Magnetic resonance imaging (MRI) is a non-invasive medical test that uses intense magnetic fields to make images of the inside of the body. Generally, the magnetic resonance (MR) image quality depends on the effect of MRI parameters on the signal-to-noise ratio (SNR) which is often calculated using samples taken from reconstructed image that will enhance image quality. Ogura, Miyai, Maeda, Fkutake and Kikumoto (2003) state that the SNR of a magnetic resonance image is common measure of imager performance. In “Magnetic Resonance Imaging: Physical and Biological Principles,” Steward C. Bushong (2003) defines SNR as comparison of the intensity of the information (signal) in the image to the intensity of noise in the image. In other words, SNR is used to describe the relative contributions to a detected signal of the true signal and random superimposed noise (Bushong, 2003).
The factors determine the SNR includes slice thickness and bandwidth, field-of-view (FOV), size of matrix, number of measurements, image parameters, magnetic field strength and choice of transmitter and receiver coils. Riddell, Richardson, Scurr and Grown (2006) note that the FOV, matrix size, slice thickness, number of signal averages (NSA), and repetition time (TR)/echo time (TE) were altered to optimize signal to noise ratio (SNR) whilst maintaining spatial resolution. These parameters are manipulated to increase SNR, define the region of interest and indirectly improve the image contrast and affect the resultant image. The higher the SNR in an image, the smoother and more appealing the image is to reader.
SLICE THICKNESS AND BANDWIDTH
Currently, high-resolution MRI with increased matrix size is routinely used. But it is necessary to consider that the effect of decreasing SNR when matrix size is increased. So, a thick slice is often used in order to compensate for a poor SNR. Increase slice thickness improves the SNR in direct proportion to the slice thickness. If the slice is thick, a good SNR can be achieved, but the contrast is reduced by the partial volume effect. Hazirolan, Gupta, Mohamed and Bluemke (2005) state that thin slice thickness decreases partial volume effects and causes a reduction in SNR. Ogura, Maeda, Miyai and Kikumoto (2005) write that thinning slices led to increased signal detection but to decreased SNR because of higher contrast in the partial volume effect. An increase in signal averaging and time scan must be applied to attain adequate SNR on thin slices for definition of structures with low contrast detectability. Increasing matrix size and slice thickness for high resolution imaging led to decrease in signal detection (Ogura et al., 2005).
Bandwidth (BW) is the amount of data that can be transmitted across a communication channel over a given period of time or the range of RF frequencies in a pulse or to which an MR receiver is tuned (Bushong, 2003). In simple words, bandwidth is the range of frequencies contained within the RF pulse. Toms, Smith-Bateman, Malcom, Cahir and Graves (2009) note that depending on the system’s vendor, the bandwidth can be set directly (in Hz/pixel) but other systems require specification of the full receiver bandwidth. An inverse relationship exists between bandwidth and SNR. Exactly, if the bandwidth is decreased by a factor of 2, the SNR will be improved by a factor of. Otherwise, a wider bandwidth causes more noise and decrease the SNR. Increasing the bandwidth produced a relative fall in the SNR of between 49% and 56% for a given matrix (Toms et al., 2009). Eventhough reducing the bandwidth is an effective way to promoting the SNR, it also increases the minimum TE and artefact known as chemical shift. Thus, reduced receive bandwidth is not suitable for T1 or PD imaging and should be used only when short TE is not required and fat is not present. The loss of SNR from increasing receiver bandwidth is preferable to long acquisition times (Toms et al., 2009).
The overlapping of adjacent frequency bandwidths cause an interference known as cross-talk. Kaur, Kumaran, Tripathi and Kushu (2006) write that cross-talk is characterized by slightly different intensity pattern in even and odd-numbered slices with reduction in contrast to noise ratio (CNR) and SNR in long TR and short TE images. Ideally, 90 and 180 RF pulses should result in perfectly rectangular slice profiles, thereby producing homogeneous excitation of the tissue within a desired slice and not affecting the tissue located near or outside the slice of interest (Kaur et al., 2006). Exactly, the frequency spectrums of the RF pulse have a bell shape and the cross-talk will appear if overlapping occurs when these frequency spectrums are placed close together. Cross-talk produces saturation effects, resulting in reduced SNR. To reduce or minimize this, a gap between the consecutive bandwidths is created, thus indirectly creating a gap between consecutive slices (interslice gap) in the actual image. Ideally, 30% gap should be used for slices 5 mm thick and 50% for slices <5 mm thick (Kaur et al., 2006).
FIELD OF VIEW (FOV) AND MATRIX
FOV is defined as anatomy contained within the volume imaged and determined by the product of acquisition matrix and pixel size (Bushong, 2003). Ideally, if we decrease the FOV (matrix size is held constant) or if we increase the matrix size (FOV is held constant), we decrease the pixel size. Crucially, there are three conceptual differences between changing the FOV and changing the matrix size. First, altering the matrix size changes the time of the examination, not the area being imaged. Second, altering the FOV alters the anatomical area being examined, not change the scan time. Third, altering the matrix size alters the pixel size only in phase-encoding direction whereas altering the FOV alters the pixel size in both direction, phase-encoding direction and frequency-encoding direction. As we know, there is a close relationship between FOV and SNR. FOV has a direct square-function relationship with SNR but SNR is inversely proportional to matrix size along the frequency axis and to the square root of the number of phase-encoding steps. Moratal, Valles-Luch, Marti-Bonmat and Brummer (2008) note that reducing the FOV in the phase encoding direction saves scan time by decreasing SNR but invariably maintains the spatial resolution. Here, we can conclude that a coarse matrix with large voxels is recommended in order to optimize the image quality by having high SNR. But, in order to attain good spatial resolution, a fine matrix with low SNR is required.
NUMBER OF MEASUREMENTS
Number of signals averaged (NSA) or also known as number of excitations (NEX) is averaging parameter that does not affect resolution and indeed reduces the noise. Bushong (2003) defines NSA as the number of times that an identical MR signal is collected for use in the same image and describes NEX as the number of signal acquisitions averaged to improve SNR. Basically, signal averaging is performed to increase SNR. Mathematically, SNR is proportional to the square root of NSA. A large NSA improves SNR and suppresses motion artefacts (Bushong, 2003). The problem with increasing NSA to increase SNR is that the scan time is undoubtedly proportional to NSA. As an example, in order to double the SNR, NSA must be increased by a factor of four and indirectly increase the total scan time. Increasing the NSA improved the SNR, an NSA of 6 was considered optimum as the scan time was increased to 9.28 min for 22 slices with NSA of 8, exceeding the maximum limit (Riddell et al., 2006).
IMAGE PARAMETERS (TR, TE and flip angle).
Time to repeat (TR), Time to echo (TE) and flip angle are parameters that influence image contrast, they also have an effect on SNR and overall image quality. TR is defined as the period between the beginning of a pulse sequence and the beginning of the succeeding and identical pulse sequence (Bushong, 2003). In other words, the amount of longitudinal magnetization that is allowed to recover before the next excitation pulse is applied depends on TR. If TR is long, full recovery of the longitudinal magnetization is allowed, so more is available to be flipped in the next repetition. In contrast, short TR results in less longitudinal magnetization available to be flipped due to incomplete recovery of the longitudinal magnetization. Thus, longer TR will increase SNR because of more complete T1 relaxation and increases scan time. In detail, as the TR increases, there will be more longitudinal magnetization to create transverse magnetization after excitation. Merkle and Dale (2005) note that the gain in SNR will be higher in T2-weighted sequences than in T1-weighted sequences because longer TRs allow more complete recovery of the longitudinal magnetization and transverse relaxation time (T2) are fairly independent of the main magnetic field strength.
Bushong (2003) defines TE as time interval between the middle of the excitation pulse and the middle of the echo signal that is observed. Basically, the amount of transverse magnetization that is allowed to decay before an echo is collected depends on TE. A long TE results in considerable decay of the transverse magnetization to occur before the echo is collected, while a short TE does not. So, if TE is increased, SNR will decreases due to less transverse magnetization available to be rephased and produce an echo. Thus, T2 weighted sequences that use a long TE will have a lower SNR than T1 or PD weighted sequences that use a short TE. The best SNR for all metabolites and tissue types for a respective scanner was obtained at the shortest TE (Englese et al., 2006). Reeder, Markl, Yu, Hellinger and Herfkens (2004) shown that a TE increment of 0.9ms decrease SNR by 17% from the maximum.
Flip angle is described as amount of rotation of the net magnetization vector produced by an RF pulse, with respect to the direction of static magnetic field (Bushong, 2003). Typically, gradient echo sequences uses flip angles between 0° and 90°, spin echo sequences uses 90° and a series of 180° pulses and inversion recovery sequences uses an initial 180° pulse followed by a 90° and 180° pulse. Actually, altering the flip angle of gradient echo sequence whilst considering on other parameter can either increase or decrease the SNR of an image. The amount of transverse magnetization is small as is the resultant SNR if the flip angle is extremely low. That means, increasing flip angle will increase SNR. However, at a point, although the flip angle keeps increasing, the signal reaches its maximum and begins to decrease. Ernst angle is the flip angle that produces the maximal signal for a given tissue at a given TR. So, if the flip angle is below the Ernst angle, lower the flip angle results in lower SNR and vice versa. But, if the angle is above the Ernst angle, lower the flip angle result in higher SNR and vice versa. Ideally, this Ernst angle specific for various tissue and depends on TR and TI time of the tissue.
MAGNETIC FIELD STRENGTH (Bo)
The magnetic field strength (Bo) is the most important parameter affecting SNR. Naturally, there are a few more protons on the lower energy level parallel to the magnetic field. The difference in number depends on the strength of the applied magnetic field. If stronger magnetic field is applied, the energy gap increases due to faster precession rate and higher precession frequency. Therefore, there are more protons pointing up than down with a resulting increase the size of net magnetic vector, which then leads to a larger transversal magnetization when disturbed by RF pulse that has same frequency. Thus, SNR also increases. In “The Promise of High-Field-Strength MR Imaging,” Joseph A. Helpern (2003) states that an 8T system has been operational for several years and there are even plans underway for the installation of 9.4T and higher human systems. One of the most obvious benefits for this continued advance in field strength is improved SNR (Helpern, 2003). Inglese, Spindler, Babb, Sunenshine, Law and Gonen (2006) write that for the same coil technology, the 3T yield the better SNR than its 1.5T counterpart. In “High-field human imaging,” Vikas Gulani (2005) notes that higher field strengths have become attractive in nearly all aspects of magnetic resonance due to the potential for increased SNR and diminished acquisition times. A straightforward analysis shows that SNR can be expected to increase at most approximately linearly with field strength (Gulani, 2005). In addition, spectroscopy at high field strengths is enhanced by the increase in SNR and higher field strengths afford improved spatial resolution in spectroscopy (Helpern, 2003). Spectroscopy or magnetic resonance spectroscopy (MRS) uses a continuous band of radio wave frequencies to excite hydrogen atoms in a variety of chemical compounds other than water. Most patients undergo an abdominal MRI study at 3.0T with a reasonable outcome in terms of image quality (Merkle & Dale, 2005). 3D acquisition technique, higher magnetic strengths (3T) and prone position provide higher SNR and could further improve accuracy of the coronary artery vessel wall imaging ( Hazirolan et al., 2005)
The type of coil has a significant effect on the signal received from the subject and therefore the SNR. Body which has capacitance and inductance becomes a source of noise during the absorption and emission of RF energy. When the body is placed in or next to a coil, loading occur in which the coil couples to the body electrically. Actually the amount of loading differs from one imaging event to the next because the body is never in exactly the same position with respect to the coil. In other words, different bodies load the coil differently. So, for each imaging task, the system must retune the coil to the resonance frequency to gain maximum SNR. That is why the use of the appropriate receiver coil plays an important role in optimizing SNR.
Quadrature coils increase SNR as two coils are used to receive signal. Phased array coils increase SNR even more as the data from several coils are added together. The SNR gain afforded by a phase-array coil upgrade is very cost-effective (Englese et al., 2006). Surface coils placed close to the area under examination also increase the SNR. Surface coils are usually used for the investigation of the specific anatomy such as spinal cord and shoulder or for the imaging of small anatomical structures. Surface coils also have improved spatial resolution because of smaller FOV. The benefit of using an endoluminal surface coil is that the coil is closer to the region of interest, thus improving SNR and as a consequence, the spatial resolution can be maximized (Riddell et al., 2006).
In general, appropriate size of the receiver coil should be chosen to ensure that the volume of tissue imaged optimally fills the sensitive volume of the coil. However, tissue outside the FOV is more likely to produce signal which result in large coils increase the likelihood of aliasing. In addition, the position of the coil also important for maximizing SNR. The coil must be positioned in the transverse plane perpendicular to Bo in order to induce maximum signal. Moreover, angling the coil also contribute in reduction of SNR.
Ideally, in order to get optimum image quality, SNR is increased by using thick slices, narrowest receive bandwidth, large FOV, coarse matrix, , long TR and short TE, flip angle of 90, stronger magnetic field, well tuned and correctly sized coil, and as many excitations and signal averages (NEX/NSA) as possible.