Concepts of Parallel Imaging - Report Example

Parallel imaging Introduction Parallel imaging is a family of techniques that often take advantage of spatial information in phased-array of radiofrequency coils in reducing the acquisition times in the magnetic resonance imaging. Recent studies have indicated that in parallel imaging, the sampled k-space lines are often reduced by a factor equal or greater than two hence shortening acquisition times. Parallel imaging techniques were not commercially available until recently. They are on the verge of being explored in clinical applications. As has been widely cited, fundamentally, their potential clinical application involves either reduction in the acquisition time or improvement in spatial resolution. Improvements in the quality of images can be realized by reducing the single-shot spin echo sequences, and the fast spin-echo’s train length. Recent studies have hinted that parallel imaging is quite attractive for both vascular and cardiac application and proves more valuable as a 3-T body. Recent studies have shown that magnetic resonance imaging (MRI) can be devoted for establishing means of increasing the acquisition speed. It is worth noting, therefore, that impressive gains have been realized in an effort to make MRI more effective in its application. This paper seeks to provide an overview of fundamental parallel imaging concepts while illustrating on potential clinical applications. In this paper, merits, demerits of parallel imaging, as well as the comparison between SENSE and GRAPPA as parallel imaging technique would be emphasized. ADVANTAGES OF PARALLEL IMAGING IN EPI: Since 1990s, the techniques of parallel MRI acquisition have increasingly become powerful imaging methods with a characteristic faster acquisition process (Bammer & Schoenberg, 2004). It is noted that it has found many application due to attributed advantages. In this case, the Parallel Imaging is often used for purposes of reducing scanning time hence help in improvement of spatial resolution in some of the magnetic Resonance Imaging applications (Bammer,R., Schoenberg, 2004). Notably, the parallel imaging demand that data get acquired simultaneously and as also independently using a multi receiver coils with the coils maintaining spatial profiles that differ (Barkhausen, 2004). Moreover, it has an advantage of not being able to alter the contrast behavior of the imaging sequence underneath (Boesiger, 2002). Described as one with the ability to decrease the time required to perform the image sequence, it causes an increase in the resolution provided there is a specific time measured or be able to perform the two (Boesiger, 2002). For instance in cases where a patient experiences acquisition time exceeding his/her breath-hold capacity, The Parallel imaging can help in addressing this issue through reducing the patient acquisition time by factor 2 or even greater Figure 01. (Glockner et al. 2004) Figure 01. The Improved visualization of segmental renal arteries in SENSE IMAGES. IN case, the patient was initially short in breath with difficulty in suspending the respiration for a standard acquisition time. The use of Parallel imaging helped reduce acquisition time from 19s to 10 seconds (Glocker et al, 2004). For the Parallel images that are used in spiral scanning and EPI, they have a faster redouts, which often help to reduce the phase error that often result from BO motion or inhomogenity (Griswold, Jakob, 2002). Through this it help mitigate the T2* decay effect. In this case, motion effects, as well as T2 decay can be reduced and can be reduced when RF echo trains apply. Within the image product, such an advantage can lead to reduced susceptibility, motion artifact, as well as in mitigating of the T2/T2* blurring (Hahn eta l. 2003). In this case, Parallel imaging helps in reducing motion artifact, alongside diminishing the venous contamination, in particular for regions in which there happens to be rapid venous return. For instance, in carotid and renal arteries (figure 02) (Glockner et al. 2004) Figure 02.  The images of partial volume and maximum intensity projection for the renal Magnetic Resonance angiography for a patient having fibro muscular dysplasia when obtained with and without SENSE. Give the STD image the early filling of left renal vein limited the left renal artery visualization. In this case, Venous contamination is reduced in the image of SENSE, obtained with a acquisition time, which was relatively reduced by half. The ccollecting system activity indicated in the SENSE image was as a result of eexcretion of the contrast material in that first non-SENSE acquisition (Glocker et al, 2005). It is also worth noting that parallel acquisitionist often help in reducing the SAR (Kassner et al. 2000). In this regard, a High RF power deposition becomes limiting whenever much of the RF pulses are needed for purposes of completing any scan for a given time. This problem is often addressed in an effective manner through the reduction of the sampling K-space density causing a reduction on the number of the RF pulses for a unit time (Kassner et al. 2000). Last but not least, parallel imaging is one of the most effective ways through which acoustic noises that is caused by the gradient switching is mitigated (Weishaupt eta l, 2004). By enabling the using of sparser k-space pattern, the approach of parallel imaging provides an allowance for the reduction in speed in the k-space. This, therefore, makes the gradient switching rate to be at 48 (Weishaupt eta l, 2004). As demonstrated above, parallel imaging has multiple benefits, which are on most occasions accompanied by a key downside, which has been common for most of the parallel techniques (Weishaupt et a l, 2004). In a likely manner as described above, Parallel imaging finds its application in many fields. One most significant area of application is in the field of angiography. As often is the case, angiographic scans are based upon the three-dimensional acquisition (Parizel et al. 2003). It strongly relies on certain contrast mechanism, for instance the contrast enhancement/ phase contrast, or inflow through exogenous agents. For the exams that demand for the patient to hold his/her breath reduction in the scan time become very important. This is one case in which the imaging of the thorax, heart and abdomen become applicable (Parizel et al. 2003). On the other hand, when dealing with cardiac imaging, breath-holding is always the limiting factor considering that the ability of doing so is, more often than not reduced, especially for the cardiac-disease affected patient. In this particular case, scan time reduction through parallel acquisition is very promising. Disadvantages of parallel imaging. The outstanding limitation often associated with parallel imaging, entails its characteristic associative signal-to-noise ratio, along with the reconstruction artifacts. From the fact that parallel magnetic resonance imaging or PMRI, can be used in many areas, it is clear that it is possible to have the speed of the MR imaging increased significantly. It has been observed that in several applications, a reduction in the acquisition times has been of great importance. For instance, Breathhold scans which makes it possible to study the dynamic heart is one of those that have benefited from such accelerated examinations. Currently, several PMRI methods are now available. As widely noted, making a choice to settle on an optimal method might not and be straight forward considering the fact each of the methods has its associated disadvantages and advantages. Auto-SENSE and GRAPPA There are two most frequently utilized PMRI methods: GRAPPA and SENSE. It is not as such quite easy to compare these two methods given pros and cons often associated with each of the two methods. For instance, in those images that are often reconstructed with the SENSE technique, signal variations arising from the individual coil sensitivities are always compensated for. This tends to give an impression that coil sensitivity is uniform. As such AUTO SENSE technique is an Image-based reconstruction, which is somewhat easier to understand. Reconstructed image can show aliasing in case date is acquired using this technique since it is fewer phases encoded. This is because less phase can encode in a similar region of the k-spacea somewhat smaller FOV. On the other hand, GRAPPA method, is a k-space-based reconstruction technique. In this technique, there is an addition of a small number of the k-space during acquisition. This helps in eliminating the need for a separate coil acquisition. As a matter of fact, the GRAPPA reconstruction algorithm provides improvements in the SNR and elimination of some artifacts. Research has shown that GRAPPA is somewhat advantageous because it provides the possibilities of performing accelerated image acquisitions using a field of view that is somewhat smaller relative to the object. What is common of these two methods is that both of them have the capability of reconstructing images from the MRI sub-sampled multimedia data. However, these two methods have outstanding differences. Research has indicated that SENSE can reconstruct an excited spin- density image directly. On the other hand, GRAPPA can be used for reconstructing estimates of a fully sampled data of a coil combining them for purposes of obtaining an image. It has also been shown that whereas SENSE requires an estimation of a coil sensitivity map before the reconstruction process, GRAPPA can be used for improvement of the sensitivity estimates of the coil. In the likely manner, with the SENSE reconstruction equations along with the sensitivity estimates of the coil, it becomes possible to make an improvement on the estimates parameters of GRAPPA reconstruction. Hence using the two approaches, one can get a higher image quality as opposed to using either of the methods alone. The appearance of the artifacts where the sense reconstruction is performed in the image domain on a pixel by pixel is different for GRAPPA and SENSE. In this case, non ideal conditions in the reconstruction process leads to local noise enhancement. In most cases, it appears localized in a folded image. On the contrary, GRAPPA algorithm generates the missing K-space lines in which an inaccurate calculation of the missing lines produces aliasing artifacts in the reconstructed image which can be seen entirely over the reconstructed image FOV. (Blaimer et al. 2004) Figure 1 shows a reduction in the number of phase encoding steps per acquisition time. From the figure above top images are produced under normal acquisition while the bottom has R=2 acceleration (Kassner et al., 2000). Comparison of artifacts in sense and Grappa reconstructions at high accelerations. (Blaimer et al. 2004) Figure (2): A comparison of the artifacts GRAPPA and SENSE reconstruction for each acceleration. In this specific example, it was possible to achieve a fourfold scan time reduction by using four coils. In A, The image from the sense technique indicate that there was a local enhancement which was as a result of the non-ideal conditioning for reconstruction. On the other hand, in B, in GRAPPA, the noise enhancement is somewhat distributed in a more evenly manner over FOV (Kassner et al., 2000). It should also be noted that SENSE provides a slightly higher image quality when used with high acceleration factors (Boesiger, 2002). GRAPPA is preferred in situations where accurate sensitivity maps are difficult to obtain, such as the lungs since it provides a more robust reconstruction (Griswold & Jakob, 2002. Auto calibration technique may also be preferred in situations where patients have a difficulty in respiration a reproducible manner thus leading to discrepancies between the parallel image acquisition and the calibration acquisition which in turn results to reconstruction artifacts (Wintersperger, et al. 2003). However, the acquisition time must be within the patient’s breath hold capacity. The choice of GRAPPA over SENSE also depends on the FOV. Where, in the SENSE like image based techniques when full FOV becomes small as compared to object, formation of artifacts occurs in the reconstructed image (Huber et al. 2004). To overcome this, aliasing is avoided in the reconstructed image. On the contrary, GRAPPA can generate partial aliased reconstructed image having no artifact to allow for the selection of a smaller FOV in situations where spatial resolution is a key consideration (Huberet al. 2004). Moreover, for the SENSE method, the scan time is directly related with the number of the encoding phase steps and subsequent incorporation of multiple coils makes it easier to reduce the sampling density. This can be illustrated as shown below (Blaimer et al. 2004) Figure (3) From figure (3), it can be observed that when the number of the phase encoding steps is reduced in the k-space view alongside a Cartesian sampling of the k space, the same k space area is collected. However, there is a considerable increase in the distance between the phase encoding lines. It is also worth noting that reduced sampling density implies that Nyqvist criterion is not satisfied, hence the use of the Fourier transformed image (Glocker et al, 2005). The figure below illustrates the relationship between sampling and image matrix, (Blaimer et al. 2004) Figure 4 above illustrates the SENSE technique used to reduce the scan time without compromising the image quality The figure below illustrates Image based reconstruction based on GRAPPA and SENSE techniques. Image source (Blaimer et al. 2004). The figure 5 above shows comparison of the image quality of (a) SENSE and (b) GRAPPA reconstructions having a reduction factor of 3. In (a) it can be observed that, with accurate coil sensitivity maps, the SENSE reconstruction obtains the best possible result with an optimized SNR (Itskovich et al. 2004). On the other hand, the Grappa reconstruction is fundamentally an approximation to SENSE reconstruction. However, on the visual scale, there is no different between the two. By and large, the SENSE technique is widely used for clinical application, and most proffered than the GRAPPA technique. This is attributed to its ability to produce higher image reconstruction. For instance, in cardiac imaging reduction in scan time, the SENSE relaxes the requirements for breath hold studies (Kuhl et al., 2003). The gain in scan time can also be used for purposes of improving spatial resolution, reducing imaging time and real time cardiac imaging. SENSE is also applied in the contrast enhanced magnetic resonance angiography. This way, it allows a higher spatial resolution at a constant time (Hu, 2004). However, image reconstruction of a single shot EPI is somewhat problematic considering that in the EPI images and sensitivity maps differ (Weiger, 2000). As noted Grappa employs the regenerative k-space technique, hence it is advantageous in the lung and abdominal MRI, and real time imaging (Huber, et al., 2004). GRAPPA technique is particularly employed in areas where the coil sensitivity maps are difficult to obtain, especially in homogenous areas such as the lungs and the abdomen. In such areas determination of precise spatial coil sensitivity is difficult. As such, algorithm provides a good quality image reconstruction in the absence of the sensitivity maps. In addition, the k space lines are fit to calculate the reconstruction parameters which involve procedure concerning the global information. It is, therefore, not affected by localized in homogeneities (Lederman, 2003). It can also be noted that k-sspace lines ensures that accurate information is available to realize good reconstruction quality. The figure below illustrates this scenario Image source (Blaimer et al. 2004) Figure 6 . The single- shot HASTE Imaging for the lung when combined with the GRAPPA, A It is a conventional Haste image having a matrix size of 256 X 128 and was acquired in the 220ms at effective TE equal to 23ms, FOV equal to 500mm X 500mm. In this image, there were 72 echoes acquired. In B, the GRAPPA acquisition having acceleration factor of 3. When compared with that reference image, it is found that the resolution is doubled with an acquisition time being reduced from the value of 220 ms to a value of 161 ms (Kassner et al., 2000). GRAPPA technique can also be used in a single shot EPI since the k space based reconstruction of the missing lines is not affected by distortions in the image (Barkhausen, 2004). Generally, GRAPPA has shown robust reconstruction without modification of the EPI or the reconstruction algorithm as shown below. (Blaimer et al. 2004). Figure 7 (a) conventional EPI (b) employs Grappa which shows reduced distortions due to the reduced interecho spacing. Grappa is unaffected by small FOV thus allows an optimal FOV acceleration for a given application (Glocker et al, 2005). In many applications only a small region within the object is relevant for a diagnosis. This implies that by choosing a smaller FOV than the object, the irrelevant parts of the image would have its imaging speed increased even in the absence of parallel imaging. However, full FOV images may result into erroneous coil sensitivity, leading to image artifacts after reconstruction. Following that the SENSE method has to use a larger FOV to prevent the aliasing of tissue from the outside. On the other hand, GRAPPA technique has the capability to generate partially aliased image construction having the same appearance as in the conventional imaging without modification of the reconstruction algorithm. (Blaimer et al. 2004) Figure 8: Image acquisition in full FOV. In (a) Artifacts can be seen this can be due to erroneous coil sensitivity maps. (b) GRAPPA to counter the problem the FOV has to be larger than the object thus preventing aliasing of tissue from outside FOV. Today, the choice of the optimal parallel imaging techniques is limited by the commercial availability of the said methods. The choice between the image domain, SENSE and the self calibrating k-space technique depends upon a number of factors. For instance, if the speed is of importance, then having an extra self calibrating k space may seem to be disadvantageous and thus SENSE stands out as the method to choose. Notably, it is a concern for low resolution images because of the reduction in the K space lines when the image matrix increases. In conclusion, parallel imaging comprises of a number of techniques. The technique is characterized by a reduced number of k-space lines and hence reduced acquisition time. As noted, the methods have a wider application in the clinical, with both having reduced acquisition time and improved spatial resolution. Table 01: Showing a comparison of the GRAPPA and AUTOSENSE GRAPPA AUTOSENSE It is k-space-based reconstruction technique, which is somewhat harder to understand. It is an Image-based reconstruction, which is somewhat easier to understand. Grappa reconstruction is an approximation to SENSE reconstruction but still on the visual scale no differences can be observed. With accurate coil sensitivity maps, the SENSE reconstruction obtains the best possible result with an optimized SNR Relatively lower image reconstruction capabilities Higher image reconstruction capabilities Addition of a small number of the k-space during acquisition. No addition of number of the k-space during acquisition. Estimates of a fully sampled data of a coil combining them for purposes of obtaining an image and cannot reconstruct an excited spin- density image directly Can reconstruct an excited spin- density image directly Preferred in situations where accurate sensitivity maps are difficult to obtain, such as the lungs since it provides a more robust reconstruction Image reconstruction of a single shot EPI is problematic due to in the EPI images and different sensitivity maps References Blaimer et al. 2004. SMASH, SENSE, PILS, GRAPPA. How to Choose the Optimal Method [Photograph]. Retrieved on 24th April 2013 from http://ee-classes.usc.edu/ee591/library/Blaimer-SMASH_SENSE_PILS_GRAPPA.pdf . Retrieved on 24th April 2013. Barkhausen, J. 2004. The Parallel Acquisition Techniques In Cardiac: Comparison of image quality and the artifacts. Magnetic Resonance; 20: 506–511. Boesiger, P, 2002. The 2D SENSE for Quicker 3D MRI. MAGMA; 14: 10–19 Griswold, M., Jakob, M, 2002. The Generalized Auto Calibrating Of Partially Parallel Acquisitions. Magnetic Resonance; 47: 1202–1210. Glockner et al. 2004. Parallel MR Imaging: A User’s Guide. RadioGraphics. The Journal of continuing medical education in radiology [photograph]. Retrieved on 24th April 2013 from http://intl-radiographics.rsna.org/content/25/5/1279.full. Retrieved on 24th April 2013. Hahn et al. 2003. Human Heat Auto Sense Profusion. Magnetic Resonance; 18(6): 702–708. Huber et al. 2004. The High Resolution MR Imaging Of The Liver: Comparison Of The Prospective Motion Correction And Respiratory Triggering. Magnetic Resonance; 20: 443–450. Hu, HH, 2004. Improved Venous Suppression And Spatial Resolution With SENSE In Elliptical Centric 3D Contrast-Enhanced MR Angiography. Magnetic Resonance; 52: 761–765. Itskovich,VV, et al. 2004. Parallel And Nonparallel Simultaneous Inversion Recovery Techniques For Vessel Wall Imaging. Magnetic Resonance Imaging; 19(4): 459–467 Kassner et al. 2000. The Contrast-Enhanced 3D MRA Using SENSE. J Magnetic Resonance; 12: 671–677 Kuhl et al. 2003. SENSE MR in The Routine Clinical Practices. Euro J Radiol2003; 46: 3–27. Lederman, RJ, 2003. The Real-Time Accelerated Interactive MRI. Magnetic Resonance Med; 50: 315–321. Parizel, PM, et al. 2003. Parallel Magnetic Resonance Imaging. Euru Radiol; 13: 2323–2327. Pruessmann., KP, 2004. Parallel Imaging At Large Field Strengths: The Synergies And Joint Potential. Magnetic Resonance; 15(4): 237–244. Pruessmann, KP., Weiger, M, 2001. The Sensitivity Encoded Cardiac MRI. Cardiovasc Magnetic Resonance; 3: 1–9. Scheidegger, MB., Besieger, P, 1999. Sensitivity Encoding For Fast MRI. Magnetic Resonance Med; 42: 952–962. Sodickson, D, 1997. Acquisition Of Spatial Harmonics: The Fast Imaging With Radiofrequency Coil Arrays. Magnetic Resonance; 38: 591–603 Maki, JH, eta l. 2004. Parallel Imaging in MR Angiography. Magnetic Resonance Imaging; 15(3): 169–185. Weiger, M., 2000. Real Time Imaging Of Cardiac Using Sense. Magnetic Resonance; 43: 177–18. Weishaupt, D, eta l, 2004. The Sensitivity Encoding For Enhancing SNR Efficiency in Steady-State MRI. Magnetic Resonance; 53: 177–185. Wintersperger, BJ, et al. 2003.Breath-Hold Real-Time Cines MR Imaging: An improved temporal resolution using GRAPPA. Euro Radiol; 13(8): 1931–1936. Read More