Factors affecting quality of MRI image.Benefits of circular or square spiral EPI methods - Essay Example

?Introduction Factors affecting quality of MRI image Accurate image reconstruction depends on resonance frequency (rf) of a spin, which is achieved by strong homogenous external field and superposition of a spatially linear gradient. If these conditions are not met, then the relative positions of anatomical structures will be misrepresented in the reconstructed MRI image. For example, if a spin is exposed to a magnetic field different from the expected value due to magnetic susceptibility, geometric distortions will be seen from the MRI image (Reinsberg, Doran, Charles-Edwards, and Leach, 2005). In particular, an increased spin produces MRI signals of low intensity, while high magnetic susceptibility areas are seen as areas of total signal loss (Zhou and Gullapalli, 2006). Figure 1 shows an example of metallic susceptibility distortions in the presence of metallic implants. Magnetic susceptibility In magnetic resonance imaging, tissues are temporarily magnetized by the large magnetic field that the imaged subject is exposed to, with the extent of magnetization varying among tissues that differ in magnetic susceptibility. For example, relative to other tissues, air and bone are less susceptible to magnetization. On the other hand, metallic implants have high magnetic susceptibility as compared to the body tissues. These differences in magnetic susceptibility cause field inhomogeneity, especially in tissue boundaries, particularly air-tissue and bone-soft tissue boundaries (Zhou and Gullapalli, 2006). That is why magnetic susceptibility distortions are sometimes seen in such boundaries, like the sinuses (Reinsberg, Doran, Charles-Edwards, and Leach, 2005). 1. Discuss various methods to correct for magnetic susceptibility distortions, while maintaining a good EPI image. EPI is highly susceptible to resonance distortions, especially since it uses more time for gradient refocusing (Zhou and Gullapalli, 2006). Basically, magnetic susceptibility distortions are reduced by preventing a significant amount of phase errors to accumulate. Without the corrective rf pulses, this can only be achieved by segmentation, reducing the echo spacing (ESP), faster data gathering, anterior-posterior phase encoding. A faster ESP implies a shorter time for magnetic susceptibility effects to develop. a.) Parameters that can optimize EPI image ESP can be decreased through the use of powerful gradients, such as increased gradient slew rate (~20, 000 T/m/s), increased maximum amplitude (350 mT/m), increased rise time rate, increased accuracy, decreased eddy currents, increased voltage, and decreased field strength. In addition, these parameters decrease signal-to-noise ratio. Meanwhile, a decreased ESP results to a smaller total train length time. In fact, it has been estimated that decreasing ESP to 40 ms limits the distortion to only one pixel wide. Similarly, increased data gathering via decreased frequency- and phase encoding steps and ramp sampling, fast analogue to digital converters and large receiver bandwidth decrease the time in which phase errors can accumulate (McMahon, 2012). On the other hand, phase and read directions can be swapped, so that the phase encoding gradient is along the same axis as susceptibility gradients. Although this may not totally prevent distortion, it shows the image in a more presentable and understandable manner (McMahon, 2012; Zhou and Gullapalli, 2006). Figure 2 shows the decrease in magnetic susceptibility upon converting from left-right phase encoding to anterior-posterior phase encoding. Other ways to minimize magnetic susceptibility distortions include decreasing the resolution, use of SE sequence, increased acquisition matrix and proper shimming before image acquisition. It should be noted, however, that heating of nearby tissues can occur when using fast SE acquisition with a high bandwidth to decrease magnetic susceptibility distortions resulting from metallic implants (Zhou and Gullapalli, 2006). b) Chain of consequences after increasing the receiver bandwidth Receiver bandwidth is a value that describes the amount of time needed to digitize the image. It is characterized by the equation 1/t, where t is the time it takes for a data to be sampled (figure 3b). However, the rf must not be too fast for the imaging process to miss out obtaining all samples necessary to make a good quality image. As such, it must obey the following conditions, RBW = y x Gx x FOV (McMahon, 2012). As can be seen in figure 3a, increased receiver bandwidth (RBW) can decrease ESP and signal-to-noise ratio and decreasing the magnetic susceptibility distortions. In addition, the increase in RBW will result to increased frequency that can be sampled during the imaging, resulting to an ample amount of data gathered and subsequent decrease in aliasing. These data can also be detected quite well during the read out phase. Sampling time will also be decreased in greater RBW, as depicted in figure 3b. All in all, increasing RBW can increase the quality of the image, although it will make it susceptible to chemical shifting (McMahon, 2012). 2. Benefits of circular or square spiral EPI methods Circular and square spiral EPI methods both use sinusoidal waveforms driven by resonant circuits, which result to a faster gradient response. In that case, gradients can be applied smoothly, lowering the risk of peripheral nerve stimulation and decreasing eddy currents (McMahon, 2012). Spirals are specifically used to decrease the time needed to cover the entire k-space. This method also requires spectral interleaves to maintain ample resolution. On the other hand, circles are more efficient samplers, and it requires spatial interleaves instead (Furuyama, Wilson and Thomas, 2012). The Spiral EPI, two-dimensional spiral in particular, is also less sensitive to physiological noise (Sangill, Wallentin, Ostergaard, and Vestergaard-Poulsen, 2006). In addition, the self-refocusing spiral acquisition patterns can decrease blurring resulting from velocity-dependent phase changes, also known as motion artifact (McMahon, 2012). What it does is to spread the blurring over a larger area to dilute the artifact in neighboring pixels (Sangill, Wallentin, Ostergaard, and Vestergaard-Poulsen, 2006). Spiral EPI is thus suitable for imaging the blood vessels and brain motion (McMahon, 2012; Sangill, Wallentin, Ostergaard, and Vestergaard-Poulsen, 2006). Spiral EPI methods in single-shot EPI, together with lower expected resolution and high capabilities of gradient systems, also decreases the time needed to produce an image to 60-150 ms. Partly, this is because the spiral trajectory only uses 58-61 ms to cover the entire k-space in one segment or in a few segments, as can be seen in figures 4 and 5 (Sangill, Wallentin, Ostergaard, and Vestergaard-Poulsen, 2006; Voiron and Lamalle, 2006). Also, it limits the image displacement to at most 0.6 mm (Sangill, Wallentin, Ostergaard, and Vestergaard-Poulsen, 2006). 3. Advantages and disadvantages of using segmentation Image segmentation is the division of an image into non-overlapping components that are homogeneous with respect to intensity or texture. This is done by dividing the phase-encoding steps into multiple shots, which are the number of samplings required to cover the whole k-space, depicted in figure 7. The resulting value is the echo-train length, such that 256 phase encoding steps sampled in 32 shots have an echo-train length of 256/32, or 8 (McMahon, 2012). It is done to allow the delineation of anatomical structures of interest, making it possible to quantify tissue volumes, to diagnose and localize pathology, and to plan treatment and surgery (Pham, Xu and Prince, 2000). Compared to single shot EPI, segmentation puts less stress to gradients, making the latter functional in conventional systems. In addition, phase errors do not build up in segmentation as much as they do in single shot EPI, thus minimizing the magnetic susceptibility artifacts in the image. As a result, segmentation decreases geometric distortions, as well as provides better resolution and T1 weighting (McMahon, 2012), as can be seen in figure 6.In fact, it reduces geometric distortions by a factor proportional to the blind width (Holdsworth et al., 2011). More importantly, if accuracy and precision are still subpar, then segmented atlases provide valuable information to improve segmentation (Pham, Xu and Price, 2000). Figure 5 shows the increased resolution of segmented MRI image. These, however, come with a price. Segmentation takes longer to accomplish. Thus, despite being able to minimize magnetic susceptibility artifacts, motion artifacts become more likely when using multi-shot EPI (McMahon, 2012). This is the reason why increasing computational efficiency, either by multi-scale processing or parallel imaging is considered very important in making segmentation more feasible in practical settings. However, the amount of manual interactions demanded to optimize segmented images make the methodology susceptible to reliability issues (Pham, Xu and Price, 2000). Other artifacts may also be evident because of the complex coverage of k space. In particular, discontinuation of k-space may cause ghosting, which may also occur when varying signal amplitude (McMahon, 2012). The Mosaic EPI, in gact, is not commercially available because of the magnetic susceptibility, chemical shifts and nyquist ghosts its images are considerably vulnerable to (Schmitt, Stehling and Turner, 1998). 4. Difference between SMASH and SENSE reconstruction in parallel imaging Parallel MRI (pMRI) To prevent image wrapping resulting from the under-sampling of the field of interest, two receiver coils can be used simultaneously to get as much samples as possible (McMahon, 2012). In addition, parallel MRI (pMRI) also decrease the time needed for MR measurement time during therapeutic MRI to as much as half (Mueller et al., 2004) by using R number of coil arrays with various coil sensitivities sampling a single field of view (FOV), and effectively decreasing the phase-encoding steps. By doing so, the signal-to-noise ratio (SNR) is also significantly reduced by a factor of (Figure 8). However, these improvements in imaging require multiple receiver pathways and ample knowledge of coil sensitivities and coil positions (Blaimer et al., 2004). Simultaneous acquisition of spatial harmonics (SMASH) and sensitivity encoding (SENSE) reconstruction are just some of the ways by which parallel imaging can be done. SMASH The first successfully implemented parallel imaging technique, the SMASH algorithm, developed in 1997, uses k-space raw data for imaging, with missing phase-encoding steps derived from known individual coil sensitivities (Blaimer et al., 2004; McMahon, 2012; Mueller et al., 2004) using convolution and phase-encoding magnetic field gradients (Blaimer et al., 2004; Kellman, 2004). In particular, the linear combination of measured sensitivity values and linear weights are combined to calculate composite sensitivity profiles. The consolidation of measured and estimated k-space data is then used to produce the image (Figure 10). This algorithm needs to use coil configurations that are optimally designed to accurately generate the ideal spatial harmonics in phase-encoding direction. If spatial harmonics are inaccurate, residual artifacts resulting from phase cancellations and eventual unintended signal losses arise in the image, compromising the quality of the MRI image (Blaimer et al., 2004). Since its development, it has been used for harmonic fits, coil-by-coil image reconstruction and generalized matrix formulation (Kellman, 2004). The theory behind SMASH is also currently being practiced through Generalized Autocalibrating Partially Parallel Acquisitions, or GRAPPA, which is a much improved and more commonly used k-space-based algorithm, eliminating disadvantages such as poor reconstruction quality, suboptimal fit procedure and limitations in coil configurations that were evident in the pure SMASH imaging (Blaimer et al., 2004). Its effects on image quality are clearly elucidated in figure 9. SENSE The most commonly used parallel imaging technique in clinical setting, SENSE, or unfolding algorithm, uses the image domain data from a previously completed database to process the image (Blaimer et al., 2004; McMahon, 2012). As seen in figure 10, it is thus necessary to complete the sampling of image data from k-space data before these can be consolidated (Mueller et al., 2004). To do this, the algorithm deriving the image from k-space data must be repeated for every pixel location in the reduced FOV image to reconstruct the full FOV. Therefore, it is important that the number of pixels to be separated must not be greater than the number of coil arrays used in parallel algorithm. By following these conditions, the presence of artifacts in the resulting MRI image may be prevented (Blaimer et al., 2004). SENSE is a more advantageous parallel imaging technique because it can be performed using arbitrary coil configurations. It is not limited to linear coil configurations or localized sensitivities. However, using non-linear coil configurations reduce SNR substantially. In addition, the coils’ encoding efficiency, as indicated by the geometry factor (g factor), decreases the SNR in SENSE algorithm (Blaimer et al., 2004). Since its development, the use of SENSE has already been extended to two-dimensional SENSE and multi-slice parallel acquisitions (Kellman, 2004). SMASH vs. SENSE With the improvements both SMASH and SENSE have undergone throughout the years, the quality of images they can produce is essentially similar (figure 11). However, in cases wherein the inaccurate sensitivity maps of SENSE compromise the image quality, especially in inhomogeneous regions such as lung and abdomen, the use of k-space-based parallel imaging such as SMASH or GRAPPA is much more suitable. In addition, k-space lines are not influenced by the smaller FOV used in reconstruction. Single-shot EPI is also better paired with GRAPPA, than with SENSE. In any other cases, especially when an accurate coil sensitivity map can be obtained, the SENSE algorithm must be used because it provides the best possible reconstruction, SNR and image quality as compared to other parallel MRI imaging techniques (Blaimer et al., 2004). References Blaimer, M. et al. 2004. SMASH, SENSE, PILS, GRAPPA. Top Magn Reson Imaging, 15(4), pp. 223-236. Furuyama, J. K., Wilson, N. E. and Thomas, M. A. 2012. Spectroscopic Imaging Using Concentrically Circular Echo-Planar Trajectories in vivo. Magnetic Resonance in Medicine, 67, 1515-1522. Holdsworth, S. J., et al. 2011. Clinical Application of Readout-Segmented_ Echo-Planar Imaging for Diffusion-Weighted Imaging in Pediatric Brain. American Journal of Neuroradiology, 32, pp. 1274-1279. Kellman, P. 2004. Parallel Imaging: The Basics. 12th Annual ISMRM Meeting. Kyoto, Japan. 16 May 2004. McMahon, K. 2012. MRES7005. Fast Imaging Techniques. Queensland, Australia: University of Queensland.   Mueller, S. et al. 2004. Real-Time Parallel Image Reconstruction for Interventional MRI. Second International Workshop on Parallel MRI: Latsis Symposium 2004. Zurich, Switzerland. 15-17 October 2004. Pham, D. L., Xu, C. and Prince, J. L. 2000. Current Methods in Medical Image Segmentation. Annu Rev Biomed Eng, 2, pp. 315-337. Reinsberg, S. A., Doran, S. J., Charles-Edwards, E. M., and Leach, M. O. 2005. A complete distortion correction for MR images: II. Rectification of static-field inhomogeneities by similarity-based profile mapping. Physics in Medicine and Biology, 50, pp. 2651-2661. Sangill, R., Wallentin, M., Ostergaard, L., and Vestergaard-Poulsen, P. 2006. The impact of susceptibility gradients on cartesian and spiral EPI for BOLD fMRI. Magn Reson Mater Phys, 19, pp. 105-114. Schmitt, F., Stehling, M. K. and Turner, R. 1998. Echo-Planar Imaging: Theory, Technique and Application. New York: Springer. Voiron, J. and Lamalle, L. 2006. Spiral MRI: Principles and in vivo Applications at High Field. SpinReport. 157/158, pp. 9-17 Zhuo, J. and Gullapalli, R. P. 2006. MR Artifacts, Safety, and Quality Control. RadioGraphics, 26(1), pp. 275-297. Read More