by Danqing Xiao
Bandwidth, sample averaging (NEX) and turbo factor (ETL) in related to scan time and image quality
MRI image quality is determined by signal to noise ratio (SNR), contrast to noise ratio (CNR), spatial resolution and scan time (Westbrook, 2008,2011; McRobbie, Moore, Graves & Prince, 2010). These factors are controlled by different parameters. Often time, changing one parameter has the advantage to increase one factor of imaging quality, but compromise the other. Therefore, trade-offs have to be made for optimum quality for each individual case, either the patient, or the equipment. Here I will focus on how scan time and SNR may be affected by bandwidth (BW), number of signal averages (NSA) and echo train length (ETL) (see the edited screenshot below).
1.Two types of bandwidth related to slice thickness, sampling time, SNR, and chemical shift artifact
Two types of bandwidth, transmit bandwidth and receive(r) bandwidth, are entirely distinct parameters (e.g. Lipton,2008; mr-tip.com). Once a certain gradient slope is applied, transmit bandwidth is a range of radiofrequencies (RFs), centering about the Larmor frequency, transmitted to excite the slice, which should match the difference in precessional frequency between two points. Transmit bandwidth is used to define the slice thickness as
Slice width = RF BW/ (γ*Gz).
γ: gyromagnetic ratio. Gz : magnetic gradient (slope).
A steep slice select slope and/or narrow transmit bandwidth is used to achieve thin slices, whereas, a shallow slice select slope or broad transmit bandwidth is used to achieve thick slices (see the edited screenshot; and Fig.1).
Fig.1. BW and slice thickness, from (Song, 2005).
The receive(r) bandwidth is the range of frequencies the receiver can sample accurately, which must be mapped across the FOV. This is determined by the upper and lower limits of frequencies on the either side of the center frequency of the echo. Based on the Nyquist theorem,
Receive bandwidth = 2 x the highest frequency (Nyquist frequency)
Alternatively (Allison,Wright &Lavin,2012),
Receive bandwidth = pixel bandwidth x number of phase encoding samples
A higher receive(r) bandwidth will give rise to faster sampling time, low minimum TE and reduce SNR due to the increase of noise (see the above screenshot), as SNR = Signal/Noise. On the other hand, it can be narrowed to block out the noise, and increases SNR (e.g. Li and Mirowitz, 2003). However, low bandwidth tends to cause chemical shift (misregistration) artifacts, caused by the different precessional frequency of fat and water. Because decreasing the bandwidth will lead to a smaller range of frequencies per voxel, and any displacement of signal due to the chemical shift effect will be extended beyond the voxel and manifest as a shift (pixel displacement) within the image.
2.Signal averaging parameters (NEX/NSA) increase both SNR (“good”) and scan time (“bad”).
The NEX (the number of excitations) or NSA (the number of signal averages or acquisitions) is an averaging parameter in imaging acquisition and reconstruction. NSA is the number of times each line in the K space is filled with data with the same slope of phase encoding gradient (see video by Magritek, 2012). The signal can be sampled more than once by maintaining the same slope of phase gradient over several TRs instead of changing it every TR. As there are more data in each line, the resultant image has a higher signal to noise ratio, canceling any “blurring”. The presence of random noise means that doubling the NEX only increases the SNR by square root of 2, i.e., 1.4, but, proportionally increases scan time (see video “MRI made easy”). For example, doubling the NEX doubles the scan time and vice versa. Therefore increasing the NEX is not the best way of increasing the SNR.
3.The echo train length (ETL) speeds up the scan time.
ETL is a unique parameter used in the fast spin echo sequence. ETL or the turbo factor is the number of 180o rephrasing pulses performed per TR (time to repetition) corresponding to the number of echoes produced and the number of lines of K space filled, i.e. different phase encoding for each echo. In the conventional spin echo, only one line is filled per TR, whereas several lines (i.e. ETL) in the K space are filled in fast spine echo, and K space is filled more quickly (see video by Magritek, 2012). The higher the turbo factor, the shorter the scan time, as more phase encoding steps are performed per TR. The scan time is reduced to the 1/ETL of the total scan time of a conventional spin echo. Additionally, the higher the ETL, the more of a mixture of weighting it is. For example, ETL of 2-4 achieves T1 weighting, whereas, ETL of 4-16 renders T2 weighting. Therefore, the long ETL has the disadvantage of increased image blurring and reduced number of slices/TR.
In summary, trade- offs or compromises among the “good” and the “bad” are required when all different factors and controlling parameters are concerned. For example, receive(r) bandwidth is a parameter that the technologist can control to adjust for optimum SNR and sampling time. Reducing receiver BW has the advantage of increasing SNR and the disadvantage of introducing chemical shift artifact. Similarly, increasing signal averaging (NEX, NSA) can increase SNR, but in the mean time proportionally increasing the scan time. ETL is a parameter for fast spin echo. Too much reduction of scan time with a long ETL will result in mixture of image weighting and blurring. Therefore, trade-offs are the key in consideration of setting these parameters for image optimization.
Allison, J., Wright, C., & Lavin, T. (2012). Mri physics course. Retrieved from www.radiology.georgiahealth.edu
Hoe, D. (Producer). (2009). Signal-to-noise-ratio. [Web Graphic]. Retrieved from http://www.imaios.com/en/e-Courses/e-MRI/Image-quality-and-artifacts/Signal-to-noise-ratio lipton, M. L. (2008). Totally accessible mri. New York,NY: Springer Science Business Media,LLC. Westbrook (2008), Handbook of MRI Technique.
Li, T. and Mirowitz,S.A. (2003). Fast T2-weighted fast imaging: impact of variations in pulse sequence parameters on image quality and artifacts. Magnetic Resonance Imaging. 21: 745-753. Retrieved from: http://ac.els-cdn.com/S0730725X03001735/1-s2.0-S0730725X03001735-main.pdf?_tid=9fbe81d38daebf27a1780773ee2622c6&acdnat=1332097262_dfc4e1ec1b2406cfd0cd649124abb767
Lipton, M. L. (2008). Totally accessible mri. New York,NY: Springer Science Business Media,LLC.
Magnetic Resonance-Technology Information Portal. (2012). Bandwidth. Retrived from: http://www.mr-tip.com/serv1.php?type=db1&dbs=Bandwidth
Magritek. (Producer). (2012). 2D mri. [Web Video]. Retrieved from http://www.magritek.com/support-videos
McRobbie, D. W., Moore, E. A., Graves, M. J., & Prince, M. R. (2010). Mri from picture to proton. New York, NY: Cambridge University Press.
Schering (producer). MRI made easy. [pdf]. MRI Made Easy
Song, A. W. (2005). Principles of mri: Image formation. Retrieved from www.biac.duke.edu
Westbrook, C. (2008). Handbook of mri technique. (3rd ed.). Oxford, United Kingdom: Wiley-Blackwell.
Westbrook, C., Roth, C. K., & Talbot, J. (2011). MRI in practice (4th ed.). Oxford, UK: Wiley-Blackwell.