for smaller digital matrix sizes (Fig, 13-37, 256 versus 128 matrix). Ringing artifact commonly occurs at skull/brain interfaces, where there is a large transition in signal amplitude A 7 8. Frequency |’ ‘pati harmonics | domain | 1 rectangie | '| : estimate | 1+ Bed eaten |. —- esteem a7). — FIGURE 13-36 The syrthesis of a spa-_C. Sharptransition in MR image: ‘val object occurs by the summation of frequency harmonics in the MR image ‘A. Left column: frequency harmonics that estimate a rectangle function wth progres- sively higher frequencies and lower ampt- “Ringing” tudes are shown. B. Middle column. As NX higher frequency harmonics are included, the summed result more faithfully rep: resents the object shape, in this example a rectangle with two vertical edges. The ‘number of frequencies encoded in the MR image is dependent on the matrix size Right column: A sharp transition bourd- ay i an MR image is represented with 256 72256 samples better than with128 samples ({re- samples quency harmonics in k-space). The amount fringing caused by insutficent samping is reduced with a larger number of samples.Chapter 13 * Magnetic Resonance Imaging 485, 256 (vertical) x 128 (horizontal) 256 x 256 FIGURE 13-37 Example of ninging artifacts caused by a sharp signal tran- sition at the skull in a brain image for a 256 X 128 matrix (eft) along the short (horizontal axis, and the elimi: nation of the artifact in a 256 X 256 matrix (right). The short axis defines the PEG direction Wraparound Artifacts The wraparound artifact is a result of the mismapping of anatomy that lies out- side of the FOV but within the slice volume. The anatomy is usually displaced to the opposite side of the image. It 1s caused by nonlinear gradients or by under- sampling of the frequencies contained within the returned signal envelope. For the latter, the sampling rate must be twice the maximal frequency that occurs in the object (the Nyquist sampling limit). Otherwise, the Fourier transform can- not distinguish frequencies that are present in the data above the Nyquist [re- quency limit, and instead assigns a lower frequency value to them (Fig. 13-38) Frequency signals will “wraparound” to the opposite side of the image, masquer- ading as low-frequency (aliased) signals. In the frequency encode direction, a low-pass filter can be applied to the acquired tume domain signal to eliminate frequencies beyond the Nyquist frequency. In the phase encode direction, aliasing anifacts can be reduced by increasing the number of phase encode steps (the trade-oll is increased image time). Another approach is to ‘move the region of anatomic interest to the center of the imaging volume to avoid the ‘Sampling rate FIGURE 13-38 Left. Wraparound artifacts are caused by aliasing. Shown isa fixed sampling rate and net pre- Cessional frequencies accurnng at poston A and positon B within the FOV that have identical frequencies but it ferent phase. f sgnal from position Cis at twice the frequency of B and insutfiiently sampled, the same frequency and phase wall be assigned to C as that assigned to A, and therefore will appear at that location. Right. A wrap- around artfact example csplaces anatomy from one side of the image (or outude of the FOV} to the other side486 Section Il * Diagnostic Radiology overlapping anatomy, which usually occurs at the periphery of the FOV. An “antialias- ing” saturation pulse just outside of the FOV is yet another method of eliminating high-frequency signals that would otherwise be aliased into the lower frequency spec- trum. This example of wrap-around artifact is easy to interpret. In some cases, the antfact is not as well delineated (e.g., the top of the skull wrapping into the brain) Partial Volume Artifacts Partial volume artifacts anse from the finite size of the voxel over which the signal is averaged. This results in a loss of detail and spatial resolution. Reduction of partial volume artifacts is accomplished by using a smaller pixel size and/or a smaller slice thickness. With a smaller voxel, the SNR is reduced for a similar imaging time, result- ing ina noisier signal with less low-contrast sensitivity. Of course, with a greater NEX (averages), the SNR can be maintained, at the cost of longer imaging time. 36 Magnetic Resonance Spectroscopy Magnetic resonance spectroscopy (MRS) is a method to measure tissue chemistry (an “electronic” biopsy) by recording and evaluating signals from metabolites by iden- tilying metabolic peaks caused by frequency shifts (in parts per million, ppm) rela- live to a frequency standard, In vivo MRS can be performed with 'H (proton), Na Godium), and "P (phosphorus) nuclei, but proton spectroscopy provides a much higher SNR and can be included in a conventional MRI protocol with about 1010 15 min extra exam time. Uses of MRS include serial evaluation of biochemical changes in tumors, analyzing metabolic disorders, infections and diseases, as well as evalua- tion of therapeutic oncology treatments for tumor recurrence versus radiation dam- age. Early applications were dedicated to brain disorders, but now breast, liver, and prostate MRS are also performed. Correlation of spectroscopy results and MR images are always advised before making a final diagnosis In MRS, signals are derived from the amplitude of proton metabolites in targeted tissues In these metabolites, chemical shifts occur due to electron cloud shielding of the nuclei, causing slightly different resonance frequencies, which exist in a fre- quency range between water and fat. The very small signal amplitudes of the metabo- lites require suppression of the extremely large (~10,000 times higher) amplitudes due to bulk water and fat protons, as shown in Figure 13-39. This is achieved by using specific chemical saturation techniques, such as CHESS (Chemical Shift-Selec- tive) or STIR (see Chaptet 12). In many cases, the areas evaluated are away from fat structures, and only bulk water signal suppression is necessary; however, in organs such as the liver and the breast, suppression of both fat and water are required. Once the water and fat signals are suppressed, localization of the targeted area volume is achieved by either a single voxel or multivoxel technique Single voxel MRS sampling areas, covering a volume of about 1 cm? are delin- eated by a STEAM (Stimulated echo acquisition mode) or a PRESS (Point Resolved Spectroscopy) sequence. The STEAM method uses a 90-degree excitation pulse and 90-degree refocusing pulse to collect the signal in conjunction with gradients to define each dimension of the voxel. The PRESS sequence uses a 90-clegree excitation and 180-degree refocusing pulse in each direction. STEAM achieves shorter echo times and superior voxel boundanes, but with lower SNR_ After the voxel data are col- lected, a Fourier transform is applied to separate the composite signal into individual frequencies, which are plotted as a trace for a normal brain spectrum (Fig. 13-40) The resulting line widths are based on homogeneity of the main magnetic field asChapter 13 * Magnetic Resonance imaging 487 ‘Water peak FIGURE 13-39 MRS metabolites of interest in comparison to the water and fat peaks commonly used for imaging, Jn order to isolate the very small signals, chemical saturation of the water (and fat 7 when present) signal is essential. Fat peak Relative Amplitude: 5 4 3 2 4 0 Frequency shift (ppm) well as the magnetic field strength. Higher field strengths (¢.g.. 3.0 T) will improve resolution of the peaks and corresponding SNR. Mubivoxel MRS uses a CSI (Chemical Shift Imaging) technique to delineate multiple voxels of approximately 1 cm’ volume in 1, 2, or 3 planes over a rectangular block of sev- eral centimeters, achievable with more sophisticated equipment and longer scan times This is followed by MRSI (Magnetic Resonance Spectroscopic Imaging) where the signal intensity of a single metabolite in each voxel is color encoded for each voxel according to concentration and the generated parameter maps superimposed on the anatomical MR image. In practice, the single voxel technique is used to make the initial diagnosis because the SNR is high and all metabolites are represented in the MRS trace. Then, a multivoxel acquisition to assess the distribution of a specific metabolite is performed. Proton MRS can be performed with short (20 to 40 ms), intermediate (135 to 145 ms), or long (270 to 290 ms) echo times. For short TE, numerous resonances from metabolites of lesser importance (with shorter T2) can make the spectra more difficult to interpret, and with long echo times, SNR losses are too severe. Therefore, most MRS acquisitions use a TE of approximately 135 ms at 1.5 T. Metabolites of interest for brain spectroscopy are listed in Table 13-2 Applications of MRS are achieved through the interpretation of the spectra that are oblained from the lesion, from its surroundings, and presumably healthy tissue in RactoGrapics 2006: 265173-5180 MR Spectrum trom anaplastic cligoastrocytoma Choline / Creatine ratio map BE FIGURE 13-40 Left: intermediate echo (TE=135 ms) single voxel spectrum s shown, positioned over an anaplastic ‘ligoastrocytoma brain lesion. Note the elevated Choline peak and lowered Creatine and NAA peas. Right: Mult ‘voxel spectrum is color codes to the Choline / Creatine rato. ilstrating the regional variaton of the metabolites cor- responding to tumor. From Al-Okail RN, Kreza J, Wang S, Woo JH, Meihem ER. Advanced MR imaging Techniques in the Diagnosis of itraaxial Brain Tumors in Adults. Radiographics 2006; 26. $173-5189. By permission.